key: cord-344020-8poerd09 authors: vermeulen, tom d; reimerink, johan; reusken, chantal; giron, sandra; de vries, peter j title: autochthonous dengue in two dutch tourists visiting département var, southern france, july 2020 date: 2020-10-01 journal: euro surveill doi: 10.2807/1560-7917.es.2020.25.39.2001670 sha: doc_id: 344020 cord_uid: 8poerd09 we report dengue virus (denv) infection in two dutch tourists who visited département var, southern france, in july and august 2020. as some autochthonous dengue cases have occurred in europe in recent years, awareness among physicians and public health experts about possible intermittent presence of denv in southern europe is important to minimise delay in diagnosis and treatment. quick diagnosis can lead to timely action to contain the spread of vector-borne diseases and minimise transmission. we report dengue virus (denv) infection in two dutch tourists who visited département var, southern france, in july and august 2020. as some autochthonous dengue cases have occurred in europe in recent years, awareness among physicians and public health experts about possible intermittent presence of denv in southern europe is important to minimise delay in diagnosis and treatment. quick diagnosis can lead to timely action to contain the spread of vector-borne diseases and minimise transmission. travel-related diseases may serve as sentinels of transmission of disease in the visited area. prompt diagnosis and notification of such diseases may assist in the detection and control of disease outbreaks. when we diagnosed dengue in a dutch tourist who visited southern france, we coordinated joint action between the patient and clinical and public health experts. this led to rapid international notification and consecutive outbreak control efforts by french authorities. a second, related, dutch patient with a recent fever was retrospectively also diagnosed with dengue. a previously healthy woman in her 20s (patient 1) spent a 2-week holiday with relatives in la croix valmer, france, from 13 to 31 july 2020. in the first week of august, patient 1 stayed with other friends in another house nearby. on 1 august, she developed fever accompanied by myalgia in her calves and neck, as well as a painful skin (day 1 of the disease episode). on day 5 post onset of symptoms (pos), she was nauseous and vomited once. on day 6 pos, she returned to the netherlands. a test for severe acute respiratory syndrome coronavirus 2 (sars-cov-2), on a nasopharyngeal swab, was negative. on day 8 pos, the patient noticed an itchy erythematous rash on her hands and lower legs. on day 11 pos, she consulted her general practitioner who, suspecting petechiae, referred her to our hospital. in addition to the reported signs and symptoms, she mentioned a blurred, colourful spot in the field of vision of her left eye. during childhood, the patient was vaccinated according to the dutch national vaccination scheme; she had never received any vaccinations for yellow fever, japanese encephalitis or tick-borne encephalitis. interestingly, one of her family members upon our clinical suspicion of dengue virus infection spontaneously acknowledged having seen tiger mosquitoes (aedes albopictus) around their holiday home. physical examination showed normal vital signs and revealed slight erythematous exanthema on her hands and upper limbs and a confluent petechiae-like exanthema on both legs. the presumptive diagnosis of dengue was made, common laboratory tests including dengue virus (denv) serology were ordered and she was referred to the ophthalmologist. fluorescein angiography of the eyes showed an inflammatory foveolitis in her left eye. laboratory results on day 12 pos showed a mild thrombocytopenia and leukocytopenia with plasmocytosis, and moderately elevated serum levels of the liver enzymes (table 1 ). on the same serum, comparative igm and igg serology (immunofluorescence arbovirus fever mosaic 1, euroimmun ag, lübeck, germany) against chikungunya virus (chikv), denv and japanese encephalitis virus (jev) was performed at the dutch national institute for public health and the environment (rivm) laboratory. high concentrations of denv-specific igm and igg antibodies were detected ( table 2 ). there was a slight igg response, without igm response, against jev, interpreted as non-specific cross-reactivity. rt-pcr was not done. patient 2 spent the holiday in la croix valmer, together with patient 1, from 13 to 31 july. he returned to the netherlands 1 week before patient 1. on 1 august, he noticed a mild pain in his right ear (day 1 of the disease episode). on day 2 pos he had a high fever (40.5° c). on day 4 pos, suspecting bacterial otitis, he was prescribed amoxicillin. on day 5 pos, he noticed a mild pain behind his eyes. on day 9 pos, 1 day after defervescence, he noticed a slight rash on his trunk and extremities and interpreted this as allergy to the amoxicillin. patient 2's blood sample of 7 september (day 37 pos) tested positive for igm and igg antibodies against denv, with high titres ( table 2 ). the igg response against jev was interpreted as non-specific cross-reactivity. also this patient had been vaccinated according to the dutch national vaccination scheme. on august 27, upon confirmation of the serological results, patient 1 was reported by the rivm to the french authorities through the early warning and response system of the european union as an autochthonous denv infection probably acquired in france with cross-border implication. the french authorities contacted the patient and announced the case by press release on 8 september 2020 [1] . santé publique france advised that also other family members with symptoms should be tested. that identified patient 2. the other seven dutch family members visiting the holiday home between 13 and 31 july did not develop any disease symptoms and neither did the individuals with whom patient 1 stayed during the first week of august in the other house nearby. none of the household members had recently travelled outside europe. notification to the french authorities led to a prompt local public health response including mosquito control around the holiday home (with deltametrine and bacillus thuringiensis israelensis on a 150 m radius) and door-to-door investigations in order to identify other cases and raising awareness among local healthcare professionals and the public [1] . here we describe two patients with dengue from the same family, who acquired the disease in department var, southern france. the signs and symptoms, as well as the plasmocytosis, of patient 1 were typical for dengue [2] . the list of differential diagnoses was therefore very short and the high igm and igg titres for denv were considered confirmative, even though definite confirmation would require demonstration of virus or serodiagnosis on paired samples [3] . pcr was not conducted to detect denv in blood or urine because the chance for a positive test was considered low in this rather late stage of disease and it was not deemed necessary for confirmation of disease, clinical management, notification or public health measures. dengue is endemic in large parts of the world and a common illness among returning travellers from (sub) tropical regions [4] . because of globalisation in travel and trade and under changing ecological conditions, the geographical distribution of the vector of denv, ae. albopictus, gradually expanded over the last decades and may continue to do so [5] . imported infections can continue to cause autochthonous outbreaks. a lack of awareness and a long interval between the viraemic episode of the patient with imported dengue and the first registration in the public health system were identified as possible drivers of local outbreaks [6] . aedes albopictus has been established in france since 2004 and currently, the mosquito species is endemic in large parts of the country including one area close to the belgian border [7, 8] . the presence of ae. albopictus in multiple sites in europe means that also other diseases can be transmitted. upon introduction by returning viraemic travellers, european cases of chikv, denv and zika virus infection have been reported [9, 10] . as recently as august 2020, five patients in vicenza province, northern italy, were confirmed to have a denv infection 2 weeks after a household member infected with denv returned from west sumatra [11] . autochthonous denv infection was first reported in france in 2010 and has since been reported at an almost yearly basis [12, 13] . in 2020, by the end of september, six other autochthonous denv cases had been reported by french authorities, one in the department hérault and five in the department alpes-maritime [1, 14, 15] . our case signalled the first evidence of local denv activity in département var in 2020. in the recent past, between 2010 and 2019, six cases of autochthonous transmission were confirmed in the départment var [12] . however, our cases do not seem to have any connection with the other autochthonous cases identified in southern france this year. the cases reported here again illustrate that travel medicine can have a role as a sentinel for detection of silent circulation of infectious diseases [16] . clinicians should be aware of the possibility of 'tropical' vectorborne diseases acquired by travellers within european areas where competent vectors are present, even when cases have not been reported (yet) by local authorities. rapid notification by clinicians and communication between national authorities is essential to ensure timely local risk management and disease control. none declared. tom d. vermeulen: clinical description of case. johan reimerink: serology and co-authoring manuscript. chantal reusken: international notification, epidemiological perspective, co-authoring manuscript. sandra giron: french public health perspective, co-authoring manuscript. peter j. de vries: clinical case management and description, corresponding author. marseille: agence régionale de santé high incidence of peripheral blood plasmacytosis in patients with dengue virus infection dengue guidelines for diagnosis, treatment, prevention and control: new edition. geneva: who travel-associated illness trends and clusters past and future spread of the arbovirus vectors aedes aegypti and aedes albopictus from importation to autochthonous transmission: drivers of chikungunya and dengue emergence in a temperate area chronology of the development of aedes albopictus in the alpes-maritimes department of france european centre for disease prevention and control (ecdc) vector-borne transmission of zika virus in europe, southern france ongoing and emerging arbovirus threats in europe first autochthonous dengue outbreak in italy émergences de dengue et de chikungunya en france métropolitaine bilan de la surveillance des arboviroses en 2019: transition vers une surveillance des cas confirmés de chikungunya, dengue et d'infection à virus zika en france métropolitaine. [review of arbovirus surveillance in 2019: transition to surveillance for confirmed cases of chikungunya, dengue and zikavirus in metropolitan france an outbreak of indigenous cases of dengue detected in the alpes-maritimes cinq cas autochtones de dengue détectés à nice a case of dengue type 3 virus infection imported from africa to italy license, supplementary material and copyright this is an open-access article distributed under the terms of the creative commons attribution (cc by 4.0) licence. you may share and adapt the material, but must give appropriate credit to the source, provide a link to the licence and indicate if changes were made.any supplementary material referenced in the article can be found in the online version. key: cord-252950-eiphxwmn authors: trouillet-assant, sophie; viel, sebastien; gaymard, alexandre; pons, sylvie; richard, jean-christophe; perret, magali; villard, marine; brengel-pesce, karen; lina, bruno; mezidi, mehdi; bitker, laurent; belot, alexandre title: type i ifn immunoprofiling in covid-19 patients date: 2020-04-29 journal: j allergy clin immunol doi: 10.1016/j.jaci.2020.04.029 sha: doc_id: 252950 cord_uid: eiphxwmn covid patients in icu present a high mortality rate and immunoprofiling reveals heterogeneous ifn-α2 production with about 20% of critically-ill patients unable to produce ifn-α2, highlighting the immune response heterogeneity and opening avenues for targeted therapies. sophie trouillet-assant 1,2* , phd, sebastien viel 2,3,4,5* , pharmd, phd, alexandre gaymard merazga for their excellent work. we thank fabien subtil for his helpful advice for statistical analysis. 41 we also thank the life (lyon immunopathology federation) community for fruitful discussion. 42 capsule summary: 43 covid patients in icu present a high mortality rate and immunoprofiling reveals heterogeneous α2 production with about 20% of critically-ill patients unable to produce ifn-α2, highlighting the 45 immune response heterogeneity and opening avenues for targeted therapies. 46 to the editor, 48 49 severe acute respiratory syndrome coronavirus 2 (sars-cov-2) infection (covid-19) is characterized 50 by a wide spectrum of disease encompassing asymptomatic carriage, mild to severe upper 51 respiratory tract illness that can evolve into respiratory failure or rapidly progressing severe viral 52 pneumonia with acute respiratory distress syndrome (ards). disease severity depends on viral strain 53 and host risk factors have been identified such as age and male gender. in addition, an excessive 54 immune response has been identified in patients showing a cytokine storm associated with ards 1 . 55 various immunosuppressive drugs, including il-6 blockers or jak-stat signaling inhibitors have been 56 suggested for the treatment of sars-cov-2 infection 2 whereas additional clinical trials are evaluating 57 the use of recombinant interferon to foster host antiviral response. (clinicaltrials nct04315948, 58 nct04293887). type i interferons (ifn-i) are major components of the innate immune system and 59 represent critical antiviral molecules 3 . to date, ifn-i response has not been evaluated in covid-19 60 patients and its contribution to the viral control and inflammation is unknown. 61 in this study, we assessed the kinetics of plasma ifn-i in covid-19 patients with a spectrum of 62 severity degree. this study was approved by an ethical committee for biomedical research (comité 63 de protection des personnes hcl). (supplemental material and method of this article online 64 repository). 65 firstly, we explored three patients issued from the first covid cluster diagnosed in france (les 66 contamines, haute savoie, france) in february 2020. we took advantage of the new digital elisa 67 technology single-molecule arrays (simoa) 4 and analyzed the kinetics of plasma inflammatory 68 cytokines. interleukin (il)-6, c-reactive protein (crp) and interferon γ-induced protein 10 (ip-10) 69 were elevated in the two symptomatic patients (pt1, 3) (supplementary figure 1 in the online 70 repository). strikingly, no ifn-α2 was detectable in these two patients. in contrast, il-6, crp and ip-71 elevation of plasmatic ifn-α2 was observed. viral loads were low with no obvious quantitative 73 difference between all three patients. 74 we further explored a larger cohort of 26 critically ill covid patients from one of the intensive care 75 unit (icu) at hospices civils de lyon (lyon, france). of note, all the patients were treated with 76 standard of care and none received antiviral or immunotherapies. considering the first 28 days of 77 infection, more than half of critically ill patients required invasive mechanical ventilation (14/26). we 78 observed that patients demonstrated a peak in ifn-α2 at day 8-10 of symptoms onset corresponding 79 to the viral replication phase, that decreased overtime to low but still detectable ifn-α2 the timing of interferon exposition may be critical to control the virus and avoid 98 immunopathogenesis. channappavanar et al. have shown that delayed ifn-i expression can be 99 detrimental in mice in the context of sars-cov-1 infection 6 . our data suggests that screening 100 patients for ifn production is instrumental to select those who could benefit from early intervention 101 with ifn. following day 10, il-6 remains increased while ifn-α tapered. this kinetics highlight that 102 cytokine inhibitors could be helpful at the second phase of the disease following ifn-i decrease. viral 103 characteristic or individual genetic susceptibility should be explored to understand the defect of ifn-104 α production in some covid patients. some ifn-α2 positive patients also experienced fatal outcome 105 highlighting the multifactorial causes of disease severity. we acknowledge limitations of this study, 106 related to the small number of included patients and the technical limitation for the measurement of 107 ifn-β and ifn-λ, in this proof of concept study. 108 here, we provide new argues for an early intervention with recombinant ifn-α2 and we also 109 highlight the window of opportunity for immunosuppressors at the second phase of the disease, 110 delay between symptom onset and icu admission (days) 7 [1-11] 7[0-15] 0.769 bacterial co-infection during icu stay (n (%)) 3 (60%) 7(33%) diabetes (n (%)) 1 (20%) 3(14%) chronic obstructive pulmonary disease (n (%)) 0 (0%) 3(14%) cardiovascular disease (n (%)) 2(40%) 9 (43%) hypertension (n (%)) 3 (60%) 7 (33%) cancer (n (%)) 1 (20%) 3 (14%) active smokers (n (%)) 0 (0%) 1(5%) mortality at d28 after symptom onset(n(%)) 2 (40%) 8 (38%) 1.000 crp -c-reactive protein, icu -intensive care unit, bmi -body mass index table 1 -clinical characteristics of covid-19 patients in intensive care unit 130 p-value are calculated using mann-whitney test for quantitative values and using fisher-exact test for qualitative ones. a. plasma ifn-α concentrations (fg/ml) were determined by single molecule array (simoa) b.c.d. il-6, crp and ip-10 concentrations were measured using a multiplexed assay with the ella platform. e. viral load is represented as cycle threshold of ip2 rt-qpcr using assay designed by pasteur institut in paris. ifn-interferon ; il-6 -interleukin 6 ; crp -c-reactive protein ; ip-10 -interferon γ-induced protein 10 a. ifn score is a transcriptionnal signature defined by 6 interferon-stimulated gene (isg) quantified using nanostring technology and obtained from paxgene tubes in 4 covid-19 patients. b-d. normal values for healthy volunteers was indicated by grey area. vertical bar indicates median delay between symptom onset and icu admission. concentrations of ifn-γ were quantified in only 16/26 patients because of lack of material. clinical features of patients infected with 114 2019 novel coronavirus in wuhan, china covid-19: 116 consider cytokine storm syndromes and immunosuppression type i interferons (α/β) in immunity 118 and autoimmunity /679 and directive 95/46/ec) and the french data protection law (law n°78-17 on 06/01/1978 and décret n°2019-536 on 29/05/2019), we obtained consent from each patient or his next of kin usa) on plasma samples of covid-19 patients. the assay was based on a 3-step protocol using an hd-1 analyzer (quanterix). il-6, crp and interferon γ-induced protein 10 (ip-10) concentrations were measured using a multiplexed assay with the ella platform (protein simple© ca, usa), according to manufacturer's instructions. plasma il28a/b and il-29 (type iii interferon) have been quantified by elisa (pbl laboratories rna integrity was then evaluated by agilent rna microarray (agilent technologies© data standardization was obtained using the geometric mean of internal control and housekeeping genes count number. interferon score was calculated as previously described 1 . virus quantification load viral load was quantified from nasopharyngeal swabs or endotracheal aspirates. rna extraction was performed by the automated nuclisens® easymag® (biomérieux, marcy l'etoile, france) using manufacturer's instructions. a 25 μl reaction contained 5 μl of rna p-value were calculated using mann-whitney test for quantitative values and using fisher-exact test for qualitative ones comparison of rt-qpcr and nanostring in the measurement of blood interferon response for the diagnosis of type i interferonopathies walzer international center of research in infectiology, lyon university, inserm u1111, cnrs umr 5308, ens, ucbl, lyon, france we explored the first three sars-cov-2 positive patients diagnosed in france (les contamines, france) in february 2020. patient 3 : a high risk contact (a 54-year-old man) initially negative for sars-cov-2 developed fever and cough with respiratory crackles at auscultation on the fifth day of hospital isolation. a bilateral interstitial syndrome at the ct-scan with bilateral ground-glass opacification predominant on the left. sars-cov2 was detected from endotracheal aspirates (eta), all nasopharyngeal swabs were always negative. the daily follow-up revealed a short-lasting excretion with only two successive eta for these three patients, no other respiratory pathogens were detected. these patients did not need oxygenation, nor antibiotics, steroids or antiviral agents. plasma samples and paxgene® tubes were collected from covid-19 patients hospitalized in the university hospital of lyon (hospices civils de lyon), france. diagnosis of covid-19 was confirmed in all patients by rt-pcr.all critically ill patients, admitted to icu, were included in the mir-covid study. this study was registered to the french national data protection agency under the number 20-097 and was approved by an ethical committee for biomedical research (comité de protection des personnes hcl) under the number n°20-41. in agreement with the general data protection regulation (regulation key: cord-257325-pvf0uon3 authors: zeitoun, jean-david; faron, matthieu; lefèvre, jérémie h. title: impact of local care environment and social characteristics on aggregated hospital-fatality rate from covid-19 in france: nationwide observational study date: 2020-10-10 journal: public health doi: 10.1016/j.puhe.2020.09.015 sha: doc_id: 257325 cord_uid: pvf0uon3 objectives we aimed to investigate possible differences in aggregated hospital-fatality rate from covid-19 in france at the early phase of the outbreak, and to determine whether factors related to population or healthcare supply before the pandemic could be associated with outcome differences. study design nationwide observational study including all french hospitals from january 24, 2020 to april 11, 2020. methods we analysed aggregated hospital-fatality rate. a poisson regression was performed to investigate associations between characteristics pertaining to populational health, socioeconomic context and local healthcare supply at baseline, and the chosen outcome. results on april 11, 2020, a total number of 30 960 patients were hospitalized among the 3 046 french healthcare facilities, including 6 832 patients in intensive care unit (icu). a total of 8 581 deaths due to covid-19 had been recorded, with a median mortality rate per 10 000 people per department of 0.53 (iqr: 0.29-1.90). there were significant variations between the 95 french departments even after adjusting on outbreak inception (p<0.001). after multivariable analysis, four factors were independently associated with a significantly higher aggregated hospital-fatality rate: a higher icu capacity at baseline (estimate=1.47; p=0.00791), a lower density of general practitioners (estimate=0.95; p=0.0205), a higher fraction of activity from the for-profit private sector (estimate=0.99; p<0.001), and the ratio of people over 75 (estimate=0.91; p=0.0023). conclusions aggregated hospital-fatality rate from covid-19 in france seems to vary among geographic areas, with some factors pertaining to local healthcare supply being associated with outcome. first cases of coronavirus disease 19 , the viral pneumonia related to severe acute respiratory syndrome coronavirus 2 (sars-cov-2), were officially identified in december 2019 in china and were notified to the world health organization (who) on december 31, 2020. 1 since then, the epidemic has expanded well beyond china and the pandemic has officially been declared by the who on march 11, 2020. 2 while italy has been the earliest disease cluster in europe 3 , france has rapidly followed. on february 23, 2020, the french ministry of health issued the phase i of the national epidemic. phases ii and iii were respectively announced on february 29, 2020 and march 14, 2020. 4 fatality rate, defined as the number of deaths of patients in whom covid-19 was confirmed, divided by the total number of covid-19 cases, seems to vary among countries. italian reports have shown a casefatality rate ranging from approximately 7% to 10% 5 , while other countries such as south korea have observed much lower figures. 6 even if there is uncertainty due to variations in case recording, we lack definitive explanations for possible differences in case-fatality rates between countries. the number of tests that could be made to screen and insulate patients has been raised as a possible factor contributing to differences. also, it is not known whether this outcome varies within a country. several factors can likely explain differences such as affected population profile, healthcare environment and quality of care. there has been concern in france regarding critical care capacity with respect to the probable high number of simultaneous severe cases during the outbreak peak. 7 it has been estimated by the french ministry of health that there were approximately 5,000 intensive care unit (icu) beds in france yet with differences between regions. estimates forecasted that this capacity would be exceeded. 7 j o u r n a l p r e -p r o o f therefore, we sought to measure aggregated hospital-fatality rate from covid-19 in france, and to examine the association between populational and local healthcare supply characteristics, and this outcome. we used official and publicly available sources to retrieve and gather the needed data: we also retrieved the number of hospital beds per 10 000 people, including surgery beds, medicine beds, obstetrical beds, physical medicine beds, psychiatry beds and those in long-term care facilities (2017) according to a 2019 report from the french ministry of health, 9 and the total number of adult intensive care beds in each department at baseline, i.e. before the outbreak (2020). last, the fraction of hospital care activity as measured by hospital-days, performed by the for-profit private sector was collected (2017). for each department, the following health indicators were retrieved: overall mortality aggregated hospital-fatality rate was chosen as study outcome (i.e. for each day of the study period, the number of hospital deaths divided by the number of admitted patients). we chose not to analyze case-fatality rate since it would be unreliable in the french case. indeed, france has not performed systematic or large sars-cov-2 testing, and the number of recorded cases has repeatedly been recognized as being orders of magnitude below actual frequency. conversely, all serious cases of suspected covid-19 were required to be tested for confirmation. hospitalized cases, whether in regular wards or intensive care units (icus), therefore represent a reliable denominator for calculation. for each day of study period and in each of the 95 french departments, the number of hospitalized covid-19 patients and the number of covid-19 patients in icus were collected. also, for each day of study sample, the j o u r n a l p r e -p r o o f cumulative number of covid-19-related in-hospital deaths over study period was collected. to account for gaps in outbreak start between areas, the time origin for each department was set to the first day where at least 10 deaths due to covid-19 had been recorded in total. to investigate the relationship between our covariates and the selected outcome, a mixed-effects poisson generalized linear regression was used. models were adjusted for the number of people living in the department and the corrected day since the beginning coded as a third order polynomial as fixed effects. to account for the hierarchical structure of our data, the department (grouping variable) was used as a random effect. both a random intercept and random slope (for the corrected days since the beginning) were used. any variable achieving a pvalue < 0.2 in the univariable analysis was proposed in the multivariable model. in there were a total number of 3046 healthcare facilities (including public hospitals, table 1 . the median area of the 95 departments was 5 880 km 2 (iqr: 4 977-6 817 km 2 ). the study included data from january 24, 2020 (first french case) to april 11, 2020. the details of univariate and multivariable analyses are given in table 1 . following univariate analysis, eleven factors were included in the multivariable analysis. apart from the population, four factors were independently associated with a significantly higher aggregated hospital-fatality rate from covid-19: a higher icu capacity at baseline (estimate=1.47; p=0.00791), a lower density of general practitioners (estimate=0.95; p=0.0205), a higher fraction of activity from the for-profit private sector (estimate=0.99; p<0.001) and the ratio of people over 75 (estimate=0.91; p=0.0023). no health indicator was associated with our outcome in the multivariable analysis. in this nationwide observational study regarding covid-19 in france, we found significant differences between areas in terms of aggregated hospital-fatality rate. four factors were associated with our study outcome: a higher density of icu beds at baseline, a lower fraction of hospital care activity from the for-profit private sector, a j o u r n a l p r e -p r o o f lower density of general practitioners, and a greater proportion of people over 75 were all predictors of higher aggregated hospital-fatality rate in the current model. our study has several strengths. first, it is a nationwide analysis gathering exhaustive data from reliable sources. for most of covariates, year of availability was very recent, thereby limiting timeliness issues. in addition, the variables of interest are unlikely to significantly change across a relatively short period of time. second, we collected a very diverse set of data regarding demographics, populational health, wealth, and also characteristics of care supply and local healthcare ecosystems. populational health data were in particular critical to incorporate in the model since they are factors likely to influence disease outcome. we had very fine health data beyond age, namely prevalence of chronic conditions that have already been recognized as risk factors for covid-19 outcome. 3, 11, 12 third, we used a robust statistical model to analyse the data, namely a poisson linear model as the variables were daily counts and a mixed model as the observed data were not independent (repeated measures within a department), which allows separate intercept and slopes for each department. also, time-adjustment was made so as to align all departments on a similar basis and take into account timeliness issues. our findings have implications. critical care capacity has been a matter of concern regarding covid-19 outbreak. it has been predicted that france did not have enough icu beds to absorb all of the patients in need along several days or weeks. yet we found no evidence that less icu beds at baseline in a given area were associated with a worst outcome. conversely, we found that areas with an initial higher density of icu beds were associated with a higher aggregated hospital-fatality rate. we do not have any certain explanation for those unexpected findings. it may be that critically ill patients were more often transferred from rural areas or smaller facilities to more j o u r n a l p r e -p r o o f comprehensive facilities. it also should be underlined that hospitals have anticipated the outbreak progression by resetting their organization and creating new icu capacity in other wards. we could not measure actual icu beds at a given time since those data were not consistently reported. this will need further investigation. we also found that areas in which the density of general practitioners was higher were associated with a better outcome. even though this should be interpreted with caution, one may hypothesize that general practitioners played a critical role in the epidemic, through adequate orientation of covid-19 patients to hospitals while maintaining others at home. last, it is remarkable that social and wealth factors were not associated with the chosen outcome. the relationship between wealth and health has been consistently documented by a huge body of literature. again, we cannot certainly explain why herein departments with more deprivation were not associated with a higher aggregated hospital-fatality rate yet it should be recalled that france has a very protective social system with a great safety net. perhaps it helped to attenuate the social risk in the case of the epidemic. this study has limitations. firstly, as an observational study, it cannot establish definitive causality. we cannot exclude the possibility that our results might be confounded by factors that were not measured. in particular, we cannot rule out that criteria for admitting patients were different among areas and that some hospitals had more serious cases than others, whether in regular wards or icus. also, we did not have access to age-and gender-structure of hospitalized patients. last, we did not take into account control measures implemented in the different departments even though those measures were thought to be very similar. secondly, the follow-up was intentionally limited. however, given the high urgency that many healthcare systems are currently facing worldwide, we aimed at rapidly providing a first evaluation of j o u r n a l p r e -p r o o f hospital-fatality rates from covid-19 in a markedly affected country. subsequent work over the outbreak course will say whether local differences and their associated factors persist. thirdly, we did not have access to hospital data or patient data. thus, we could not calculate individual hospital-fatality rate and had to deal with aggregate measures which have been updated on a daily basis at the department level over the study period. fourth, we intentionally excluded nursing home since the related data were not available across the whole study period. this represents a selection bias. last, as of march 28, 2020, the french government decided to implement targeted transfers of seriously ill patients by medicalized trains or helicopters in order to improve resource allocation within the whole territory. those transfers may have interfered with our results even though we believe it is unlikely. indeed, reported counts of those transfers showed it involved very few patients as compared to the magnitude of the epidemic. it seems implausible that it significantly influenced the findings from the regression analysis, which were otherwise consistent over time. in conclusion, we found significant differences in aggregated hospital-fatality rate across french areas over the early period of the covid-19 outbreak. several factors pertaining to local healthcare supply were associated with a worst outcome, such as a higher icu capacity at baseline and a lower involvement from the private sector as well as a lower density of general practitioners. those findings clearly deserve further investigation with hospital-or patient-level data and over a longer follow-up. those departments have been chosen to illustrate the heterogeneity of situations across the whole french territory (see figure 1 ). world health organization. who director-general's opening remarks at the media briefing on covid-19 -11 critical care utilization for the covid-19 early experience and forecast during an emergency response arrêté du 14 mars 2020 portant diverses mesures relatives à la lutte contre la propagation du virus covid-19 2020 case-fatality rate and characteristics of patients dying in relation to covid-19 in italy transmission potential and severity of covid-19 in south korea coronavirus : les simulations alarmantes des épidémiologistes pour la france health as an independent predictor of the 2017 french presidential voting behaviour: a crosssectional analysis les établissements de santé -édition the association between income and life expectancy in the united states clinical characteristics of coronavirus disease 2019 in china characteristics of and important lessons from the covid-19) outbreak in china: summary of a report of cases from the chinese center for disease control and prevention key: cord-284616-jgtsl06q authors: harkouk, hakim; jacob, chantal; fletcher, dominique title: urgent development of an anaesthesiology-based intensive care unit for critical covid-19 infected patients date: 2020-05-04 journal: anaesth crit care pain med doi: 10.1016/j.accpm.2020.04.011 sha: doc_id: 284616 cord_uid: jgtsl06q nan the global covid-19 pandemic requires anaesthesiologists to adapt themselves to this unprecedented situation 1 . beyond this first adaptation, the major influx of patients imposes to rapidly manage critical patients outside the usual intensive care structures, in addition to required care of surgical patients. in france, the first cases are diagnosed on january 24 th , 2020. on march 14 th , all unessential institutions are shut down. since march 17 th at noon, the population is confined at home with strict rules. on april 1 st , 6017 patients are hospitalised in intensive care units while national maximal admissions are estimated to be 5000 patients. we shortly describe the use of professional skills and existing structures in a french anaesthesia department to deal with this covid-19 crisis. french anaesthesiologists have 5 years of training with mixed skills in anaesthesia (3 years) and resuscitation (2 years). our 390 beds university hospital is part of the assistance publique -hôpitaux de paris, the first hospitals group in ile de france, a region that is severely affected by the covid-19 pandemic. our structures include 10 operating theatres and 14 beds of recovery room (rr), performing an average of 10.000 scheduled or urgent surgical interventions a year, in trauma, visceral and vascular surgeries. the medical team includes 14 anaesthesia consultants and 8 residents; the paramedic team includes 25 nurse anaesthetists and 13 rr nurses. the intensive care unit (icu), managed by intensivists, has a capacity of 12 resuscitation beds and 14 continuous care beds. a regulatory team headed by an anaesthesiologist with the help of surgeons, usually meeting once a week, decides a new organisation evaluating rapidly both management of critical negative and positive covid-19 patients and surgical activity; chronological details are listed in table 1 the man-power includes 40 anaesthetist nurses and rr nurses and 10 anaesthesiologists (2 anaesthesiologists present 24 hours a day). all these professionals work in 12-hour shifts, 24 hours a day, 7 days a week. this radical reorganisation within 2 weeks of an operating theatre and a rr relies on the professional, structural and material resources of an anaesthesia department to create an icu with 10 beds dedicated to critical covid-19 infected patients while maintaining the management of selected scheduled and emergency surgery. problems to overcome are numerous, covering both patient care, professional protection and urgency of management; we only discuss here three of them: 1. the isolation of covid or non-covid patients by restructuring the circulation areas, using the advantages of an operating theatre (clean and contaminated circuit, closed operating theatre under negative pressure) and identifying two separates care team selected both on professional skills and risk of viral exposition (age over 60 years old and/or comorbidities). all successive decisions were validated with hospital hygiene team; 2. making the best of existing structures for patient care and professional protection: the rr allows easy centralised monitoring of patients but exposes to aerosolised virus, especially with high oxygen flow and requires enhanced protection for nursing professionals (ffp2 mask changed every 8 hours, dressing and take-off procedures, gown) but also an adaptation of the rr 1,2 . negative pressure was installed on day 2 after admission of first critical patients and rr was equipped with 3 air extractors plasmair© (dalkia) which allow treating 7.500 m 3 of air per hour (i.e. 10 volumes per hour for a 750 m 3 rr) 3 ; 3. the medical and paramedical anaesthesia teams had to upgrade rapidly their skills to be able to use high and very high oxygen flow therapy, ventilation of the patient with severe adult respiratory distress syndrome and to be kept informed of additional therapeutic solutions specific to these patients in collaboration with intensivists. the target physician/patient ratio was set to 1/5 (2 anaesthesiologist 24 hours a day) and 1/2 for nurses in the acute phase and 1/2.5 in the steady phase. page 4 of 7 j o u r n a l p r e -p r o o f 4 after 10 days of functioning as icu for critical covid-19 infected patients, 20 patients were admitted with 7 patients with mechanical ventilation. patients start to be discharged from icu and hospital (respectively 12 and 3) and 1 patient is deceased. we report our experience with mobilisation of an anaesthesia team and use of existing structures for urgent creation of an icu managing critical covid-19 patients in a pandemic which exceeds the usual resources of resuscitation structures. j o u r n a l p r e -p r o o f 6 funding statement: support was provided solely from institutional and/or departmental sources conflicts of interest: the authors declare no competing interests service d'anesthésie, hôpital ambroise paré, assistance publique hôpitaux de paris france; for icu organisation and validation of the manuscript service d'anesthésie, hôpital ambroise paré, assistance publique hôpitaux de paris france; for icu organisation and validation of the manuscript service d'anesthésie, hôpital ambroise paré, assistance publique hôpitaux de paris france; for icu organisation and validation of the manuscript service d'anesthésie, hôpital ambroise paré, assistance publique hôpitaux de paris france; for icu organisation and validation of the manuscript service d'anesthésie, hôpital ambroise paré, assistance publique hôpitaux de paris france; for icu organisation and validation of the manuscript service d'anesthésie, hôpital ambroise paré, assistance publique hôpitaux de paris france; for icu organisation and validation of the manuscript service d'anesthésie, hôpital ambroise paré, assistance publique hôpitaux de paris france; for icu organisation and validation of the manuscript service d'anesthésie, hôpital ambroise paré, assistance publique hôpitaux de paris france; for icu organisation and validation of the manuscript service d'anesthésie, hôpital ambroise paré, assistance publique hôpitaux de paris france; for icu organisation and validation of the manuscript service d'anesthésie, hôpital ambroise paré, assistance publique hôpitaux de paris france; for icu organisation and validation of the manuscript references covid-19 infection: implications for perioperative and critical care physicians aerosol and surface stability of sars-cov-2 as compared with sars-cov-1 supplemental treatment of air in airborne infection isolation rooms using high-throughput in-room air decontamination units key: cord-253801-y1nherse authors: lepelletier, didier; andremont, antoine; choutet, patrick title: risque d’introduction et voies d’importation par l’homme de maladies infectieuses exotiques : cas particulier de l’émergence de bactéries pathogènes multirésistantes aux antibiotiques, importées en france à l’occasion de voyages internationaux ou du rapatriement de patients hospitalisés à l’étranger date: 2009-11-30 journal: bulletin de l'académie nationale de médecine doi: 10.1016/s0001-4079(19)32416-1 sha: doc_id: 253801 cord_uid: y1nherse summary the spread of multidrug-resistant bacteria has become a major problem in france in recent years, owing to increasing antibiotic exposure, growing international exchanges, repatriation of hospitalized french patients, and treatment of french and foreign travelers in french hospitals. this article examines how different pathogens may become endemic in france. the spread of multidrug-resistant bacteria has become a major problem in france in recent years, owing to increasing antibiotic exposure, growing international exchanges, repatriation of hospitalized french patients, and treatment of french and foreign travelers in french hospitals. this article examines how different pathogens may become endemic in france. les êtres humains jouent un rôle important dans l'introduction de maladies infectieuses. en toutes ces infections, très médiatisées, ne doivent pas sous-estimer d'autres risques comme l'introduction en france de bactéries résistantes. en effet l'apparition de bactéries pathogènes devenues résistantes aux antibiotiques et leur diffusion constituent un des phénomènes émergents majeurs de ces trente dernières années. certaines espèces bactériennes sont devenues résistantes à plusieurs antibiotiques et parfois à l'ensemble des antibiotiques disponibles : on parle alors de bactéries multirésistantes, de bactéries extrêmement résistantes ou de bactéries panrésistantes [1] [2] [3] . ces terminologies expriment l'évolution de la mulitrésistance et concernent des bactéries cause d'infections associées aux soins pouvant, dans certains cas, aboutir à des impasses thérapeutiques [4] . les niveaux très élevés de la résistance qui sont observés actuellement résultent de l'exposition massive aux antibiotiques à laquelle ont été soumis les humains et les animaux au cours des cinquante dernières années. les résistances aux antibiotiques affectent non seulement les bactéries pathogènes mais aussi, et probablement même beaucoup plus, les bactéries commensales qui colonisent les individus (humains et animaux) et qui sont beaucoup moins facilement repérables car le portage ne s'associe à aucun signe clinique. toutefois, selon les spécificités locales de la pression de sélection et des modes de vie des populations, les niveaux de résistance ne sont pas égaux dans tous les pays. a partir d'une zone où elle a émergé, la diffusion de cette résistance est ensuite facilitée par l'intensité des échanges internationaux et la mondialisation. si les déplacements des animaux, voire de produits de l'agriculture jouent un rôle dans la dissémination mondiale de la résistance, nous allons ici analyser des évènements infectieux liés aux déplacements des populations humaines. en france, les brassages de population sont importants. notre pays est le numéro un mondial des arrivées de touristes avec plus de 80 millions de voyageurs étrangers chaque année. dans la même période, 19,4 millions de français voyageaient à l'étranger [5] . par ailleurs, 1,4 million de français vivent à l'étranger dont 48 % en europe, 20 % en amérique, 15 % en afrique, 8,5 % en asie-océanie et 6,6 % au proche et moyen orient [6] . le rapatriement sanitaire de français hospitalisés à l'étranger, mais aussi les simples retours de voyage et la prise en charge sanitaire d'étrangers en voyage en france, quelle que soit leur nationalité, expose donc la population française à des bactéries multirésistantes aux antibiotiques qui auraient pu être acquises dans des zones de haute prévalence de résistance. ce risque d'émergence et de diffusion à partir des brassages de population est mal évalué quantitativement à l'heure actuelle en france. on sait toutefois qu'il est réel et des évènements sporadiques ou épidémiques ont été observés concernant des pathogènes tels que mycobacterium tuberculosis multirésistant, clostridium difficile de ribotype 027, klebsiella pneumoniae productrice de carbapénèmase, acinetobacter baumannii multirésistant, staphylococcus aureus résistant à la méticilline commu-bull. acad. natle méd., 2009, 193, n o 8, 1821-1833, séance du 24 novembre 2009 nautaire producteur de la leucocidine de panton-valentine ou entérocoques résistant à la vancomycine. dans ce contexte, la question se pose de savoir quel degré de priorité de sécurité sanitaire doit être attribué à la surveillance et au contrôle de la diffusion de ces pathogènes multirésistants aux antibiotiques importés en france, à l'occasion du retour d'un voyage ou du rapatriement de patients hospitalisés à l'étranger. habituellement isolées [12] . [18] . si la souche 027 de c. difficile était la souche prédominante lors des épidémies survenues au québec, elle n'était cependant pas la seule souche cause de l'augmentation de l'incidence observée dans les différents pays.. en particulier, une souche toxine a négative/toxine b positive a causé des épidémies d'icd dans plusieurs hôpitaux d'abord au canada [19] , aux pays-bas [20] et en irlande [21] , associant également une résistance aux fluoroquinolones et aux macrolides. il est probable que la forte augmentation de l'utilisation des fluoroquinolones dans les années qui ont précédé a favorisé ce phénomène. aussi, la surveillance et le contrôle des icd est devenue une priorité nationale depuis 2006 afin d'identifier les cas liés à la souche épidémique mondiale et de maîtriser sa diffusion en france. l'émergence d'entérobactéries résistantes aux carbapénèmes depuis le début des années 1990 est inquiétante, laissant entrevoir des impasses thérapeutiques [22] . les carbapénèmes étant utilisés dans le traitement des infections sévères à entérobactéries productrices de β-lactamases à spectre étendu (e-blse), l'explosion de l'épidémiologie des e-blse est ainsi à l'origine de l'émergence des bactéries résistantes aux carbapénèmes. leur large utilisation crée une pression de sélection qui favorise ensuite l'émergence des souches d'entérobactéries qui y sont résistantes. les souches d'entérobactéries résistantes aux carbapénèmes ainsi sélectionnées appartiennent essentiellement à l'espèce klebsiella pneumoniae mais aussi à d'autres espèces comme escherichia coli. la première souche de klebsiella pneumoniae a été isolée aux etats-unis en 1996, en caroline du nord et dénommée kpc-1. par la suite, d'autres souches de kpc ont été décrites à travers les etats-unis (kpc-2 à kpc-7) sur des modes sporadiques ou épidémiques [22] [23] [24] . la première épidémie de kpc en dehors du territoire américain a été rapportée en israël, à partir de voyageurs et/ou de patients ayant transité entre les deux pays [25] . depuis, de nombreux continents ont rapporté l'émergence de kpc, comme l'amérique du sud, et l'asie. en europe, le phénomène semble rare mais les kpc ont été isolées de manière sporadique en suède, en irlande, au royaume-uni [26] et en grèce qui représente (figure 1 ). en france, plusieurs cas sporadiques ont été isolés chez des voyageurs, rapatriés des états-unis après avoir été hospitalisés [28, 29] , mais d'autres cas ont été importés en provenance d'autres pays de la communauté européenne, notamment de grèce [30] . une attention toute particulière doit être portée à l'importation en france de ce type de bactéries multirésistantes, n'ayant pas encore diffusé sur un mode épidémique, à partir de voyageurs rapatriés et ayant été hospitalisés à l'étranger, a fortiori dans un pays de haute prévalence. le haut conseil de santé publique s'est saisi de ce problème. les sarm communautaires ont émergé aux états-unis à la fin des années 1990, dans des populations jeunes, sans facteur de risque. ces souches étaient génétiquement différentes de celles provenant des hôpitaux et provoquaient principalement des infections cutanées et des pneumopathies nécrosantes [35] , souches dont la virulence est liée à la présence de la toxine de panton-valentine. la diffusion des différents clones de sarm communautaire est complexe et mal élucidée. le clone le plus répandu est le clone américain usa300, particulièrement épidémique. cette souche usa 300 est présente en france, mais le clone principalement détecté sur notre territoire est le clone européen st80. si le clone st80 n'était détecté, en europe, seulement en france et en suisse avant 2003 [35] , on le retrouve maintenant dans de nombreux pays, comme la belgique, le royaume-uni, l'écosse, la suède, la norvège, la finlande, la grèce, la roumanie, l'allemagne, la croatie, les pays-bas, le danemark, la slovénie, mais aussi en algérie et à singapour [36] . d'autres clones sont également présents en france (st5, st8, st59, st377 depuis une dizaine d'années, les autorités sanitaires internationales ont dû faire face à l'émergence et à la diffusion rapide à travers le monde de nouvelles souches de virus grippal, du syndrome de détresse respiratoire aiguë, du chikungunya, et de la tuberculose multirésistante aux antibiotiques... les transports modernes et l'augmentation du tourisme, les voyages d'affaires et l'immigration ont contribué à la dissémination de ces pathogènes à haut impact épidémique [43, 44] . les bactéries multirésistantes aux antibiotiques représentent aussi un risque important [45, 46] . l'augmentation des voyages internationaux de populations à haut risque infectieux, nécessitant une prise en charge médicale ou chirurgicale, et de migrants recherchant des soins spécifiques n'existant pas dans leur pays d'origine, a déjà des implications internationales dans l'émergence et la diffusion de la résistance bactérienne aux antibiotiques [47] . les données de la littérature sur le dépistage systématique de patients hospitalisés à l'étranger et rapatriés dans leur pays d'origine sont peu nombreuses et relativement anciennes. cependant, elles apportent des éléments de réflexion intéressants sur la prise en charge des patients hospitalisés à l'étranger et rapatriés et sur la diffusion de souches mulitrésistantes de pays à pays [48, 49] . cette réflexion doit être intégrée dans les politiques nationales de diminution et de contrôle de la diffusion de la résistance bactérienne aux une large utilisation de céphalosporines favorise l'émergence d'entérocoques. il est probable que l'interdiction de l'utilisation des dérivés des glycopeptides comme promoteurs de croissance en élevage depuis 1997 et l'utilisation plus parcimonieuse de la vancomycine, notamment orale, en médecine humaine ont protégé la france d'une explosion des entérocoques résistants aux glycopeptides. de plus des recommandations ont été rédigées par le comité technique des infections nosocomiales et des infections liées aux soins (ctinils) en 2005 pour stopper la diffusion de la résistance à partir d'un cas sporadique ou de réduire son importance en cas d'épidémie installée. -pandrug resistance (pdr), extensive drug resistance (xdr), and multidrug resistance (mdr) among gram-negative bacilli: need for international harmonization in terminology emergence of extensively drug-resistant and pandrug-resistant gram-nagative bacilli in europe a. -the diversity of definitions of multidrugresistant (mdr) and pandrug-resistant (pdr) acinetobacter baumannii and pseudomonas aeruginosa has the era of untreatable infections arrived institut national de la statistique et des études économiques (insee) ministère des affaires etrangères -la tuberculose en france est-elle d'actualité ? les cas de tuberculose maladie déclarés en france en surveillance de la résistance aux antituberculeux en france : données récentes wordwilde emergence of extensively drug-resistant tuberculosis -le poids de la tuberculose en afrique et ses enjeux internationaux emergence of clostridium difficile toxinotype iii, pcr-ribotype 027-associated disease clostridium difficile infection in patients discharged from us short-stay hospitals a portrait of the geographic dissemination of the clostridium difficile north american pulsed-field type 1 strain and the epidemiology of c. difficile-associated disease in quebec -outbreak of clostridium difficile infection in an english hospital linked to hypertoxin-producing strains in canada and the us van den broek p.j. -clostridium difficile ribotype 027, toxinotype iii in the netherlands first isolation of clostridium difficile pcr ribotype 027, toxinotype iii in belgium. eurosurveillance weekly -first cluster of c. difficile toxinotype iii, pcr-ribotype 027 associated disease in france: preliminary report an outbreak of toxin a negative, toxin b positive clostridium difficile-associated diarrhea in a canadian tertiary-care hospital nosocomial outbreak of clostridium difficileassociated diarrhoea due to a clindamycin-resistant enterotoxin a-negative strain emergence and control of fluoroquinolone-resistant, toxin a-negative, toxin b-positive clostridium difficile the real threat of klebsiella pneumoniae carbapenemaseproducing bacteria novel carbapenem-hydrolyzing â-lactamase kpc-1 from a carbapenem-resistant strain of klebsiella pneumoniae molecular epidemiology of kpc-producing klebsiella pneumoniae isolates in the united states: clonal expansion of multilocus sequence type 258 emergence of kpc-2 and kpc-3 in carbapenem-resistant klebsiella pneumoniae strains in an israeli hospital arrival of klebsiella pneumoniae producing kpc carbapenemase in the united kingdom kpc-2-producing klebsiella pneumonia infections in greek hospitals are mainly a hyperepidemic clone intercontinental travels of patients and dissemination of plasmid-mediated carbapenemase kpc-3 associated with oxa-9 and tem-1 plasmid-mediated carbapenem-hydrolyzing β-lactamase kpc in a klebsiella pneumoniae isolate from france plasmid-mediated carbapenem-hydrolyzing β-lactamase kpc in klebsiella pneumoniae isolate from greece veb-1 extended-spectrum β-lactamase-producing acinetobacter baumannii multi-resistant infections in repatriated patients after natural disasters: lessons learned from the 2004 tsunami for hospital infection control multidrug-resistant organisms in military wounds from iraq and afghanistan community-acquired methicillin-resistant staphylococcus aureus carrying panton-valentine leukocidin genes: worldwide emergence global distribution of panton-valentine leukocidin-positive methicillin-resistant staphylococcus aureus antimicrobial resistance and molecular epidemiology of vancomycin-resistant enterococci from north america and europe: a report from the sentry antimicrobial surveillance program emergence ans spread of vancomycin resistance among enterococci in europe rapide control of an outbreak of vancomycin-resistant enterococci in a french university hospital successful control of a hospital-wide vancomycin-resistant enterococcus faecium outbreak in france -les entérocoques résistants aux glycopeptides (erg) : situation épidémiologique, mesures de contrôle actuelles et enjeux à venir debate-guidelines for control of glycopeptide-resistant enterococci (gre) have not yet worked human migration and infectious diseases globally mobile populations and the spread of emerging pathogens evaluation of repatriation parameters: does medical history matter? international aeromedical evacuation population mobility, globalization, and antimicrobial drug reistance bacterial colonization of patients undergoing international air transport: a propsective epidemiologic study vanderbroucke-grauls c. -carriage of resistant microorganisms in repatriates from foreign hospitals to the netherlands godeau quelle est la responsabilité d'une antibiothérapie préalable inadaptée dans la sélection d'entérocoques résistants ? key: cord-335297-vxhta6a4 authors: véran, emilie; gallay-lepoutre, julie; gory, guillaume; guillaumot, pierre; duboy, julie title: chyloabdomen in a cat with pancreatic carcinoma date: 2018-11-24 journal: open vet j doi: 10.4314/ovj.v8i4.16 sha: doc_id: 335297 cord_uid: vxhta6a4 a 12-year-old spayed female domestic shorthair cat was evaluated for a 3-week history of abdominal distension. chyloabdomen secondary to pancreatic carcinoma was diagnosed. the cat was palliatively managed using rutin and a low-fat diet. the etiology, diagnosis and management of chyloabdomen are discussed. in human and veterinary medicine, chyloabdomen is a rare condition. it results from leakage of lipid-rich lymph into the peritoneal cavity, secondarily to obstruction or increased permeability of lymphatic vessels. in cats, it is believed to be mostly of neoplastic origin. medical investigation relies on a systematic approach and definitive diagnosis often requires histopathological samples. this report describes the exploration and management of a chyloabdomen secondary to pancreatic carcinoma in a 12-year-old cat. case details a 12-year-old neutered female domestic shorthair cat was referred for evaluation of an abdominal effusion. the owners reported an abdominal distension of three weeks duration, with conservation of general demeanor and appetite. the referring veterinarian detected an abdominal effusion. a chyloabdomen was suspected, according to the gross appearance of the liquid. the cat was housed mainly outside. her vaccination status was not up-to-date. on physical examination, the cat was bright, alert and normothermic. the body condition score was of 4 out of 9 and she had an unkempt haircoat. a marked abdominal distension was noted. no pain was elicited on abdominal palpation. cardiorespiratory parameters were within normal limits. abdominocentesis was performed under sedation. approximately 300 ml of milky fluid were removed. triglycerides content in the effusion was highly increased (102.84 g/l; reference range 0.30-1.60). the cytologic analysis of the fluid showed 1000 erythrocytes/µl and 1500 leukocytes/µl, including neutrophils (48%), monocytes (40%) and small, mature lymphocytes (12%). no bacteria nor neoplastic cells were visualized. based on these findings, chyloabdomen was confirmed. complete blood count and biochemistry profile were within normal limits. in-house tests for feline immunodeficiency virus antibody and feline leukemia virus antigen were negative. a feline coronavirus research by polymerase chain reaction on effusion was negative. echocardiography was unremarkable. abdominal ultrasound showed remaining abdominal effusion and an ill-defined heterogeneous mesenteric aggregate in the cranial abdomen, assumed to be mesenteric fat reaction due to chronic effusion. no mass was identified and the pancreas showed no significant ultrasonographic abnormalities. computed tomodensitometry (ct) with lymphangiography was planned. thoracic ct scan was within normal limits. on abdominal ct scan, the amorphous mesenteric aggregate, previously identified on ultrasound, was seen, surrounding the portal vein ( fig. 1) . the caudal extremity of the right lobe of the pancreas was slightly thickened and it was in contact with the mesenteric aggregate. it had ill-defined heterogenous contrast enhancement (fig. 2) . ct scan lymphangiography was performed by injection of contrast media in perianal subcutaneous tissue, as previously described in dogs by ando et al. (2012) . the perianal area was clipped and surgically prepared. using a 25-gauge needle, a warmed water-soluble contrast media (iodixanol, visipaque tm 320, ge healthcare sas, vélizy-villacoublay, france) was injected in the subcutaneous tissue surrounding the anus at 0.6 ml/kg. the administration site was subsequently massaged for 5 minutes. images were obtained with a multi-detector helical ct scan at 5, 10, 15, 20, and 25 minutes after injection. only two lymph nodes in the sacral region and one hypogastric were marked with iodixanol. the remaining lymphatic system was not correctly visualized. http://www.openveterinaryjournal.com e. véran et al. open veterinary journal, (2018) , vol. 8(4): 452-457 ________________________________________________________________________________________________________ http://www.openveterinaryjournal.com e. véran et al. open veterinary journal, (2018) , vol. 8(4): 452457 ________________________________________________________________________________________________________ 454 as no definitive diagnosis had been made, and given the mesenteric lesion of undetermined origin, an exploratory laparotomy was performed. diazepam (valium® roche 10mg/2ml, roche, boulogne-billancourt, france), 0.3 mg/kg body weight (bw), iv, was given as premedication. anesthesia was induced using alfaxalone titrated to effect (alfaxan® 10mg/ml, dechra veterinary products sas, montigny-le-bretonneux, france), 3 mg/kg bw, iv. oro-tracheal intubation was readily performed with a 3-mm cuffed tube. anesthesia was maintained with isoflurane (vetflurane® 1000mg/g, virbac, carros, france) vaporized in oxygen. analgesia was provided by morphine (morphine lavoisier 10mg/ml, c.d.m. lavoisier, paris, france), 0.2 mg/kg bw, iv, q2h. a midline laparotomy was performed. a large amount of chylous effusion was removed. the mesentery was folded upon itself by necrotic adhesions. those were released and biopsies of the abnormal mesentery close to the pancreas and of mesenteric and pancreaticoduodenal lymph nodes were obtained. no other abnormalities were seen on thorough examination of abdominal cavity. abdomen wall was closed routinely. the cat recovered well from the anesthesia. she was discharged 2 days after surgery with amoxicillin/clavulanic acid, 20 mg/kg bw, po, q12h for 5 days (késium® 62.5mg, ceva santé animale, libourne, france), rutin, 62.5 mg/kg bw, po, q8h, and a low-fat diet (royal canin gastro-intestinal low fat, royal canin sas, aimargues, france). biopsies revealed exocrine pancreatic tissue infiltrated by solid sheets of large polygonal cells with oval nuclei, prominent nucleoli and coarse chromatin (fig. 3) . anisokaryosis was moderate and mitotic rate was low. mesenteric fat tissue showed necrotic areas surrounded by vacuolated macrophages (fig. 4) . lymph node biopsies were free from metastatic lesions. pancreatic carcinoma with mesenteric necrotic and granulomatous remodeling was diagnosed. standard chemotherapy and targeted therapy (tyrosine kinase inhibitor) were declined. the cat was palliatively maintained under rutin and low-fat diet (royal canin gastro-intestinal low fat, royal canin sas, aimargues, france). three days after discharge, the cat was presented to the referring veterinarian for inappetence and depression. non-steroidal anti-inflammatory drugs (metacam® 0,5 mg/ml suspension orale pour chats, boehringer ingelheim france division santé animale, reims, france) were prescribed for a few days and the cat improved. on telephonic follow-up, three weeks after discharge, the patient was in good general condition. no relapse of the abdominal effusion was reported. however, one month after surgery, the cat was presented to her regular veterinarian for decreased appetite. 200 ml of abdominal effusion were removed. despite administration of prednisolone (dermipred® 5mg, ceva santé animale, libourne, france), maropitant (cérénia® 16mg, zoetis, paris, france), and mirtazapine (norset® 15mg, msd france, courbevoie, france), the cat's condition worsened with rapid recurrence of effusion. she was euthanized 52 days after the surgery. discussion chyloabdomen is an uncommon condition in veterinary medicine, with sparse data available in the literature. it results from leakage of triglycerides-rich lymph formed in intestinal lacteals into the peritoneal cavity, due to damage or obstruction of the lymphatic system or one of its tributaries (al-busafi et al., 2014) . unlike chylothorax, which is frequently diagnosed as idiopathic, no idiopathic chyloabdomen has been http://www.openveterinaryjournal.com e. véran et al. open veterinary journal, (2018) , vol. 8(4): 452-457 ________________________________________________________________________________________________________ described. the differential diagnosis includes tumor, congestive heart failure, infectious diseases (feline infectious peritonitis, feline immunodeficiency), ruptured cisterna chyli, mesenteric root strangulation or lymphatic vessels malformation. cases in dogs were reported in association with intestinal lymphangiectasia (peterson, 1996) , mediastinal lymphangiosarcoma (myers et al., 1966) , acute pancreatitis (lott et al., 2015) , abdominal lymphatic rupture (fossum et al., 1992) , complication of mesenteric lymphangiography for chylothorax (fossum et al., 1992) , and lymphatic obstruction secondary to thrombus formation (fossum et al., 1992) . in contrast, chylous ascites in cats have mainly been reported in association with neoplastic disease. in the unique published case series of feline chyloabdomen, seven of 9 cats had intra-abdominal malignancy: 4 were diagnosed with a nonresectable solid tumor (hemangiosarcoma, paraganglioma), 2 with lymphoma of the small intestine infiltrating the mesenteric lymph nodes, and one with lymphangiosarcoma of the abdominal wall (gores et al., 1994) . the remaining 2 cats had nonneoplastic diseases: biliary cirrhosis, and steatitis caused by vitamin e deficiency (gores et al., 1994) . chyloabdomen has also been associated with feline immunodeficiency virus (börkü et al., 2005) , feline infectious peritonitis (savary et al., 2001) and hypertrophic cardiomyopathy (nelson, 2001) . a case of chylous pleural and peritoneal effusion with no underlying cause was described in a cat, with no postmortem examination performed (thompson and carr, 2002) . in this case report, pancreatic carcinoma was identified as the underlying cause of the chyloabdomen. linderman et al. (2013) previously reported 34 feline cases of pancreatic carcinoma; chylous ascites was present in one cat in this series. in human medicine, direct malignant cells invasion into lymphatic vessels and obstruction of lymphatic flow by lymph node metastatic infiltration are thought to explain chyloabdomen formation during neoplastic disease (al-busafi et al., 2014) . in this particular case, the mesenteric adipose tissue showed necrotic and granulomatous remodeling on histopathology. this might have led to an obstruction of lymphatic flow and subsequent leakage of chyle. ct lymphangiography is an imaging technique used to assess lymphatic networks through injection of contrast agents. it is frequently used for exploration of chylous effusion when first-line diagnostic procedures have failed to identify the underlying cause of the effusion. ultrasound-guided percutaneous lymphography by mesenteric or popliteal lymph node injection has been recently described in cats (kim et al., 2011; lee et al., 2012) . however, the mobility of the lymph nodes and the volume of contrast agent to administer (1.5ml) can make the procedure tedious (kim et al., 2011; lee et al., 2012) . laparotomy or more recently laparoscopy may be needed to perform the mesenteric injection (brisson et al., 2006) . in this case report, an approach described by ando et al. (2012) was applied. ando et al. (2012) reported an appropriate visualization of the thoracic duct 5 minutes after contrast media injection in a healthy beagle dog. iwanaga et al. (2016) successfully used this protocol in a shiba inu suffering from thoracic duct rupture, with visualization of the duct 10 minutes after injection. in this cat, however, no interpretable visualization of lymphatic system was obtained, with only a few caudal lymph nodes detected. failure of the lymphangiography in this case might be explained by species differences in absorption of subcutaneously injected contrast media, inappropriate dose of iodixanol, a different contrast media (iopamidol in the previous descriptions) or differences in the underlying disease responsible for chylous effusion. in human medicine, the precision of magnetic resonance lymphangiography images is improved if subjects have ingested a high fat meal 3-4h prior to examination rather than fasting (chen et al., 2017) . it could have been a way to enhance lymphatic networks visualization in our case. management of chyloabdomen first relies on treatment of the underlying cause when possible, as in this case pancreatic carcinoma. pancreatic carcinoma is an uncommon tumor in cats with a high metastatic rate and a poor prognosis (linderman et al., 2013) . abdominal effusion is a negative prognostic indicator, with a median survival time of only 30 days (linderman et al., 2013) . gemcitabine is a nucleotide analogue used as a firstline agent in human pancreatic carcinoma (teague et al., 2015) . it has been evaluated in cats alone or in combination with other drugs, like carboplatin or tyrosine kinase inhibitors (martinez-ruzafa et al., 2009; linderman et al., 2013) . chemotherapy provided an improvement of quality of life, but survival time remained poor, with a median of 165 days (linderman et al., 2013) . only four cats were reported to live longer than a year after diagnosis; all of them received gemcitabine-based chemotherapy (martinez-ruzafa et al., 2009; linderman et al., 2013) . in our case, chemotherapy was declined, due to poor long-term prognosis. a palliative treatment with rutin and a low-fat diet was instaured. a low-fat diet may decrease the amount of fat in the effusion, which may improve the animal's ability to resorb fluid from the cavity (hawkins and fossum, 2009 ). rutin is a benzopyrone flavonoid extracted from plants. the exact mechanism of action is unknown; it might reduce leakage from blood vessels, increase proteolysis and removal of protein from tissues, and enhance http://www.openveterinaryjournal.com e. véran et al. open veterinary journal, (2018) , vol. 8(4): 452-457 ________________________________________________________________________________________________________ 456 macrophage phagocytosis of chyle (meadows et al., 1993; gould, 2004; kopko, 2005) . in cats, it is recommended for management of idiopathic chylothorax (gould, 2004; kopko, 2005) and it was successfully used in a case of chylothorax due to cryptococcal mediastinal granuloma (meadows et al., 1993) . by analogy, rutin has been unsuccessfully used in a case of chyloabdomen secondary to hypertrophic cardiomyopathy and a chylous pleural and peritoneal effusion with no underlying condition (nelson, 2001; thompson and carr, 2002) . in this present case, owners have reported a great improvement of the cat general condition after initiation of rutin therapy. the hair coat was shinny and smooth, and the cat was bright and alert. however, effusion rapidly relapsed despite rutin. in conclusion, chylous ascites is an uncommon condition in dogs and cats. neoplastic disease is a leading cause of chylous abdominal effusion in cats. the identification of the underlying cause relies on a systematic and often fastidious approach with analysis of the effusion as a first step. regarding the lymphangiography method, the results obtained in our case were disappointing, even if the injection into the perianal tissue is easier and less invasive than into popliteal or mesenteric lymph nodes. a prospective study would be needed to validate this technique in small animal imaging and establish a standardized protocol for cats. management of chyloabdomen relies on treatment of the underlying cause. by analogy with chylothorax, management with rutin and low-fat diet may be attempted, but to this date, no studies have demonstrated its efficacy. chylous ascites: evaluation and management. isrn hepatology computed tomography and radiographic lymphography of the thoracic duct by subcutaneous or submucosal injection chylous pleural and peritoneal effusion in a cat with feline immunodeficiency virus; diagnosis by lipoprotein electrophoresis comparison of mesenteric lymphadenography performed via surgical and laparoscopic approaches in dogs non-enhanced mr lymphography of the thoracic duct: improved visualization following ingestion of a high fat mealinitial experience chylous ascites in three dogs chylous ascites in cats: nine cases (1978-1993) the medical management of idiopathic chylothorax in a domestic long-haired cat kirk's current veterinary therapy xiv thoracic duct lymphography by subcutaneous contrast agent injection in a dog with chylothorax ultrasound-guided mesenteric lymph node iohexol injection for thoracic duct computed tomographic lymphography in cats the use of rutin in a cat with idiopathic chylothorax ct thoracic duct lymphography in cats by popliteal lymph node iohexol injection feline exocrine pancreatic carcinoma: a retrospective study of 34 cases acute chylous peritonitis associated with acute pancreatitis in a staffordshire bull terrier tolerability of gemcitabine and carboplatin doublet therapy in cats with carcinomas chylothorax associated with cryptococcal mediastinal granuloma in a cat chylothorax and chylous ascites in a dog with mediastinal lymphangiosarcoma chyloabdomen in a mature cat postcaval thrombosis and delayed shunt migration after pleuro-peritoneal venous shunting for concurrent chylothorax and chylous ascites in a dog chylous abdominal effusion in a cat with feline infectious peritonitis advanced pancreatic adenocarcinoma: a review of current treatment strategies and developing therapies hyponatremia and hyperkalemia associated with chylous pleural and peritoneal effusion in a cat the authors declare that there is no conflict of interest. ___________________________________________ key: cord-314884-110nqkej authors: lansiaux, édouard; pébaÿ, philippe p.; picard, jean-laurent; son-forget, joachim title: covid-19 and vit-d: disease mortality negatively correlates with sunlight exposure date: 2020-07-23 journal: spat spatiotemporal epidemiol doi: 10.1016/j.sste.2020.100362 sha: doc_id: 314884 cord_uid: 110nqkej the novel covid-19 disease is a contagious acute respiratory infectious disease whose causative agent has been demonstrated to be a new virus of the coronavirus family, sars-cov-2. alike with other coronaviruses, some studies show a covid-19 neurotropism, inducing de-myelination lesions as encountered in guillain-barré syndrome. in particular, an italian report concluded that there is a significant vitamin d deficiency in covid-19 infected patients. in the current study, we applied a pearson correlation test to public health as well as weather data, in order to assess the linear relationship between covid-19 mortality rate and the sunlight exposure. for instance in continental metropolitan france, average annual sunlight hours are significantly (for a p-value of 1.532 × 10(−32)) correlated to the covid-19 mortality rate, with a pearson coefficient of -0.636. this correlation hints at a protective effect of sunlight exposure against covid-19 mortality. this paper is proposed to foster academic discussion and its hypotheses and conclusions need to be confirmed by further research. the novel covid-19 disease is a contagious acute respiratory infectious disease whose causative agent has been demonstrated to be a new virus of the coronavirus family, sars-cov-2. alike with other coronaviruses, some studies show a covid-19 neurotropism, inducing de-myelination lesions as encountered in guillain-barré syndrome. in particular, an italian report concluded that there is a significant vitamin d deficiency in infected patients. in the current study, we applied a pearson correlation test to public health as well as weather data, in order to assess the linear relationship between covid-19 mortality rate and the sunlight exposure. for instance in continental metropolitan france, average annual sunlight hours are significantly (for a p-value of 1.532x10 -32 ) correlated to the covid-19 mortality rate, with a pearson coefficient of -0.636. this correlation hints at a protective effect of sunlight exposure against covid-19 mortality. this paper is proposed to foster academic discussion and its hypotheses and conclusions need to be confirmed by further research. keywords: covid-19; coronavirus; france; correlation; vitamin d; phototherapy; uv. la nouvelle infection au covid-19 est une maladie respiratoire infectieuse sévère dont l'agent causal a été identifié comme un nouveau virus de la famille des coronavirus , sars-cov-2. comme les autres coronavirus, des études montrent un neurotropisme du covid-19, induisant des lésions démyélinisantes comme dans le syndrome de guillain-barré. plus particulièrement, une note italienne conclue qu'il y a un déficit significatif en vitamine d chez les patients infectés par le covid-19. . patients with the coronavirus pneumonia typically exhibit a fever, with temperature above 38 degrees © and other symptoms such as dry cough, fatigue, dyspnea, difficulty breathing, and diarrhea 1.5 . furthermore, this diseases has a relatively high transmission rate as compared to other upper respiratory illnesses. as a result of this and other factors such as international travel and trade, the initial epidemic has turned into a pandemic in march 2020, with hundreds thousands of individuals confirmed to be infected worldwide -and most likely millions of unreported cases 5 . similar to other coronaviruses-caused illnesses 6 , covid-19 infection has shown some amount of neurotropism [7] [8] [9] , with lesions not unlike those of the guillain-barré demyelination 8 or hemorrhagic necrotizing encephalopathy 7, 9 . meanwhile, it has long been noted that in the case of guillain-barré syndrome, vitamin d deficiency, in relation with high latitude climates, is both a causal and a risk factor 10, 11 . furthermore, a recent italian note has demonstrated a significant vitamin d deficiency in a cohort of covid-19 infected elderly women 12 . therefore, it is important to assess the effect of vitamin d blood levels on covid-19 infection rate and disease course, as it may offer preventative and/or curative options in the context of the ongoing pandemic. specifically in the context of continental metropolitan france, the correlation between sunlight exposure and sars-cov-2 infection will be studied in this article, by using an adjusted pearson test applied to public health and weather data [13] [14] [15] . we conducted a descriptive observational cross-sectional study in order to define a hypothetical relationship between sunlight exposure and sars-cov-2 infection. the source and targeted populations are the whole humanity in view of the ongoing covid-19 pandemic. the eligible population is constituted by the residents of metropolitan continental france. the study was conducted by a consortium of two data analysts, a md-phd specialized in radiology, and a medical student in clinical years. nexgen analytics had no role in making the decision to submit manuscript to the publication, nor did it receive any fee or compensation in the context of this work. the first author vouches for the data and analyses, as well as for the fidelity of this report to the study protocol. we gathered covid-19-related data from various public health and social sources 13, 15 . a parallel multiple group analysis was performed. we excluded the population from the non-metropolitan jurisdictions of france(guyane, mayotte, martinique, reunion, guadeloupe, etc.), due to (1) the fact that their climates vastly differ from that of metropolitan france, and (2) the substantially lower access to healthcare in these areas. moreover, albeit part of metropolitan france, the island of corsica was excluded from this study because of poorer access to healthcare there than on the continent. we chose to use covid-19 mortality rate as the primary variable to evaluate the role of sars-cov-2 infection in our hypothetical correlation. sunlight exposure was evaluated by using the average annual hours of sunshine exposure, as reported by that country's national weather service ("météo france") 14 . our null hypothesis (h 0 ) was the non-correlation between average sunlight hours at the locality (x) and covid-19 mortality rate (y). in order to assess the potential effect of confounding factors, we also considered (1) finally, in order to further sustain our analysis, we also considered the confirmed covid-19 infection cases as well as the number of verified recovered covid-19 patients. we began by computing several descriptive statistics for each variable: arithmetic mean, sample variance, standard deviation and the corresponding confidence intervals (justified by having shapiro-wilk tested each of these variables). obviously unrelated to covid-19 mortality, the 2019 birth and death rates were kept off the analysis. furthermore, age was also eliminated from this analysis as the national statistics in this regard are provided in the form of age classes not directly usable in the context of pearson correlation analysis. all other variables were treated using the pearson correlation test, and the corresponding p-value are reported here in order to assess the statistical significance of these correlations. the this population was subsequently partitioned by region of residence (nb: "region" is the largest sub-national jurisdiction of france), as summarized in table 1 . we note that none of the resulting subgroups was found to exhibit values significantly outside of their respective confidence intervals, per a manova-wilk test performed at the 5% significance level (table 2) . the primary outcome of this analysis was the pearson coefficient between sunlight exposure and covid-19 mortality rate, for which we found a value of -0.6368. with a corresponding p-value 1.532*10 -32 , this allows us to reject the null hypothesis h 0 ( we have shown via pearson correlation that sunlight exposure is significantly correlated (p-value: 1.532*10 -32 ) covid-19 mortality rate in continental metropolitan france (table 3) , which is the main outcome of this study. besides, we acknowledge an interesting secondary finding: namely, the protective effect of life expectancy (pearson r: 0.512; p-value: 7.951*10 -31 ) and discuss it further as it appears counter-intuitive, as older age is already been broadly documented as being associated with worse covid-19 outcomes. however, we also note that in our sample life expectancy is strongly positively correlated with sunlight exposure (pearson r: 1.628*10 (table 3) , and possibly other unknown population confounding variables. nevertheless, our regression, linked with the hypothesized physiopathological mechanism 12 , suggests a first order effect at least. we thus contend that the findings presented in our analysis should be taken into account, in order to envision possibly effective yet inexpensive diagnostic and therapeutic options against the novel covid-19. our conclusions could easily be tested and further assessed by screening the prevalence of covid-19 infected among vitamin d deficient patients. in addition, in vitro cell studies and animal models could be of interest to test our statistical correlation and the physiopathological hypothesis. clinical characteristics of 138 hospitalized patients with 2019 novel coronavirus-infected pneumonia in wuhan, china clinical features of patients infected with 2019 novel coronavirus in wuhan, china. the lancet clinical characteristics of coronavirus disease 2019 in china epidemiologic and clinical characteristics of novel coronavirus infections involving 13 patients outside wuhan, china hrct imaging features in representative imported cases of le potentiel neurotrope des coronavirus. médecine/sciences covid-19-associated acute hemorragic necrotizing encephalopathy: ct and mri features. radiol guillain-barré syndrome associated with sars-cov-2 infection: causality or coincidence? the lancet neurologic manifestations of hospitalized patients with coronavirus disease pulmonary activation of vitamin d3 and preventive effect against interstitial pneumonia vitamin d defiency in patients with primary-immune mediated key: cord-319418-ao0df0q7 authors: chire saire, j. e.; oblitas cruz, j. f. title: study of coronavirus impact on parisian population from april to june using twitter and text mining approach date: 2020-08-18 journal: nan doi: 10.1101/2020.08.15.20175810 sha: doc_id: 319418 cord_uid: ao0df0q7 the fast spreading of coronavirus name covid19, generated the actual pandemic forcing to change daily activities. health councils of each country promote health policies, close borders and start a partial or total lockdown. one of the first countries in europe with high impact was italy. besides at the end of april, one country with a shared border was on the top of 10 countries with more total cases, then france started with its own battle to beat coronavirus. this paper studies the impact of coronavirus in the poopulation of paris, france from april 23 to june 18, using text mining approach, processing data collected from social network and using trends related of searching. first finding is a decreasing pattern of publications/interest, and second is related to health crisis and economical impact generated by coronavirus. officially declared as a global pandemic by the world health organization (who) on march 11, 2020, covid-19 outbreak (coronavirus 19 disease) has evolved at an unprecedented rate. the covid-19 pandemic has resulted in over 20 million confirmed cases and over 700,000 deaths globally. it has also sparked fears of an impending economic crisis and recession [1] . social distancing, self-isolation and travel restrictions have lead to a reduced workforce across all economic sectors and caused many jobs to be lost. schools have closed down, and the need for commodities and manufactured products has decreased. in contrast, the need for medical supplies has significantly increased. all countries that have been affected by covid 19 have followed a similar pandemic growth curve, where the number of cases of sars-cov-2 coronavirus infection continues to grow, and, as time goes by, together with prevention policies, the rate of contagion will start to decrease progressively until the situation is controlled. this has been observed in realities such as those of european countries, which shows that there is certain universality in the temporary evolution of covid-19. this is demonstrated by the time lag graphs of infected populations confirmed in countries such as france, china and italy, which follow the same power law on average [2] . in order to help public health and to make better decisions regarding public health and to help with their monitoring, twitter has demonstrated to be an important information source related to health on the internet, due to the volume of information shared by citizens and official sources. twitter provides researchers an information source on public health, in real time and globally. thus, it could be very important for public health research [3] within the context of covid 19, users from all over the world may use it to identify quickly the main thoughts, attitudes, feelings and matters in their minds regarding this pandemic. this may help those in charge to make policies, health professionals and public in general to identify the main problems that concern everybody and deal with them more properly [4] particularly, focusing on a densely populated region of france, we document evidence that the highest economic "indicators of precariousness," such as unemployment and poverty rates, lack of formal education and housing, are important factors in determining mortality rates for covid-19. therefore, measuring what happens after having the pandemic under control is essential, and the economic issue is important to be monitored, since it goes hand in hand with public health policies for the containment of the pandemic [5] , so that our study will help to show changes in issues that concern the french population at this stage. the actual paper uses data mining approach to perform an exploratory analysis of the dataset of brazilian patients of sao paulo state. the methodology to explore data is presented in section 2, the experiments and results in section 3. conclusion states in section 4, final recommendations and future work are presenten in section 5, 6. the conducted work follows a methodology inspired in crisp-dm [6] . this methodology is explained in the next subsections, from collecting data, processing and visualization to support the study. twitter is a social network, where users can post/share ideas, opinions, thoughts about any topic. then, it is possible to collect text using twitter api(aplication programming interface). the parameters for accesing the data are: terms: covid19, coronavirus -date collection: 23/04/2020 -18/06/2020 -geolocalization: paris, france -language: french -radius: 50 kilometers . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted august 18, 2020. . https://doi.org/10.1101/2020.08.15.20175810 doi: medrxiv preprint a cleaning process is necessary to avoid characters with no meaning for the scope of this analysis. first, convert text to lowercase, remove french accents, remove non alphanumerical values. later, delete stopwords, i.e. articles, pronouns, etc. the scope of this paper is to analyze how reacted french population during the range of date: final week of may until third week of june. besides, know how was the perception around economy situation in paris, france. this step is important to know what kind of graphics will be useful to answer the questions related to the study. the collected date is textual then filtering, organizing it to show proper graphics that support analysis and let a better understanding about the situation in paris, france. cloud of words are useful to get a general overview, bar plots for frequency or histograms, and filtering process removing some terms can help to get a better view of terms. the present paper presents the description of dataset and results in the next subsections 3.1, 3.2. the results presents the interest of parisian inhabitants about health crisis originated by coronavirus and concern around econonomy topic. the collected data has the next features: all countries, including france, in response to 'flattening the curve', generated policies and rules for actions, such as border closures, travel restrictions and quarantine, which is a serious blow to one of europe's largest economies. these actions gave results, achieving a control of the epidemic, evidenced in the figure 1 , where it is clearly observed that since may a constant control of this one was achieved. [7] . . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted august 18, 2020. . https://doi.org/10.1101/2020.08.15.20175810 doi: medrxiv preprint a general ovierview of the collected is presented in figure 2 , after fig. 3 presents the hourly distribution of downloaded twitter posts. it is possible to appreciate that the process of downloading data recovered data from march to june 2020, where clearly the interest evidenced in the web begins to decrease with respect to topics related to fear of the disease, which was very high in previous periods. this assessment is based on the discussion about fear of covid 19 on twitter and the period in which the code to download the data was executed. the issues of fear of this pandemic have been related to issues of quarantine exhaustion, anxiety, depression and fear [8] . is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted august 18, 2020. . https://doi.org/10.1101/2020.08.15.20175810 doi: medrxiv preprint it is necessary to remark there is a clear pattern of publication in population of paris from april to june(see 3). people start interaction at 6h, continues during noon passing afternoon and decreases from 19 -21h. besides, considering image 2, duplicating number of june to have an estimation of the total number for this month. there is a decreasing pattern of publications. by the other hand, a small valley is starting to appear around 13-14h, on may and june. helping the visualisation a cloud of words is presented in fig. 5, and 6 , it can be seen the regions including words related to "corp lutter", "tue comment", "plus parisien", "plus personnes", "trump", these tweets reflect the early interesting around the coronavirus health global crisis on this social network. all the collected data were searched using the keyword "coronavirus". prior to the outbreak of covid-19, people already relied on social media to gather information . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted august 18, 2020. . https://doi.org/10.1101/2020.08.15.20175810 doi: medrxiv preprint and news, and since the outbreak in january 2020, people in many countries have relied on social media like twitter to obtain information about the virus. april april may may may but along with this, the emergence of new causes of anxiety, as detected in the words associated "crise", "ouverture", "pleine crise", "criseéconomique" and "avant crise prix" (fig. 7) , with this analysis, is evident, being the main finding the fear of an imminent economic crisis and recession in france. the covid-19 pandemic has had an unprecedented impact on the global economy as well as individuals' economic well-being [9] [10] the shock of the coronavirus pandemic and shutdown measures to contain it have plunged the global economy into a severe contraction in countries where the pandemic has been the most severe and where there is heavy reliance on global trade, tourism, commodity exports, and external financing. according to world bank forecasts, the global economy will shrink by 5.2 percent this year. [11] this is reasonable as france's leading newspapers already talked about the impact of covid 19 on french economy and made efforts to try to understand the effect it would have, focusing on one of the most important sectors, which is tourism industry with all associated services [12] , including impacts on both the supply and demand of travel [13] . as a direct consequence of covid-19, the . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted august 18, 2020. world travel and tourism council warned that 50 million jobs in the global travel and tourism sector may be at risk [14] . the economic recessions are estimated to affect significantly on the people mental health and wellbeing by magnitude the relative and attributable risks. research [15] indicates a significant adverse effect of job loss and unemployment on mental health sufferings like depression, stress, etc. with this we can show that monitoring and using text mining techniques can detect changes in concerns and fears in the evolution of a population during and after the health emergency by covid 19. this, along with the widespread popularity of social media that will provide the public with a fast platform to measure trends [16] , makes this technique an important public health tool, as it measures in a short time the continuous evolution of communication strategies generated by government institutions. the information analysis on twitter indicated by the detected rates can help to monitor the evolution of the interests of a population like that of france, within the phase of control of the outbreak of the current covid-19 pandemic, showing that public interest in fear of health issues decreased and new fears arose, such as the issue of economic crisis, which is relevant information to generate effective communication policies meeting the needs of a population within the framework of public health. . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. (which was not certified by peer review) the copyright holder for this preprint this version posted august 18, 2020. . https://doi.org/10.1101/2020.08.15.20175810 doi: medrxiv preprint for researchers interested to work with this approach, consider: -select a topic to study and check if social networks are a source for your work. every country has different number of active users and preferences about social networks. -consider to use some tool to get an overview of geographical zone before of using a data collection of some city/state. -remember languages has patterns about how writing(grammar) besides slang or common phrases are dependant of the location, then if you can find one collaborator from the zone of study, this will be very valuable to support the analysis. -involve more people to avoid bias for your own thoughts/ideas and of course, invite specialists around the topic of analysis, they will give you the key terms, intuition about what is useful or not and enforce the project. covid-19, sars and mers: a neurological perspective analysis and forecast of covid-19 spreading in china, italy and france association of the covid-19 pandemic with internet search volumes: a google trendstm analysis top concerns of tweeters during the covid-19 pandemic: infoveillance study the role of economic structural factors in determining pandemic mortality rates: evidence from the covid-19 outbreak in france the crisp-dm model: the new blueprint for data mining coronavirus en francia: 206,696 casos y 30 frontline nurses' burnout, anxiety, depression, and fear statuses and their associated factors during the covid-19 outbreak in wuhan, china: a large-scale cross-sectional study intersecting ethnic and native-migrant inequalities in the economic impact of the covid-19 pandemic in the uk e-commerce y su importancia enépocas de covid-19 en la zona norte del perú covid-19 to plunge global economy into worst recession since world war ii tourism and covid-19: impacts and implications for advancing and resetting industry and research the socio-economic implications of the coronavirus pandemic (covid-19): a review this is how coronavirus could affect the travel and tourism industry covid-19 suicides in pakistan, dying off not covid-19 fear but poverty? -the forthcoming economic challenges for a developing country shindo, and scientific and technical advisory group for infectious hazards who. covid-19: what is next for public health? the lancet perform a deeper analysis about topics related to main sectors: economy, social, health, education. invite collaborators, i.e. economist, sociologist, physicists, teachers to do a global analysis, how one sector can impact/influence to anthers as chain effect. key: cord-262046-bb8ekgdm authors: unlu, e.; leger, h.; motornyi, o.; rukubayihunga, a.; ishacian, t.; chouiten, m. title: epidemic analysis of covid-19 outbreak and counter-measures in france date: 2020-05-01 journal: nan doi: 10.1101/2020.04.27.20079962 sha: doc_id: 262046 cord_uid: bb8ekgdm covid-19 pandemic has triggered world-wide attention among data scientists and epidemiologists to analyze and predict the outcomes, by using previous statistical epidemic models. we propose to use a variant of the well known seir model to analyze the spread of covid-19 in france, by taking in to account the national lockdown declared in march 11, 2020. particle swarm optimisation (pso) is used to find optimal parameters for the model in the case of france. we propose to fit the model based only on the number of daily fatalities, where an r2 score based error metric is used. as number of confirmed cases shall not be fully representative due to low testing especially in the first phases of the outbreak, we present that basing the model optimisation on the fatalities can provide legitimate results. in december 2019, novel coronavirus-sourced atypical pneumonia cases were reported in wuhan, china. rapidly, it evolved into an epidemic in its city of origin [1] . despite taken counter-measures, the outbreak has gradually spread out globally. the world health organisation (who) declared it a pandemic in march 11, 2020 and called for augmented enforcing policies to all governments [2] . countries have responded to outbreak with varying degrees of containment and other preventive actions; also with varying latency [3] [4] [5] [6] . the pandemic is ongoing by the date april 22, 2020 when this paper is written, with total confirmed cases more than 2.5 million and 180,000 deaths. this work focuses on france, where the first confirmed case was reported on january 24, 2020. the situation then rapidly deteriorated, leading authorities to execute more draconian policies [7] [8] . a nation-wide, strict lockdown has been initiated on march 17 and was announced to continue till may 11. by april 22, there are around 159,000 confirmed cases and 20,800 deaths caused by covid19 . french government has announced that [9] [10] lockdown will be lifted starting on may 11 , with strict measures. the exact details of the lockdown lifting policy are not publicized yet, however it is known that the allowed proportion of population and the commercial and public establishments will be gradually increased week by week. the reopening date of high risk establishments such as restaurants or schools is still unknown. it would not be speculative to state that the reproduction number r 0 will also increase gradually over this period, however we expect that it will be much lower compared to pre-lockdown values due to increased awareness and public health measures. considering this setting, we aim to analyze the current situation and its future evolution using a variation of a widely employed mathematical epidemiology model: the seir (susceptible, exposed, infected, recovered) model [11] [12] . we have made the assumption to use two different reproduction numbers: r b 0 and r q 0 . r b 0 is used to describe reproduction rate between the first confirmed death until the start of the lockdown and r q 0 from the start to the end of the lockdown. the parameters of the model are fitted using particle swarm optimization (pso) [13] [14] . however, unlike other approaches we have decided to base our optimization on the number of fatalities only. in addition, rather than mean squared error (mse) or mean squared logarithmic error (msle); we minimize an r 2 score based metric [15] for daily deaths. the rationale behind this approach is that due to largely unknown dynamics of the novel coronavirus such as degree of infectiousness, length of incubation period and limited testing capability, using confirmed cases with such limited information may highly disrupt the validity of the model. however, the number of deaths caused by covid-19 is much more definitive, especially in the first days of the outbreak and thus expected to increase the accuracy. note that, with this approach the initial state values are also defined in terms of proportion of fatalities, hence they also are parameters of the pso optimization. according to our initial results, we estimate the reproduction number at around 3.56 before quarantine and 0.74 after total lockdown, in agreement with various recent studies around the globe for different scenarii [16] [17] . using the developed model we predict that if lockdown continues with strict measures, the total number of covid-19 fatalities should topple below 50,000 (which is currently around 20,000) by late august, 2020; where the effects of the epidemic start to significantly diminish. as it is not possible to predict the reproduction number for the forthcoming lockdown lift, we propose two scenarii with a reproduction number increase of respectively 5% and 10% per week during 3 consecutive weeks. for these two scenarii, it is estimated that total number of fatalities may reach up to 70,000-80,000, and that an epidemic situation could continue till november, 2020. the seir model has been one of the keystone components of statistical epidemiology for a long time, and has proven its validity also for relatively recent regional or global epidemics such as mers, sars and ebola [18] [19] [20] . the seir model is classified as a closed dynamical epidemiological model, as it divides the total population into four distinct categories, where the proportion of individuals belonging to each category evolves over time, subjects passing from one state to other. the temporal transitions between states are defined by several differential equations [11] . initially, the entire population is considered as susceptible except very few infected individuals. gradually over time, according to base differential equations, susceptible individuals are exposed to disease, proportional to the reproduction number r 0 . exposed subjects get infected, recover or die also based on the scalar parameters used in base differential equations. generally, these parameters are calculated for the considered case using optimization algorithms [21] . in the past, researchers have also developed noteworthy number of extended seir/sir models, where additional number of states are added, considering the special circumstances of the evaluated epidemic [18] [22] . we further explain the statistical formulation of an extended seir model in the next section. covid-19 pandemic has immediately gained attention among the research community and numerous different approaches using seir model have been proposed. first studies were published in january, 2020, for the initial epicenter wuhan, less than a month after the novel coronavirus was identified [23] [24] . one particular work aims to extend the model according to characteristics of covid-19, by a research group from the university of basel [25] [26] . authors identify 3 new states: h (hospitalized), c (critical) and d (dead). we base our methodology by following this guideline, as it appears to fit the dynamics of the on going pandemic. as mentionned in the previous section, we developed a seir-hcd model as in [25] . the state transition diagram is shown in fig. 1 . infected individuals may be hospitalized after a certain period. a proportion of the infected agents turn into critical cases, requiring intensive care; whilst the rest recovers. among the critical cases, a certain proportion of individuals eventually dies. these proportion constants are among the parameters of the model to be optimized. initially there are only a few infected agents, i(t 0 ), while rest of the population is susceptible s(t 0 ) = n − i(t 0 ); where n is the population of france (65 millions). also note that, at any time the sum of all 7 states must be equal to n . one of the innovative proposal we make is to base our model completely on the number of deaths. as mentioned previously, the most definitive data for covid-19 case is the number of fatalities, due to current lack of pathological and epidemiological information about the disease and the low number of tests. especially, in the first phase of dissemination of the virus in france, the number of tests was much lower, further decreasing the validity of using confirmed cases as a model initiator. therefore, we propose to estimate the initial number of infected people, i(t 0 ) from the initial number of fatalities d(t 0 ), by simply reverse tracing the state transition diagram, using the proportion of hospitalized h(t 0 ) and critical c(t 0 ) cases. it is important to note that these proportion constants are parameters of the seir-hcd model we aim to optimize along a temporal axis. in other words, one novel outcome of our proposed algorithm is to be able to intrinsically calculate the initial number of infected citizens. it is highly reasonable to indicate that the number of infected people at the time where the first confirmed cases are announced shall be exponentially higher, due to very high proportion of asymptomatic cases [27] . the transitions between states are explained by this set of differential equations in terms of proportion of the total population [25] [26] : all rights reserved. no reuse allowed without permission. was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint (which this version posted may 1, 2020. . where t inc is the incubation period of the coronavirus, t inf is the infectiousness period of an infected agent, t hsp is the duration it takes for an infected agent to check in to a health facility and finally t crt is the duration it takes for an hospitalized person to turn into a critical case since the initial check-in. p a , p c , p f respectively refer to the proportion of asymptomatic infected individuals, the hospitalized agents who switched to a critical case and the critical cases resulting in death. r 0 is the basic reproduction number for the coronavirus. we have decided to use the reported data for france, starting from february 15, when the first fatality was confirmed. without loss of generalization, we assume a binary reproduction number; r b 0 for the interval between february 15, 2020 and march 11, 2020 for the pre-quarantine period and r q 0 for lockdown period, which is still ongoing. finally, we need to optimize the model to find these 9 parameters : r q 0 , r b 0 , p a , p c , p f , t inc , t inf , t hsp , t crt . we employ a particle swarm optimization (pso) for this task, which is a powerful evolutionary algorithm, well suited for this setting [13] . as we propose to set our model solely on the initially reported covid-19 related fatalities, the the initial states of each dynamic component can be denoted as : other than suggesting a model optimization based on fatalities, we have also observed that the most adapted error metric is the r 2 score of differentials of number of deaths (i. e. series of daily fatalities). this avoids overfitting, while preserving the parameter optimization in plausible ranges. the r 2 score is a metric ranging from [−∞, 1], where 1 denotes the full accuracy, so we aimed to minimize the following value: the gradual lockdown lift, starting on may 11, is modeled as a %5 to %10 increase of quarantine time reproduction number for each week for next 3 weeks (assuming reproduction number does nor grow after 3 weeks), presenting two different scenarii for france. with pso optimization based on the defined error metric, the optimal parameters for france are found in table 1 . we have estimated the reproduction number at 3.56 prior to lockdown and 0.74 after lockdown. in an independent research for france, these numbers are reported as 3.3 and 0.5, where their method is not exposed; confirming integrity of our approach from a parallel perspective [28] . we also report an approximate average reproduction number before lockdown that is consistent with the initial various studies on the issue [25] [26] , where the average of the studies suggests a value around 3.6. as shown in table 1 , all other parameters of the suggested model are within the range of other published research on covid-19 [25] . for instance, asymptomatic ratio of 0.79, incubation period of 5.10 days, infectiousness period of 2.79 days, hospitalisation period of 5.14 days, transition to critical state period of 14.06, ratio of 0.12 for infected becoming critical and fatality ratio of 0.33 for critical patients are all close to the medians of the other reported work [25] . these results are particularly interesting, since our seir model optimization approach with suggested error metric for number of fatalities is able to converge to the optimal point, and directly coronavirus related metrics such as incubation period, hospitalisation period etc. (assuming demographics etc. of a country does not influence considerably) are close to the average of the previous works. this asserts the validity of binary reproduction number approach for lockdown and its calculated values. note that, in complex dynamics of a seir model, slight variance of parameters may have a drastic impact on the outputs, such as the estimated number of deaths. we have calculated the mortality ratio of infected people (1 − p a )p c p f as %0.83, where [28] proposes %0.5 for france. by the date april 22, our model estimates the proportion of already infected population as %6.6 ( fig. 2-3) , while [28] also reports as %6. it is quite surprising to observe independent works show similar results, while our seir model optimization only takes fatalities into account for optimization. based on our model, we estimate that if lockdown is maintained, the number of fatalities for france might never pass the limit of 50,000. if reproduction number grows by 5% each week during 3 weeks following the lockdown lift; this upper limit might reach 70,000 (fig. 4) . we observe that, if strict measures are respected till september, 2020 the epidemic almost terminates around this date. in case this growth rate all rights reserved. no reuse allowed without permission. was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. becomes 10% per week, the upper death toll limit might reach 80,000 (fig. 5) ; where epidemic situation lasts till november, 2020. considering the fact that, french government has already prepared a versatile and extensive plan for lifting, by limiting social gatherings, augmented surveillence and nationwide distribution face masks, these scenarii with %5-%10 per week increase of quarantine time reproduction number seem legitimate. in this paper, we have proposed a seir model for covid-19 epidemic in france, similar to [25] . unlike other similar attempts, we have used only confirmed number of deaths as an optimization metric. rationale behind this proposal is that deficiency in testing coverage, especially in the early phases of the outbreak, greatly underestimates the number of confirmed cases. however, the number of confirmed fatalities is a much more solid evidence in this setting. an error metric based on the daily death toll is presented and model parameters are all rights reserved. no reuse allowed without permission. was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint (which this version posted may 1, 2020. . optimized using pso algorithm. also note that, the initial state of the model (initial values of each state) is also defined based on the proportion of fatalities, hence all initial state values are also parameters for the pso optimization. we believe that, for epidemics at this scale starting a seir models with a hypothetical single infected individual may greatly lower the accuracy; especially considering their potential drastic impact. coronavirus disease 2019 (covid-19): situation report world health organization declares global emergency: a review of the how will country-based mitigation measures influence the course of the covid-19 epidemic? modeling the control of covid-19: impact of policy interventions and meteorological factors covid-19: extending or relaxing distancing control measures the effect of travel restrictions on the spread of the 2019 novel coronavirus analysis and forecast of covid-19 spreading in china, italy and france first cases of coronavirus disease 2019 (covid-19) in france: surveillance, investigations and control measures lifting lockdown: france looks ahead to options for easing coronavirus restrictions france to unveil end-of-lockdown plan within 2 weeks global dynamics of a seir model with varying total population size a fractional order seir model with vertical transmission pyswarm: particle swarm optimization (pso) with constraint support particle swarm optimization pseudo-r2 measures for some common limited dependent variable models the reproductive number of covid-19 is higher compared to sars coronavirus novel coronavirus 2019-ncov: early estimation of epidemiological parameters and epidemic predictions a modified seir model for the spread of ebola in western africa and metrics for resource allocation epidemic modelling using sars as a case study dynamical transmission model of mers-cov in two areas a comparison of delayed sir and seir epidemic models survey of models, methods and techniques for computational epidemiology nowcasting and forecasting the potential domestic and international spread of the 2019-ncov outbreak originating in wuhan, china: a modelling study epidemic analysis of covid-19 in china by dynamical modeling about covid-19 scenarios seir-hcd model quantifying undetected covid-19 cases and effects of containment measures in italy covid-19 : seuls 6% des français infectés par le coronavirus le 11 mai key: cord-272085-4mqc8mqd authors: roques, lionel; klein, etienne k.; papaïx, julien; sar, antoine; soubeyrand, samuel title: impact of lockdown on the epidemic dynamics of covid-19 in france date: 2020-06-05 journal: front med (lausanne) doi: 10.3389/fmed.2020.00274 sha: doc_id: 272085 cord_uid: 4mqc8mqd the covid-19 epidemic was reported in the hubei province in china in december 2019 and then spread around the world reaching the pandemic stage at the beginning of march 2020. since then, several countries went into lockdown. using a mechanistic-statistical formalism, we estimate the effect of the lockdown in france on the contact rate and the effective reproduction number r(e) of the covid-19. we obtain a reduction by a factor 7 (r(e) = 0.47, 95%-ci: 0.45–0.50), compared to the estimates carried out in france at the early stage of the epidemic. we also estimate the fraction of the population that would be infected by the beginning of may, at the official date at which the lockdown should be relaxed. we find a fraction of 3.7% (95%-ci: 3.0–4.8%) of the total french population, without taking into account the number of recovered individuals before april 1st, which is not known. this proportion is seemingly too low to reach herd immunity. thus, even if the lockdown strongly mitigated the first epidemic wave, keeping a low value of r(e) is crucial to avoid an uncontrolled second wave (initiated with much more infectious cases than the first wave) and to hence avoid the saturation of hospital facilities. covid-19 epidemic was reported in the hubei province in china in december 2019 and then spread around the world reaching the pandemic stage at the beginning of march 2020 (1) . to slow down the epidemic, several countries went into lockdown with different levels of restrictions. in the hubei province, where the lockdown has been set long before the other countries (on january 23), the epidemic has reached a plateau, with only sporadic new cases by april 15 [from the data of johns hopkins university center for systems science and engineering (2) ]. in france, the first cases of covid-19 were detected on january 24, and the lockdown has been set on march 17. this national lockdown means important restrictions on movement, with a mandatory home confinement except for essential journeys including food shopping, care, 1 h individual sporting activity and work when teleworking is not possible, and closing of the borders of the schengen area. it also includes closures of schools and universities as well as all non-essential public places, including shops (except for food shopping), restaurants, cafés, cinemas, and nightclubs. the basic reproduction number r 0 corresponds to the expected number of new cases generated by a single infectious case in a fully susceptible population (3) . several studies, mostly based on chinese data, aimed at estimating the r 0 associated with the covid-19 epidemic, leading to values from 1.4 to 6.49, with an average of 3.28 (4) . as the value of r 0 can be interpreted as the product of the contact rate and of the duration of the infectious period, and since the objective of the lockdown and associated restriction strategies are precisely to decrease the contact rate, an important effect on the number r e of secondary cases generated by an infectious individual is to be expected. this value r e is often referred to as "effective reproduction number, " and corresponds to the counterpart of r 0 in a population that is not fully susceptible (5) . if r e > 1, the number of infectious cases in the population follows an increasing trend, and the larger r e , the faster this trend. on the contrary, if r e < 1, the epidemic will gradually die out. the control measures in china have been shown to have a significant effect on the covid-19 epidemic, with growth rates that shifted from positive to negative values (corresponding to r e < 1) within 2 weeks (6). the study (7) showed that containment policies in hubei province also led to a subexponential growth in the number of cases, consistent with a decrease in the effective reproduction number r e . fitting a seir epidemic model to time series of reported cases from 31 provinces in china, tian et al. (8) found a basic reproductive number r 0 = 3.15 before the implementation of the emergency response in china, a value that was divided by more than 20 once the control measures were fully effective. using contact surveys data for wuhan and shanghai it was estimated in zhang et al. (9) that the effective reproduction number was divided by a factor 7 in wuhan and 11.5 in shanghai. standard epidemiological models generally rely on sir (susceptible-infected-removed) systems of ordinary differential equations and their extensions [for examples of application to the covid-19 epidemic, see (10, 11) ]. with these models, and more generally for most deterministic models based on differential equations, when the loss of information due to the observation process is heavy, specific approaches have to be used to bridge the gap between the models and the data. one of these approaches is based on the mechanistic-statistical formalism, which uses a probabilistic model to connect the data collection process and the latent variable described by the ode model. milestone articles and textbook have been written about this approach or related approaches (12) , which is becoming standard in ecology (13, 14) . the application of this approach to human epidemiological data is still rare. in a previous study (15) , we applied this framework to the data corresponding to the beginning of the epidemic in france (from february 29 to march 17), with a sir model. our primary objective was to assess the infection fatality ratio (ifr), defined as the number of deaths divided by the number of infected cases. as the number of people that have been infected is not known, this quantity cannot be directly measured, even now (on april 15). the mechanistic-statistical framework allowed us to compute an ifr of 0.8% (95%-ci: 0.45-1.25%), which was consistent with previous findings in china (0.66%) and in the uk (0.9%) (16) and lower than the value previously computed on the diamond princess cruse ship data (1.3%) (17) . in this previous study, we also computed the r 0 in france, and we found a value of 3.2 (95%-ci: 3.1-3.3). although the number of tests at that stage was low, an advantage of working with the data from the beginning of the epidemic was that the initial state of the epidemic was known. here, we develop a new mechanistic-statistical approach, based on a sird model (d being the dead cases compartment), in the aim of • estimating the effect of the lockdown in france on the contact rate and the effective reproduction number r e ; • estimating the number of infectious individuals and the fraction of the population that has been infected by the beginning of may (at the official date at which the lockdown should be relaxed). we obtained the number of positive cases and deaths in france, day by day from santé publique france (18) , from march 31 to april 14. we obtained weekly data on the number of individuals tested (in private laboratories and hospitals) from the same source. we assumed that during each of these weeks the number of tests per day was constant. this assumption is consistent with the small variations between the number of tests during the first week (111,690) and the second week of observation (132,392 the mechanistic-statistical framework consists in the combination of a mechanistic model that describes the epidemiological process, a probabilistic observation model and an inference procedure. the dynamics of the epidemic are described by the following sird compartmental model: with s the susceptible population, i the infectious population, r the recovered population, d the number of deaths due to the epidemic and n the total population. for simplicity, we assume that n is constant, equal to the current french population, thereby neglecting the effect of the small variations of the population on the coefficient α/n. the parameter α is the contact rate (to be estimated) and 1/β is the mean time until an infectious becomes recovered. based on the results in zhou et al. (20) , the median period of viral shedding is 20 days, but the infectiousness tends to decay before the end of this period: the results in he et al. (21) indicate that infectiousness starts 2-3 days before symptom onset and declines significantly 8 days after symptom onset. based on these observations we assume here that the mean duration of the infectiousness period is 1/β = 10 days. in li et al. (22) , the duration of the incubation period was estimated to have a mean of 5.2 days. thus, the mean duration of the non-infectious exposed period is relatively short (about 2-3 days), and can be neglected without much differences on the results, as shown in liu et al. (23) . inclusion of an exposed compartment (as in seir models) is particularly relevant when exposed individuals can indirectly transmit the disease e.g., through insect vectors [e.g., (24) ], which is seemingly not the case for coronaviruses. the parameter γ corresponds to the death rate of the infectious (to be estimated). the model is started at a date t 0 corresponding to april 1st. the initial number of infectious i(t 0 ) = i 0 is not known and will be estimated. the total number of recovered at time t 0 is also not known. however, as the compartment r has no feedback on the other compartments, we may assume without loss of generality that r(t 0 ) = 0, thereby considering only the new recovered individuals, starting from the date t 0 . we fixed d(t 0 ) = 3523, the number of deaths at hospital by march 31. the initial s population at the beginning of the period, should still be close to the total french population: by march 31 only 52,128 cases had been observed in france, corresponding to 0.08% of the total population. a factor 8 had been estimated in roques et al. (15) between the cumulated number of observed cases and the actual number of cases at the beginning of the epidemic. even though this factor may have changed, this means that the proportion of the total population that has been infected by march 31 is still small. we can get an upper bound for the cumulated number of cases by march 31 by dividing the number of hospital deaths at the end of the observation period (10,129 by april 14) by the hospital ifr [0.5%, as estimated in (15)] leading to about 2 million cases. this means that the value of s(t 0 ) is between 65 and 67 million cases. for our computation, we assumed that s(t 0 ) = 66 · 10 6 , corresponding to about 98.5% of the french population. as shown in figure s3 , our results are not much sensitive to the value of s(t 0 ) (at least when s/n remains close to 1). the ode system (1) was solved thanks to a standard numerical algorithm, using matlab r ode45 solver. the number of cases tested positive on day t, denoted byδ t , is modeled by independent binomial laws, conditionally on the number of tests n t carried out on day t, and on p t the probability of being tested positive in this sample: the tested population consists of a fraction of the infectious cases and a fraction of the susceptibles: n t = τ 1 (t) i(t)+τ 2 (t) s(t). thus, with κ t : = τ 2 (t)/τ 1 (t), the relative probability of undergoing a screening test for an individual of type s vs an individual of type i. we assumed that the ratio κ was independent of t over the observation period. the coefficient σ corresponds to the sensitivity of the test. in most cases, rt-pcr tests have been used and existing data indicate that the sensitivity of this test using pharyngeal and nasal swabs is about 63 − 72% (25). we assumed here σ = 0.7 (70% sensitivity). each day, the number of new observed deaths (excluding nursing homes), denoted byμ t , is modeled by independent poisson distributions conditionally on the process d(t), with mean value d(t) − d(t − 1) (which measures the daily increment in the number of deaths): note that the time t in (1) is a continuous variable, while the observationsδ t andμ t are reported at discrete times. for the sake of simplicity, we used the same notation t for the days in both the discrete and continuous cases. in the formulas (2) and (3) i(t), s(t), and d(t) are computed at the end of day t. the unknown parameters are α, γ , κ, and i 0 . we used a bayesian method (26) to estimate the posterior distribution of these parameters. the likelihood l is defined as the probability of the observations (here, the increments {δ t ,μ t }) conditionally on the parameters. using the observation models (2) and (3), and using the assumption that the incrementsδ t andμ t are independent conditionally on the underlying sird process and that the number of tests n t is known, we get: with t i the date of the first observation and t f the date of the last observation. in this expression l(α, γ , κ, i 0 ) depends on α, γ , κ, i 0 through p t and d(t). the posterior distribution corresponds to the distribution of the parameters conditionally on the observations: p(α, γ , κ, i 0 |{δ t ,μ t }) = l(α, γ , κ, i 0 ) π(α, γ , κ, i 0 ) c , where π(α, γ , κ, i 0 ) corresponds to the prior distribution of the parameters (detailed below) and c is a normalization constant independent of the parameters. regarding the contact rate α, the initial number of infectious cases i 0 and the probability κ, we used independent noninformative uniform prior distributions in the intervals α ∈ (0, 1), i 0 ∈ (1, 10 7 ) and κ ∈ (0, 1). to overcome identifiability issues, we used an informative prior distribution for γ . this distribution, say f g , was obtained in roques et al. (15) during the early stage of the epidemic (f g is depicted in figure s1 ). in roques et al. (15) , the number of infectious cases i 0 at the beginning of the epidemic was known (equal to 1), and did not need to be estimated. thus, we estimated in roques et al. (15) the distribution of the parameter γ by computing the distribution of the infectious class and using the formula d ′ (t) = γ i(t) together with mortality data (which were not used for the estimation of the other parameters, unlike in the present study). finally, the prior distribution is defined as follows: π(α, γ , κ, i 0 ) = 1 (α,κ,i 0 )∈(0,1)×(0,1)×(1,10 7 ) f g (γ ). the numerical computation of the posterior distribution is performed with a metropolis-hastings (mcmc) algorithm, using 5 independent chains, each of which with 10 6 iterations, starting from the posterior mode. to find the posterior mode we used the bfgs constrained minimization algorithm, applied to − ln(l) − ln(π), via the matlab r function fmincon. in order to find a global minimum, we applied this method starting from 4,000 random initial values. the matlab r codes are available as supplementary material. denote by (α * , γ * , κ * , i * 0 ) the posterior mode, and s * (t), i * (t), r * (t), d * (t) the solutions of the system (1) associated with these parameter values. the observation model (2) implies that the associated expected number of cases tested positive on day t is n t p * t (expectation of a binomial) with the observation model (3) implies that the expected cumulated number of deaths on day t is d * (t). to assess model fit, we compared these expectations and the observations, i.e., the cumulated number of cases tested positive, t : = c 0 + {s=t 0 ,...,t 0 +13}δs with c 0 the number of cases tested positive by march 31 (c 0 = 52, 128) and the cumulated number of deaths m t : = m 0 + {s=t 0 ,...,t 0 +13}μs , with m 0 the number of reported deaths (at hospital) by march 31 (m 0 = 3 123). the results are presented in figure 1 . we observe a good match with the data. the pairwise posterior distributions of the parameters (α, i 0 ), (α, γ ), (α, κ), (γ , i 0 ), (γ , κ), (κ, i 0 ) are depicted as figure s2 . with the exception of the parameter γ (figure s1 ), for which we chose an informative prior, the posterior distribution is clearly different from the prior distribution, showing that new information was indeed contained in the data. the effective reproduction number can be simply derived from the relation r e = α/(β + γ ) when s is close to n (3). the distribution of r e is therefore easily derived from the marginal frontiers in medicine | www.frontiersin.org posterior distribution of the contact rate α (since we assumed β = 1/10; see section 2.2). it is depicted in figure 2 . we observe a mean value of r e of 0.47 (95%-ci: 0.45-0.50). the marginal posterior distribution of i 0 indicates that the number of infectious individuals at the beginning of the considered period (i.e., april 1st) is 1.4 · 10 6 (95%-ci: 1.1 · 10 6 − 1.8 · 10 6 ). the computation of the solution of (1) with the posterior distribution of the parameters leads to a number of infectious i(t f ) = 7.0 · 10 5 and a total number of infected cases (including recovered) (i + r)(t f ) = 2.0 · 10 6 at the end of the observation period (april 14). by may 10, if the restriction policies remain unchanged, we get a forecast of i(t) = 1.6 · 10 5 infectious cases (95%-ci: 1.3 · 10 5 − 2.1 · 10 5 ) and (i + r)(t) = 2.5 · 10 6 infected cases including recovered (95%-ci: 2.0 · 10 6 − 3.2 · 10 6 ). the dynamics of the distributions of i and i + r are depicted in figure 3 . by may 10, the total number of infected cases (including recovered) therefore corresponds to a fraction of 3.7% of the total french population. this value does not include the recovered cases before april 1st. many studies focused on the estimation of the basic reproductive number r 0 of the covid-19 epidemic, based on data-driven methods and mathematical models [e.g., (4, 27) ] describing the epidemic from its beginning. in average, the estimated value of r 0 was about 3.3. we focused here on an observation period that began after the lockdown was set in france. we obtained an effective reproduction number that was divided by a factor 7, compared to the estimate of the r 0 carried out in france at the early stage of the epidemic, before the country went into lockdown [a value r 0 = 3.2 was obtained in (15) ]. this indicates that the restriction policies were very efficient in decreasing the contact rate and therefore the number of infectious cases. in particular, the value r e = 0.47 is significantly below the threshold value 1 were the epidemic starts dying out. the decay in the number of infectious cases can also be observed from our simulations. it has to be noted that, although the number of infectious cases is a latent, or "unobserved" process, the mechanistic-statistical framework allowed us to estimate its value (figure 3) . the cumulated number of infected cases that we obtained by may 10 (i +r) corresponds to a fraction of 3.7% (95%-ci: 3.0-4.8%) of the total french population, without taking into account the number of recovered individuals before april 1st, which is not known. based on a value r 0 = 3.2, the herd immunity threshold, corresponding to the minimum fraction of the population that must have immunity to stop the epidemic, would be 1 − 1/r 0 ≈ 69% [a threshold of 80% was proposed in (28) ]. this proportion will probably not be reached by may 10. as emphasized by angot (29) , a too fast relaxation of the lockdown-related restrictions before herd immunity is reached or efficient prophylaxis is developed), would expose the population to an uncontrolled second wave of infection. in the worst-case scenario, the effective reproduction number r e would approach the initially estimated value of r 0 , and the second wave would start with about 1.6 · 10 5 infectious individuals (in comparison with the few cases that initiated the first wave in france) and about 64 · 10 6 susceptible individuals. keeping a low value of r e is therefore crucial to avoid the saturation of hospital facilities. we deliberately chose a parsimonious mechanistic model with a few parameters to avoid identifiability issues. possible extensions include stage-structured models, where the infectious class i and the contact rate α would depend on another variable: i = i(t, τ ) and α = α(t, τ ) with τ the time since infection, to take into account the dynamics of the viral load on the infectiousness. see e.g., murray (3) (chapter 19.6) for an introduction to such modeling approaches. another insightful extension would consist in using spatially-explicit models, e.g. reaction-diffusion models (30) to describe the spatial spread of the epidemic, and to be able to estimate local values for the parameter r e and the number of susceptible cases. although herd immunity is far from being reached at the country scale, it is likely that the fraction of immune individuals strongly varies over the territory, with possible local immunity effects [e.g., by april 4 the proportion of people with confirmed sars-cov-2 infection based on antibody detection was of 41% in a high-school located in northern france (31) ]. publicly available datasets were analyzed in this study. this data can be found here: https://www.gouvernement.fr/infocoronavirus/carte-et-donnees https://geodes.santepublique france.fr and https://ourworldindata.org/coronavirus-testing. lr, ek, jp, as, and ss conceived the model and designed the statistical analysis. lr and ss wrote the paper. lr carried out the numerical computations. all authors reviewed the manuscript. world health organization. who director-general's opening remarks at the media briefing on covid-19 an interactive web-based dashboard to track covid-19 in real time the reproductive number of covid-19 is higher compared to sars coronavirus epidemiology of transmissible diseases after elimination the effect of human mobility and control measures on the covid-19 epidemic in china effective containment explains subexponential growth in recent confirmed covid-19 cases in china an investigation of transmission control measures during the first 50 days of the covid-19 epidemic in china age profile of susceptibility, mixing, and social distancing shape the dynamics of the novel coronavirus disease 2019 outbreak in china understanding unreported cases in the 2019-ncov epidemic outbreak in wuhan, china, and the importance of major public health interventions the effect of control strategies to reduce social mixing on outcomes of the covid-19 epidemic in wuhan, china: a modelling study hierarchical bayesian models for predicting the spread of ecological processes modelling population dynamics in realistic landscapes with linear elements: a mechanistic-statistical reaction-diffusion approach dating and localizing an invasion from post-introduction data and a coupled reaction-diffusion-absorption model using early data to estimate the actual infection fatality ratio from covid-19 in france estimates of the severity of coronavirus disease 2019: a model-based analysis estimating the infection and case fatality ratio for coronavirus disease (covid-19) using age-adjusted data from the outbreak on the diamond princess cruise ship covid-19: point épidémiologique du 16 avril 2020 728 000 résidents en établissements d'hébergement pour personnes âgées en 2015 clinical course and risk factors for mortality of adult inpatients with covid-19 in wuhan, china: a retrospective cohort study temporal dynamics in viral shedding and transmissibility of covid-19 early transmission dynamics in wuhan, china, of novel coronavirus-infected pneumonia a covid-19 epidemic model with latency period analysis of transmission dynamics for zika virus on networks detection of sars-cov-2 in different types of clinical specimens preliminary estimation of the basic reproduction number of novel coronavirus 2019 to 2020: a data-driven analysis in the early phase of the outbreak impact of non-pharmaceutical interventions (npis) to reduce covid-19 mortality and healthcare demand early estimations of the impact of general lockdown to control the covid-19 epidemic in france spatial ecology via reaction-diffusion equations cluster of covid-19 in northern france: a retrospective closed cohort study. medrxiv effect of a one-month lockdown on the epidemic dynamics of covid-19 in france this manuscript has been released as a pre-print at medrxiv (32) . the supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmed. 2020.00274/full#supplementary-material conflict of interest: the authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.copyright © 2020 roques, klein, papaïx, sar and soubeyrand. this is an open-access article distributed under the terms of the creative commons attribution license (cc by). the use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. no use, distribution or reproduction is permitted which does not comply with these terms. key: cord-318701-f9j13fsc authors: chamboredon, p.; roman, c.; colson, s. title: covid‐19 pandemic in france: health emergency experiences from the field date: 2020-06-22 journal: int nurs rev doi: 10.1111/inr.12604 sha: doc_id: 318701 cord_uid: f9j13fsc aim: this paper describes the situation regarding covid‐19 emergency in france as of early may 2020, the main policies to fight this virus, and the roles and responsibilities of nurses regarding their work at this time, as well as the challenges facing the profession. background: europe continues to be affected by the covid‐19 pandemic. at the time of writing france was the fourth country with the highest number of detected cases and cumulative deaths. sources of evidence: websites of the world health organization, french government, french agency of public health, french national council of nurses and clinicaltrials.gov database, as well as the experiences of the authors. discussion: the history of the development of the pandemic in france helps explain the establishment of the state of health emergency and containment of the population. many decisions made had undesirable repercussions, particularly in terms of intra‐family violence, mental health disorders and the renunciation of care. hospitals and primary care services, with significant investment by nurses, played a key role in the care of persons with and without covid‐19. conclusion: france has suffered a very high toll in terms of covid‐19 morbidity and mortality, and effects on its people, health systems and health professionals, including nurses. implications for nursing practice: nurses are recognized for their social usefulness in france. however, it is important to consider the collateral effects of this crisis on nurses and nursing and to integrate the health emergency nursing skills established during the pandemic into the standard field of nursing competence. implications for nursing policy: the nursing profession has expectations of a reflection on and revision of nursing skills as well as of its valorization in the french healthcare system, notably carried out by the french national council order of nurses. • present a brief history of the development of the pandemic in the country, including the political decisions that have been taken to combat it; • explain the repercussions of containment measures on the health of the population; • describe the roles and responsibilities of nurses regarding their work during the pandemic, as well as the challenges facing the profession; and • summarize the current french research studies in progress about covid-19. the covid-19 pandemic is undoubtedly the most serious global health crisis in decades, causing more than 283 000 deaths worldwide as of 12 may 2020 (world health organization [who] 2020a). this is a devastating new virus. first reported in wuhan, china, on 31 december 2019, the virus gradually spread to europe and the rest of the world (who 2020b). the emergency situation was declared by who on 31 january 2020. within 100 days of the outbreak of the virus, the director-general of who found that more than 1.3 million people were confirmed as infected, of whom nearly 80 000 died (who 2020c) . at the time of writing on 12 may 2020, the situation in europe remains catastrophic: more than 1 750 000 reported cases and more than 157 000 cumulative deaths (who 2020d) . the most affected countries are spain (227 436 detected cases, 26 744 cumulative deaths), the united kingdom (223 060 detected cases, 32 065 cumulative deaths), italy (219 814 detected cases, 30 739 cumulative deaths), germany (170 508 detected cases, 7533 cumulative deaths) and france (139 519 detected cases, 26 643 cumulative deaths). the situation has necessitated the reorganization of healthcare systems and changes in population lifestyles and has led to particularly difficult economic consequences. to date, the primary strategy has been to utilize cross-contamination measures to prevent the spread of the virus such as good hand hygiene, avoiding close contact with others or social distancing and respecting respiratory hygiene rules. population containment measures have been implemented in many countries, and particularly in france, from 16 march 2020. france is the fourth most affected country in europe. the number of deaths is important, but just as important are the more than 57 000 people who underwent hospitalization for covid-19 (french public health 2020) . the data are updated daily. the most reliable indicator to date remains the incidence of covid-19 cases entering resuscitation/critical care every day, which is beginning to plateau. france must manage the first wave of the pandemic while deploying all means to avoid a second wave. health policies must then adapt to a virus whose spread is not fully known and whose treatments are currently being evaluated. these many unknowns in the equation lead to the need to adjust policy measures in france on an almost daily basis. in preparing this report from the field, relevant information was taken from the websites of who, french government, french agency of public health and french national council of nurses. the clinicaltrials.gov database was also examined. we have also drawn on our experiences as french nurses. the identification of the first three cases of covid-19 positive patients was announced by the ministry of solidarity & health (2020a) on 24 january 2020. the virus began to circulate in france, considered to have been transmitted by people who had stayed in china or singapore and had been in contact with infected people. the first death in france was announced on 15 february 2020 (ministry of solidarity & health 2020b). despite the isolation of the cases identified and the reminder to the public to practice barrier actions, covid-19 spread. subsequently, the minister of solidarity and health, olivier v eran, initiated the plan d'organisation de la r eponse du syst eme de sant e en situations sanitaires exceptionnelles (orsan) (organizational plan for health system response in exceptional health situations) under the epidemic and biological risk section on 23 february 2020, enacting the various protocols to be implemented in the context of a health crisis (ministry of social affairs, health and women's rights 2014). stage 1 of this plan consisted of isolating the identified cases and the people they had been in contact with, at the time numbering about 40 people, to slow down the spread of the virus in the country. a few days later, on 29 february 2020, france moved to stage 2, which consisted of slowing down the viral spread, following the identification of several epidemic outbreaks and the first deaths linked to covid-19. barrier measures were widely disseminated to the population, and containment measures were implemented locally in areas with identified infectious outbreaks. on 12 march 2020, when who declared the status of a pandemic concerning the novel coronavirus (who 2020e), crisis measures were taken by the president of the french republic (2020a) and his government, to control the epidemic and manage the health situation, namely, the closure of the nurseries, schools and universities for users as of 16 march 2020 ; the introduction of short-time work hours for employees whose companies cannot carry out their activities and of teleworking for all employees who have this possibility of adjusting the exercise of their profession (ministry of solidarity & health 2020c). however, a few days later, the number of cases and deaths increased. stage 3 was declared to reduce the circulation of the virus in the population and mitigate its effects. all nonessential public places were closed, and several measures put in place by the french government to manage what was becoming the country's biggest health crisis in several decades. on 16 march 2020, the president of the french republic spoke live on television, declaring that 'we are at war' against covid-19 (2020b) . the white plan corresponded to the provisions of orsan to organize health facilities in response to a major health crisis (ministry of social affairs, health and women's rights 2014). it consisted of four points: mobilizing health establishments to respond to a crisis situation, mobilizing health professionals, mobilizing the material and logistical resources of establishments and adapting their medical activity. initiated in health establishments close to identified epidemic outbreaks, the white plan was generalized throughout france when the epidemic reached stage 3. a new gradation of care began to be implemented: university and public hospitals as the first line to receive patients with covid-19, private hospitals with at least an emergency department and critical care service as the second line and private hospitals with critical care service as the third line. all other care facilities were placed in the fourth line. covid-19 units were set up in more than 150 public hospitals, and new resuscitation places were being created, increasing the capacity from 5000 to 7900 beds (prime minister of the french government 2020). healthcare professionals were mobilized as well as health students on internships or volunteers, and retired people were also called upon to strengthen healthcare teams. the french system of mobilization by the state of volunteer health professionals in exceptional health circumstances, known as the health reserve, was activated to provide support in the areas most affected by the epidemic (ministry of solidarity & health 2020d). non-urgent medical activities were deprogrammed, and the monitoring of chronic pathologies was reorganized. primary care teams, especially home care nurses, were also referred to as backup, to manage not only the usual care of the population but also the aftercare of covid-19 patients discharged from hospital or those who did not require hospitalization, only simple monitoring at home. however, as the existing legislative and regulatory measures were not sufficient to deal with the crisis, the french state introduced the state of health emergency (president of the french republic 2020c). this new state of health emergency covered parts or all of the territory (including overseas territories) in the event of a health disaster that, by its nature and severity, endangered the health of the population. within this framework, the prime minister, as head of the french government, could decree measures listed by the law: order home confinement, requisition personnel and equipment, and prohibit gatherings. the prime minister could also take temporary measures to control the prices of certain products, allow patients to have access to medicines and decide on any regulatory limits to entrepreneurial freedom. the minister responsible for health could, by ministerial order, determine other general and individual measures. the military operation 'resilience' was launched on 25 march 2020 (ministry of the army 2020). the french army served as a reinforcement to provide assistance and support to the population and public services in terms of health, logistics and protection of the entire territory. mistral and dixmude helicopter carriers were deployed in the southern indian ocean (reunion, mayotte) and in the antilles-guyana regions. implementation of containment throughout france up to 11 may 2020 to decelerate the circulation of the virus, the government implemented a containment of the french population (prime minister of the french government 2020b). travel was severely restricted. a certificate justifying individual movements was required to leave the home, and checks were carried out by the police and the army to ensure that these restrictions were respected by the population. those not respecting the confinement were fined or even sentenced to imprisonment according to the severity of the situation. economic measures were put in place urgently by the french state (president of the french republic 2020c). to safeguard jobs and reduce the risks of job insecurity, a shorttime working scheme was launched for the duration of the confinement, enabling more than 10 million people to receive at least three-quarters of their wages. an adapted sick leave scheme was set up for parents of children under 16 years old who could not telework, pregnant women in the third trimester, and vulnerable or fragile persons. unemployment benefit entitlements were extended for persons reaching the end of their entitlement. several types of aid were likewise offered to companies affected by the crisis, to safeguard them and secure jobs in france. concerning children's schooling, pedagogical continuity was achieved at a distance, in virtual classes, or through homework assignments to be carried out with the help of parents. this system had major limitations, including the absence of computer equipment in low-income families, saturation of the bandwidth of internet connections and saturation of educational platforms, which are not accustomed to such a large number of simultaneous connections. containment measures were applied in medical establishment for dependent older adults for dependent older adults (ehpads), where the circulation of the virus was particularly harmful. older adults were initially confined to their rooms, without visiting rights, and these measures were recently relaxed, with permission for visits without physical contact. the french government conferred a broadening of competences and recognition of the role of home care nurses. the health context made it possible to create the first telecare procedure related to the management of patients with covid-19 by home nurses during the period of the state of health emergency (high authority of health 2020; prime minister of the french government 2020c). for the duration of the epidemic, a patient diagnosed with covid-19 could benefit from telecare on prescription, as long as the patient guarantees their availability and mastery of the tele-monitoring tools (smartphone, computer with wi-fi connection, or, failing that, telephone). telecare would be fully covered by the french health insurance. before any care was provided to the patient with covid-19, a nurse collected general information and the care plan prescribed by the doctor for the patient (e.g. points of vigilance, monitoring rhythm). during the first contact, the nurse assessed the patient to confirm the criteria for inclusion in the telecare system, supplemented by measures related to the current situation and, in particular, the implementation of hygiene and prevention measures for the family caregiver. then, as part of the follow-up set-up according to the severity of the patient's condition as indicated by the doctor, the nurse carried out the following: determining the patient's general condition, looking for signs of worsening symptoms, collecting clinical observations at a distance (e.g. temperature, weight), looking for signs of altered consciousness, looking for signs of dehydration, reminding the family and friends of the hygiene and prevention instructions, coordinating with the doctor regarding an alert without delay if the patient's condition required it, or call for emergency medical assistance in case of distress, in parallel with the information from the doctor. if the nurses considered that the conditions would no longer enable them to carry out the follow-up, they would then go to the patient's home to carry out face-to-face monitoring and inform the attending physician, who will adjust the prescription for nursing follow-up as necessary. this new system, requested by the order of nurses, made it possible to monitor patients while drastically reducing exposure to the risk of contamination for caregivers. if telecare was not possible for patient follow-up, and to avoid the risk of spreading the coronavirus within home nursing structures, nurses could opt to follow-up their patients at home, even without specific instruction from the medical prescription. the related procedures were subject to specific coverage and price re-evaluation by the health insurance. prolonged containment can have several implications for the health of the population. the first concern to be feared was the impact on mental health, brought by social isolation, fear of illness and uncertainties in relation to the illness. a survey was conducted by sant e publique france with a sample of 2000 internet users to characterize the impact of covid-19 on the general population and to influence the political measures to be implemented to care for the population (french public health 2020b). because the abovementioned repercussions may be more severe for people with disabilities, particularly psychiatric disorders, specific measures were recommended by the high council of public health (2020a) to adapt containment measures to the problems of each person concerned. these containment measures were applied in ehpads, where the circulation of the virus was particularly harmful. the second concern was that a large, difficult-to-measure proportion of the french population seemed to have given upon their usual, acute or chronic care, mainly because of covid-19 containment measures and the fear of being contaminated. according to a recent survey by a telemedicine platform, the number of consultations with general practitioners decreased by 44% since the beginning of containment, and by 70% for specialist physicians (doctolib 2020) . to date, the effects of this situation remain difficult to assess, especially for people with particular health vulnerabilities. meanwhile, paediatricians alerted the authorities to the decrease in the number of families requesting paediatric consultations, particularly for consultations in connection with the programming of children's vaccinations (french association of outpatient pediatricians 2020). the risk of a resurgence of infectious diseases in children is becoming significant because it is not possible to identify the proportion of children who are not vaccinated according to the vaccination schedule issued by the high council of public health. third, confinement unfortunately endangers a certain number of women and children who are victims of domestic violence (usher et al. 2020) . the french government (2020) widely publicized the possibility of contacting a telephone hotline to report situations of violence. recently, these reports have doubled; however, it is difficult to obtain reliable data to date to estimate the number of collateral victims in confinement. for these reasons, the government has wished to introduce deconfinement for children, who seem less sensitive to the virus, so that a certain number of them can return to school, eat at least one balanced meal a day and escape intrafamily contexts that are harmful to them. finally, several french nurses faced threats or were subjected to malicious acts, often anonymous, by neighbours in particular: posters or anonymous letters asking the nurse to move to avoid contaminating an entire residence, vandalism on personal vehicles or in professional premises, theft of equipment and assault. the french national council order of nurses (2020a) assisted nurses who were victims of these malicious acts in legal proceedings. gradual deconfinement was being implemented as of 11 may 2020 (prime minister of the french government 2020d). the national deconfinement strategy was based on three main principles: protecting the population through barrier gestures and the wearing of masks in certain situations, testing the population on a large scale and isolating sick people and contact cases. departmental (territorial division in france) maps were established to report on situations that may or may not be conducive to deconfinement, according to three main indicators: the rate of new cases in the population over seven days, hospital resuscitation capacity, and organization of the local testing and contact case detection system. the deconfinement plan announced the opening of some public places, including schools, but advised the maintenance of teleworking as much as possible. new rules for social life were also introduced. if the indicators were unfavourable, then a department would not be deconfined. two phases were planned: a first period of deconfinement from 11 may to 2 june 2020, followed by a second period before the summer holidays. despite the exceptional purchasing and requisitioning measures by the french government of personal protective equipment (ppe) and other urgent health supplies, caregivers were left with a real lack of protection, as was the case elsewhere in the world. france was counting on its main supplier, china, without foreseeing that if china itself was exposed to a health crisis such as covid-19, stocks of chinese products would then be used primarily by china. to obtain more precise information on the situation, the french national council order of nurses (2020b) carried out an online consultation from 4 april to 7 april 2020, in which more than 70 000 nurses participated (a sample of 10% of the french nursing population). the main results were as follows: • nearly three-quarters of the nurses consulted stated that they did not have enough ppe. • of the nurses consulted, 83% said they did not have enough gowns, and 78% said they did not have enough masks. • of the nurses consulted, more than two-thirds (65%) stated that they did not have enough protective goggles. • more than half (58%) said they did not have enough overshoes. • more than half (54%) stated that they did not have enough mob caps. • nearly half (47%) stated that they did not have a sufficient quantity of hand sanitizers. the french state set up an emergency system for the purchase of ppe. it has been able to count on the solidarity of the french population and companies, which, on a voluntary basis, have developed the production of masks, gowns and hand sanitizers, although this was not their primary function. to date, 178 french studies on covid-19 have been referenced in clinical trials, 108 of which are in the process of gathering participants. these studies cover the epidemiology of covid-19, clinical trials of drug treatments and their side effects, and the effects of containment. different drug strategies are being investigated, and the results of these studies are expected to be published soon. the results of these studies are eagerly awaited by the french government, by the scientific community, as well as the population. france has suffered a very high toll in terms of covid-19 morbidity and mortality, and adverse effects on its people, economy, health systems and health professionals, including nurses. the context of the health crisis caused by covid-19 in france is leading to strategic and political changes on a daily basis. health professionals in hospitals and primary care facilities are in the front line of the health management of the crisis. however, the population, through political decisions, has a duty to support healthcare workers to reduce the circulation of the virus. after a confinement of almost two months, france is preparing to live a new life, partly deconfined, but with new habits to implement, and above all, a deep reflection on the aftermath of the pandemic. nurses play a key role in the context of the covid-19 health crisis, in hospitals, medical and social care institutions and primary care. the public is largely grateful for nurses' involvement and dedication in this context. although public gratitude may bring satisfaction and value to the profession, the collateral effects of this crisis on the nurses themselves need to be studied. the authorities likewise need to ensure that nurses remain in their profession. derogating measures that would extend the scope of nursing activities during crises also need to be considered to develop and establish them on a permanent basis in nursing practice. it would be inappropriate to withdraw recognized skills acquired during the crisis once the crisis is over. the french concerns are completely in line with the global concerns raised by the international council of nurses (icn), which calls for the recognition, respect and protection of nurses (international council of nurses 2020a). the context of this health crisis places the nursing profession in a social mandate recognized by the french population. it is imperative that nursing practice be adapted and evolved so that france can win the fight against this virus. the french national council order of nurses (2020c) has asked the french government to deploy several means to help nurses accomplish their daily mission: an intensification of efforts to equip nurses working in residential institutions for dependent older people, medico-social establishments or at home with ppe and systematic screening of health personnel; additional efforts to promote tele-nursing; the introduction of differentiated spaces and rounds of home visits (covid-19/non-covid-19); a more efficient system to ensure the quality and continuity of care for all, particularly for at-risk populations and those suffering from chronic pathologies; a strong fight against any malicious act or discrimination towards healthcare workers with regard to their employment and the covid-19 risk; the possibility for nurses to carry out the entire procedure relating to releasing death certificates instead of a doctor; and the prescription of covid-19 tests. the french national council order of nurses has also called for an accurate count of nurses infected with and died from covid-19, the recognition of occupational disease for infected caregivers, and the granting of the status of ward of the nation for the children of deceased nurses. these latter concerns appear to be global, as the icn also notes that the number of nurses who died from covid-19 appears to be underestimated (international council of nurses 2020b). these requests were made during the time of the covid-19 crisis, but the french national council order of nurses asked the french government to rethink completely its vision of the nursing profession. today, the nurse is an essential link in the patient's care journey. the nurse is a clinician, and this must be reflected in a progressive evolution of nursing skills to include skills regarding medical prescription. the international council of nurses has positioned itself to ensure that the critical role of nurses in the management of covid-19, as well as in day-to-day operations, is fully recognized by governments around the world (international council of nurses 2020c). the state of the world's nursing report provides a basis for reflection on the evolution of the nursing practice and better recognition of nurses in all countries (who 2020f). covid-19: doctolib alerts on drop in practice attendance and commits to allowing patients to return for consultation press release 16 april. available at protecting children, continuing to care for them in the midst of the pandemic (in french) the government fully mobilized against domestic and intra-family violence the order assists nurses who are victims of pressure or aggression in their legal proceedings (in french) covid-19: the national order of nurses alerts on the situation of the profession and announces new emergency measures (in french) the national order of nurses makes recommendations for priority measures for deconfinement (in french) covid-19 france covid-19: a survey to monitor behavioural and mental health changes during confinement rapid responses under covid-19 teleconsultation and telecare epidemic at covid-19: support for people with disabilities (in french). notice international nurses day: nurses deserve praise, thanks, protection amid covid-19. press release 11 may icn says worldwide death toll from covid-19 among nurses estimated at 100 may be far higher international council of nurses: nursing the world to health prime minister of the french government2020. introduction of the emergency law to deal with the covid-19 epidemic address transcript 21 march assistance in organising the provision of care in exceptional health situations (in french) three cases of coronavirus (2019-ncov) infection in france (in french) covid-19: a twelfth case confirmed in france (in french) order of 14 march 2020 laying down various measures to combat the spread of the covid-19 (in french). regulatory text 14 march ministry of solidarity and health2020d. order of 4 march 2020 on the mobilization of the health reserve (in french) ministry of the army2020. operation resilience (in french) address to the french population address to french population as an emergency measure to deal with the covid-19 epidemic decree no. 2020-293 of 23 march 2020 prescribing the general measures necessary to deal with the covid-19 epidemic within the framework of the state of health emergency (in french). regulatory text 23 march decree no. 2020-277 of adopting adapted conditions for the receipt of cash benefits for persons exposed to coronavirus (in french). regulatory text 19 march presentation of the national deconfinement strategy (in french) family violence and covid-19: increased vulnerability and reduced options for support who health emergency dashboard who (covid-19) homepage world health organization2020b. coronavirus disease (covid-19) pandemic who director-general's opening remarks at the mission briefing on covid-19 rector-general-s-opening-remarks-at-the-mission-briefing-on-covid covid-19 situation in the who european region who director-general's opening remarks at the media briefing on covid-19 state of the world's nursing report -2020 firstly, we thank all nurses in france, from all sectors of activity, for their involvement in this covid-19 crisis. secondly, we thank the french national council order of nurses for the financial support for the linguistic revision of this article. manuscript design: sc data collection: pc, cr, sc manuscript writing: pc, cr, sc critical intellectual revisions of manuscript: pc, cr, sc key: cord-296081-6coxz3l8 authors: souris, m.; gonzalez, j.-p. title: covid-19: spatial analysis of hospital case-fatality rate in france date: 2020-05-20 journal: nan doi: 10.1101/2020.05.16.20104026 sha: doc_id: 296081 cord_uid: 6coxz3l8 when the population risk factors and reporting systems are similar, the assessment of the case-fatality (or lethality) rate (ratio of cases to deaths) represents a perfect tool for analyzing, understanding and improving the overall efficiency of the health system. the objective of this article is to estimate the influence of the hospital care system on lethality in metropolitan france during the inception of the covid-19 epidemic, by analyzing the spatial variability of the hospital case-fatality rate between french districts. the results show that the higher case-fatality rates observed in certain districts are mostly related to the level of morbidity in the district, therefore to the overwhelming of the healthcare systems during the acute phases of the epidemic. however, the magnitude of this increase of case-fatality rate represents less than 10 per cent of the average case-fatality rate and cannot explain the magnitude of the variations in case-fatality rate reported by country by international organizations or information sites. these differences can only be explained by the systems for reporting cases and deaths, which, indeed, vary greatly from country to country, and not attributed to the care or treatment of patients, even during hospital stress due to epidemic peaks. since the beginning of the epidemic, the case-fatality rate of covid-19 and the 34 differences between countries have been the subject of many questions about national 35 pandemic response policies and patient treatment. most studies on the lethality of 36 the case-fatality rate (or lethality rate) is the ratio between the number of closed 39 cases (i.e. recovered or dead) and the number of deaths due to the disease, it is 40 estimated by the healthcare system based on the reporting of these two values. the 41 case-fatality rate should not be confused with the mortality rate, which is the ratio of 42 the number of deaths to the total population, or also with the morbidity rate, which is 43 the ratio of the number of cases to the total population. mortality and morbidity rates 44 depend on the extent of disease in a population, unlike case-fatality rates, which are 45 normally calculated independently of the number of infected persons [por 08]. 46 the case-fatality rate of a disease in a population is an index of severity of the 47 disease in that population, and of the capacity of the healthcare system to reduce 48 mortality. in principle, this allows to compare the effectiveness of healthcare systems 49 across regions or countries. 50 the aim of this article is to analyze the effectiveness of the healthcare system in 51 france in the context of the covid-19 epidemic. based on spatial differences in 52 lethality, this study ultimately show that the case-fatality rates published by the 53 international agency by country (may 2020) do not allow to compared the country one 54 to the others. 55 lethality depends on the intrinsic virulence of the virus but, unlike morbidity, it does 56 not depend on its contagiousness. virulence comes from the reproductive capacity of 57 the virus in the cell, its capacity for cellular degradation, and its ability to induce or not 58 an innate or specific immune response. virulence is of purely biological origin and once 59 the virus has entered the target cell where it will cause its pathogenic effect does no 60 longer depends on environmental conditions outside the host. virulence is independent 61 of the host population, but may change over time and space if there is a risk of natural 62 mutation/selection of the pathogen. contagiousness characterizes the biological 63 capacity of the virus to reach the target cell system of its host, and the ability to be 64 transmitted from one individual to another. the efficiency of transmission depends 65 largely on environmental conditions (e.g., climate, urbanization, population density, 66 mobility), which can vary greatly from one country to another. 67 in addition to the virulence of the virus, the case-fatality rate depends on biological 68 risk factors and on population vulnerability (age structure, genetic factors, prevalence 69 of co-morbidities, healthcare accessibility, etc.) as well as other factors related to the 70 health system (equipment, capacity, staff, management, care of patients, effectiveness 71 of therapies, patient management in a critical phase of the disease), and factors related 72 to the detection and registration system for cases and deaths (clinical cases definition, 73 detection, surveillance systems, case and death reporting). the evaluation of the case-74 fatality rate normally requires the detection and counting of all infected persons, 75 irrespective of their level of symptoms (i.e. disease severity). 76 when the population risk factors and reporting systems are identical, case-fatality 77 rate evaluation represents an excellent tool for analyzing, understanding and improving 78 the overall performance of the health system, particularly at the level of hospital units. 79 studying the magnitude of differences in case-fatality rates between units also makes 80 it possible to assess the impact of the quality of the health system on case-fatality. 81 there are large differences in the case-fatality rates of covid-19 published by 82 country (table 1) or calculated directly from who data. these rates vary considerably, 83 from less than 0.02 (thailand, australia, chile) to more than 0.15 (france, belgium, 84 uk), with a mean at 0.04 and a standard deviation of 0.045 (who, may 8, 2020, figure 85 1 cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. (which was not certified by peer review) the copyright holder for this preprint this version posted may 20, 2020. . in europe, the characteristics of populations (in terms of risk factor for covid-19) 96 and health systems are quite similar, but the definition, detection and reporting of cases 97 and causes of death can differ greatly from one country to another. some countries 98 conducted significantly more detection tests and hospitalizations than others (table 2) , 99 resulting in differences in the protocols for patient management. the rate of testing 100 performed (policy) and mortality rates (reporting) vary mainly according to the 101 geographical extent of the epidemic within each country. 102 103 . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. (which was not certified by peer review) the copyright holder for this preprint this version posted may 20, 2020. the virulence of the covid-19 pathogen (sars-cov-2 virus) is assumed to be 107 identical in all countries. in order to compare case-fatality rates across regions or 108 countries (and thus analyze the effectiveness of healthcare systems), it is necessary, 109 when calculating rates, to standardize population-related risk factors and to use the 110 same definitions and enumeration methods to record cases and deaths. this is not the 111 case for the current pandemic and discrepancies exist among the country systems. 112 the objective of this article is not to estimate the actual lethality of covid-19 in 113 france based on the rates published by the health authorities, but to estimate the 114 influence of the healthcare system on lethality by analyzing the spatial variability of the 115 hospital case-fatality rate (confirmed hospitalized cases and hospital deaths) in 116 metropolitan france between districts (i.e. french départements). this analysis, 117 limited to metropolitan france, makes it possible while it remains within the framework 118 of the same system for defining and counting cases and deaths. we thus assume that 119 this system of definition and enumeration was identical throughout france during the 120 period (19 march to 8 may) corresponding to the first wave (inception) of the covid-121 19 epidemic in france. therefore the study focuses on the extent of spatial differences 122 in the case-fatality rate in metropolitan france, and enable to highlight the relative 123 differences between districts, as well as to analyze the causes independently of the 124 system of definition and enumeration of cases and deaths, and also independently of 125 the main biological risk factor of severity (age) after standardization on this factor. 126 estimating the variability of the case-fatality rate attributable exclusively to hospital 127 care of patients will then allow us to compare the case-fatality rate observed in 128 metropolitan france with the one calculated for other countries. it will allow us to 129 estimate whether the variability of the case-fatality rate due to the management of 130 . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted may 20, 2020. . https://doi.org/10.1101/2020.05.16.20104026 doi: medrxiv preprint patients in the acute epidemic phase can exclusively explain the significant differences 131 in case-fatality rates observed between countries. 132 this study is based on daily hospitalization and death declaration data by district 135 in france and is accessible on the "santé publique france" website. (www.data.gouv.fr/fr/datasets/donnees-hospitalieres-relatives-a-lepidemie-de-covid-137 19) from march 19 to may 8, 2020, corresponding to 50 days lockdown (i.e. quarantine) 138 and the spread of the covid-19 epidemic in france. we also obtain demographic data 139 by districts (source: population by age, insee, 2020), as well as data on the 140 distribution of hospitalized cases according to age group (10-year age group) (santé 141 publique france). this analysis was carried out on the 96 districts of metropolitan 142 france (figure 2 ), while the french overseas districts and territories were excluded 143 from the analysis for reasons of spatial analysis and mapping. the data were 144 integrated into a geographic information system (savgis, ww.savgis.org) for analysis 145 and mapping. 146 . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted may 20, 2020. severe and asymptomatic forms (which a fortiori do not cause deaths) -it is estimated 153 that only 2.6% of infected persons were hospitalized [sal 20] -this overall lethality is 154 necessarily much lower than hospital lethality, but it will be accurately calculated only 155 at the end of the epidemic when the total number of positive cases (i.e. seroprevalence 156 survey) will be available and the total number of deaths outside hospital due to covid-157 19 will be accurately assessed. 158 all identified and hospitalized cases were tested positive (by rtpcr). all deaths 159 counted were covid-19 associated. as of may 8, 2020, not all hospitalized cases are 160 closed since the epidemic is still ongoing: deaths counted at the beginning of the study 161 period correspond to cases hospitalized but were not included in the study, and cases 162 counted at the end of the period were not closed and no deaths from these cases were 163 included in the study. 164 . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted may 20, 2020. cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted may 20, 2020. belfort) (figure 3) . 224 the hospital mortality rate (not age-standardized) has the same spatial distribution. 225 it varies from 0.01 per 1,000 (tarn-et-garonne) to 1.13 per 1,000 (territoire de belfort), 226 with a mean of 0.21 (median 0.12) and a standard deviation of 0.21. 227 . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted may 20, 2020. . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted may 20, 2020. from the calculation of age-specific case-fatality rates, the slr is between 0.28 and 255 1.67, with the mean at 0.99 and the median at 1 (figure 7) . in the following, we will 256 consider only the slrs calculated with age-specific case-fatality rates that do not take 257 into account the ile-de-france and grand-est regions. 258 . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted may 20, 2020. the spatial distribution of standardized morbidity rate (hospitalized cases) shows 261 a significant spatial autocorrelation (moran index: 1.54, p-value < 10 -6 ), and this is 262 expected for an infectious disease. the case-fatality rate shows also significant spatial 263 autocorrelation (moran index: 0.29, p-value < 0.000007), and this is no expected. the 264 analysis of the clusters clearly shows a clustering of high case-fatality rate values in 265 regions of high morbidity (particularly the grand-est), and shows some cases of 266 districts with high case-fatality rate values isolated in areas with low rates. 267 the breslow & day significance test shows districts where the slr is statistically 268 significantly different from 1, corresponding to districts with abnormally high (slr > 1, 269 red) or abnormally low (slr < 1, green) case-fatality rates. the individual significance 270 threshold is set at 0.05, and for all districts at 0.0005 to account for multi-testing ( figure 271 8). 272 . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. there is a correlation between the standardized hospitalization rate and the 277 standardized case-fatality rate (bravais-pearson index=0.40) (figure 9 ), a correlation 278 which increases (0.48) if we limit the calculation to districts whose slr is significantly 279 different from 1 (p-value < 0.05). to illustrate the increase of case-fatality rate with hospitalization rate, table 4 gives 284 the mean of the standardized case-fatality rate over the districts according to their 285 standardized hospitalization rate. the average case-fatality rate varies from 0.134 for 286 . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted may 20, 2020. the mapping of the hospitalization rate and the hospital mortality rate with the slr 296 shows the spatial correspondence of these values ( figure 10) . a typology combining hospitalization rates and case-fatality rates is proposed: low 302 rates (values below the mean by less than one standard deviation), high rates (values 303 . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted may 20, 2020. . above the mean by more than one standard deviation), so as to represent four classes 304 (low-low, low-high, high-low, high-high). the hatched areas represent those for which 305 the slr is not significant (p-value > 0.05) (figure 11 ). 306 307 figure 11 . combination of standardized hospitalization and case-fatality rates in four classes. the ratio between the rate of patients in intensive care and the rate of 309 hospitalization gives in principle an indication of the severity of the patients in hospital. this hospitalization rate and severity rate show a weak negative correlation (r=-0.22), 311 indicating a decrease in the intensive care rate when the hospitalization rate is high. 312 this trend may be due to the saturation of intensive care units. the relationship 313 between hospitalization and severity could also be interpreted as a decrease in less 314 severe hospitalizations in order to be able to manage more severe cases when the 315 healthcare system is overloaded, which would result in an increase in lethality. nevertheless, in both cases, there is no correlation between the severity rate and the 317 case-fatality rate (r=-0.1), indicating that globally, the intensity of reanimation does not 318 impact the case-fatality rate. finally, the severity rate does not have a spatial 319 distribution corresponding to the increase in the hospitalization rate ( figure 12) . 320 . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the standardized case-fatality rates (slr) of the districts in france (0.04 for tarn-325 et-garonne to 0.26 for the vosges district) remain in a ratio of 0.3 to 1.6 compared with 326 the national average of 0.14, calculated by excluding districts under stress in order not 327 to take account of possible saturation of the care systems. the relationships between 328 morbidity rates and standardized case-fatality rates in france show a correlation 329 between these two indices, the average case-fatality rate for all districts being about 330 15% higher than the average rate calculated in the 20% of districts with the lowest 331 hospitalization rates. it is therefore very likely that the increase in hospital tension over 332 the period under consideration has increased the hospital case-fatality rate: for the 20 333 districts with the highest hospitalization rates (essentially located in the grand-est and 334 ile-de-france regions), the average case-fatality rate is 20 per cent higher than the 335 average for all districts, and 25 per cent higher than the average for all other districts 336 alone. it can be concluded that hospital case-fatality rates have increased the national 337 average case-fatality rate by district from 0.145 to 0.153. it can therefore be estimated 338 that 2,425 deaths (out of the 16,732 deaths due to covid-19 in hospital in france from 339 19 march to 8 may 2020, i.e. 15% of the total number of deaths) are due to the 340 saturation of the health system in the grand-est and ile-de-france regions. 341 . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted may 20, 2020. . https://doi.org/10.1101/2020.05.16.20104026 doi: medrxiv preprint there are clearly two risk profiles: 1/ the districts where a high rate of 342 hospitalization is coupled with a high case-fatality rate, and 2/ the districts where a low 343 rate of hospitalization is coupled with a high case-fatality rate. the first category 344 probably results from an increase of lethality due to saturation of the health care 345 system. the second category is probably linked to the opposite phenomenon: a low 346 hospital case-fatality rate which would have led to an increase in lethality due to a local 347 lack of healthcare access (e.g. medical deserts, poor hospital lethality preparation). it 348 has also been noted that all these late districts are located in essentially rural areas. 349 some districts in the south of france have both a very low rate of hospitalization and 350 a very low case-fatality rate (gironde, dordogne, gers, pyrénées orientales), as a 351 result of the low circulation of the virus and the effective response of the health system. 352 another particular case, is the one of the bouches-du-rhône, which appears with a 353 high hospitalization rate ( french average is therefore very significantly higher than the world average (p-value 380 < 10-6). even if we consider only the average case-fatality rate calculated only for the 381 french districts with the lowest hospitalization rates (thus not causing saturation of the 382 health care system), this average is still very significantly higher than the international 383 average (and the rates of most european countries, such as spain, 11.73, greece, 384 5.52, germany, 4.28, etc.) (table 4) . taking into account the quality of the healthcare 385 system in france (table 1) , it can be concluded that the difference between the case-386 . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. (which was not certified by peer review) the copyright holder for this preprint this version posted may 20, 2020. . fatality rate calculated for france and the case-fatality rates presented using 387 international who data is highly probably the result of a difference in the registration 388 of cases and/or deaths and not due to the quality of health care. these differences in 389 the counting of cases and/or deaths may be due to the hospitalization and screening 390 policy specific to each country as well as the ability or willingness to hospitalize more 391 non-severe forms, to the differences in case definition, or to insufficient quality of the 392 system for detecting and reporting cases and deaths. 393 this study shows that the higher case-fatality rates observed in france in certain 395 districts during the first wave of the covid-19 epidemic (data from 19 march to 8 may 396 2020) are mostly linked to the level of morbidity in the district, and therefore to the 397 congestion of the healthcare systems during the acute phases of the epidemic. when 398 the hospitalization rate is low, high case-fatality rates concern rural districts and could 399 be linked to health care access in these districts. 400 however, the increase in the standardized case-fatality rate due to exceptional 401 situations during epidemic peaks represents less than 10% of the average case-fatality 402 rate per district in france, and the hospital case-fatality rate without these districts 403 would be reduced from 0.153 to 0.145. this increase cannot therefore explain the 404 extent of the difference observed between the average case-fatality rate in france and 405 the average of the rates reported for all countries by international organizations or 406 information sites (who, wordometer, etc.). these differences probably stem from the 407 reporting of cases and deaths, which is uneven from one country to another, and not 408 from the care or treatment of patients during hospital stress due to epidemic peaks. 409 real estimates of 411 mortality following covid-19 infection 414 [mor 20] morteza abdullatif khafaie, fakher rahim. cross-country comparison of case 415 fatality rates of covid-19/sars-cov-2. osong public health and research 416 por 08] porta m. a dictionary of epidemiology. 5th ed an empirical estimate of the infection fatality 420 rate of covid-19 from the first italian outbreak using early data to estimate the actual 422 infection fatality ratio from covid-19 in france estimating the burden of sars-cov-2 in france key: cord-322943-lvdl7puw authors: lardon, zélie; watier, laurence; brunet, audrey; bernède, claire; goudal, maryvonne; dacheux, laurent; rotivel, yolande; guillemot, didier; bourhy, hervé title: imported episodic rabies increases patient demand for and physician delivery of antirabies prophylaxis date: 2010-06-22 journal: plos negl trop dis doi: 10.1371/journal.pntd.0000723 sha: doc_id: 322943 cord_uid: lvdl7puw background: imported cases threaten rabies reemergence in rabies-free areas. during 2000–2005, five dog and one human rabies cases were imported into france, a rabies-free country since 2001. the summer 2004 event led to unprecedented media warnings by the french public health director. we investigated medical practice evolution following the official elimination of rabies in 2001; impact of subsequent episodic rabies importations and national newspaper coverage on demand for and delivery of antirabies prophylaxis; regular transmission of epidemiological developments within the french antirabies medical center (armc) network; and armc discussions on indications of rabies post-exposure prophylaxis (rpep). methodology/principal findings: annual data collected by the national reference center for rabies nrcr (1989–2006) and the exhaustive database (2000–2005) of 56 armc were analyzed. weekly numbers of patients consulting at armc and their rpepand antirabies-immunoglobulin (arig) prescription rates were determined. autoregressive integrated moving-average modeling and regression with autocorrelated errors were applied to examine how 2000–2005 episodic rabies events and their related national newspaper coverage affected demand for and delivery of rpep. a slight, continuous decline of rabies-dedicated public health facility attendance was observed from 2000 to 2004. then, during the summer 2004 event, patient consultations and rpep and arig prescriptions increased by 84%, 19.7% and 43.4%, respectively. moreover, elevated medical resource use persisted in 2005, despite communication efforts, without any secondary human or animal case. conclusions: our findings demonstrated appropriate responsiveness to reemerging rabies cases and effective newspaper reporting, as no secondary case occurred. however, the ensuing demand on medical resources had immediate and long-lasting effects on rabies-related public health resources and expenses. henceforth, when facing such an event, decision-makers must anticipate the broad impact of their media communications to counter the emerging risk on maintaining an optimal public health organization and implement a post-crisis communication strategy. media-communicated health alerts are being used more-and-more frequently by public health decision-makers to prevent consequences of a sudden event, such as, emerging and episodic zoonotic diseases. the medical community must now consider these communications to be preventive intervention tools for public health officials [1] [2] [3] . obviously, as during any effective health intervention, undesired effects may also occur, such as rapidly rising numbers of potential cases to treat, leading, in turn, to health-resource saturation, especially if the pathogen involved is rare [4, 5] . rabies is a viral encephalitis [6] that is considered to be a reemerging zoonosis throughout much of the world [7] . in western europe, rabies in non-flying terrestrial mammals was a well-known illness that has now become a rare disease, because many countries have succeeded in eradicating it. the major risk of rabies is now due to translocation of infected animals, mainly dogs, from rabies-enzootic areas and humans with rabies infection acquired abroad [8] . although untreated rabies is invariably fatal, death can be avoided by proper administration of rabies postexposure prophylaxis (rpep), e.g., antirabies vaccine, with or without antirabies immunoglobulins (arig), before disease onset [6] . thus, rapid identification of individuals potentially exposed to rabies is critical and media alerts can be extremely useful to identify people who were in contact with the rabid animal. in france (60,000,000 inhabitants, 675,417 km 2 ), primary health-care management of patients seeking rpep is delivered through an official national network of antirabies medical centers (armc), which are distributed throughout the country. rpep is administered, predominantly according to the zagreb schedule, to people bitten by an animal suspected of being infected with rabies or exposed to its saliva. clinicians conduct a risk assessment for each exposed patient, and decide to administer rpep according to the general recommendations, epidemiological data and grade of the bite [9] . the french network for rabies prophylaxis provides exhaustive national data collected by armc [10] , and laboratory diagnoses of humans suspected of having rabies [11] and animals suspected contaminating humans. from 1968 to 1998, a period during which rabies was endemic in french foxes, more than 45,600 animals were diagnosed as rabid. in 2001, france was declared free of rabies in non-flying terrestrial mammals based on world organisation for animal health (oie) criteria and, as a consequence, the number of rpep began to decline progressively. however, in summer 2004, one imported rabid dog generated unprecedented media communications by the public health director, whose official press release, dated 31 august 2004, warned, ''at least, nine people are at risk of death and are actively and intensively being sought by the health authorities…'' during this episode, antirabies vaccine stocks in armc were almost exhausted, leading to a temporary marketing license for the multidose verorab vaccine (sanofi pasteur), which had not previously been authorized in france. that arig supplies were dangerously low is illustrated by the postponement of arig injections in some armc until day 7 after starting rpep [12, 13] for several patients. controlling rabies reintroduction and communicating the risk of rabies spread remain a challenge to public health officials in rabies-free areas. in this study, we analyzed why and how the french rabies-control organization became so oversaturated. in particular, we examined the impact of newspaper reports on the numbers of patients consulting at armc, and their rpep and arig prescriptions. this research has complied with the french national guidelines and institut pasteur policy. the analysis of data collected by the national reference center for rabies (nrcr) from the amrc was done anonymously and approved by the commission french veterinary and human authorities work in close collaboration to detect cases and organize the medical responses to rabies (figure 1) , with a territorial network of 96 veterinary services and 74 armc disseminated throughout continental france, in 2004 ( figure 2 ). on the one hand, each animal responsible for human exposure is confined under veterinary surveillance. if dead and for whatever the reason, diagnostic laboratory tests are conducted at the nrcr, institut pasteur, paris, france. on the other hand, armc are the only primary care centers allowed to prescribe rpep. for each patient, a standard case-report form (table s1 ) is systematically filled out describing important epidemiological features, such as geographic location, consultation date, type of exposure, animal species, contact date with the animal, medical decision concerning rpep. based on the data collected by armc, annual reports are written, which describe the patients visiting armc and those receiving rpep (http://www.pasteur.fr/sante/clre/cadrecnr/rage/rageactualites.html). our analysis of the behavior patterns of patients consulting armc, and the rpep and arig prescribed to them between 1989 and 2006 was based on those annual data. among the 74 french armc, 56 systematically entered their data into the nrcr database between 2000 and 2005. the following statistical analysis is based on the exhaustive weekly information provided by these 56 armc. the armc network also constitutes an effective communication infrastructure coordinated by the nrcr, including conference calls and regular exchanges of information via the internet. when rabies is suspected in a human, biological specimens are sent to the nrcr. articles on rabies-related news published in three major national daily newspapers, le monde, le figaro and libération, were retrieved from the french association for auditing media circulation: an on-line service: http://www.factiva.fr. weekly numbers of patients consulting at armc, as a function of the date each was in contact with a potentially rabid animal, were used to construct times series. autoregressive moving average (arma) [14] modeling was used to determine the significance of event-associated modification of armc weekly patient numbers and its duration. because several known events could have affected the series, a step-by-step procedure was undertaken [15, 16] . before the onset of event #2, trend and/or seasonality were estimated and removed, so that the time series was obtained in a stationary mode and, autoregressive integrated moving-average (arima) modeling was done using box-jenkins procedure from sas/ets [17] ). the model was then used to predict armc consultations and their 95% confidence intervals (95% ci). an event was considered to have an impact when the number of consultations during 2 consecutive weeks exceeded the upper 95% ci. observed values were then replaced by forecasts, to obtain analyses of the subsequent weeks. similarly, 2 consecutive weeks within the 95% ci defined the end of the event's impact period. relative differences between observed and predicted values were calculated. for impacting events, the number of cases attributed to the event (ncae) was estimated by subtracting the prediction from the observed data during the impact period. an increase rate rabies has been eliminated from a large part of the european union and, thus, any newly imported cases threaten its reemergence. the 2000-2005 data derived from the exhaustive surveillance system implemented in france was analyzed to evaluate the impact on demand for and delivery of antirabies prophylaxis following introduction of five rabies-infected dogs and one infected human into this rabies-free area. using these events, we were able to illustrate the difficulties encountered in reducing the demand for and prescription of postexposure rabies prophylaxis in this context of episodic importation. moreover, we highlighted the need for public health decision-makers to anticipate the broad spectrum of consequences of their media communications and to prepare appropriate responses (in terms of health resources) to maintain an optimally effective public health organization after importation of an exotic infectious agent or its emergence. these responses are particularly relevant in the context of limited availability of rabies post-exposure prophylaxis, especially antirabies immunoglobulin. (ir) was then calculated as the ratio of the ncae/number predicted for the impact period. with the aim of evaluating potential repercussions of an identified event impacting on rpep prescriptions, two other time series were investigated: the weekly rpep rate, defined as the number of rpep prescribed/the number of consulting armc patients, e.g. rabies vaccine with or without arig; and the weekly arig rate, corresponding to the ratio of the number of arig/the number of consulting armc patients. during the period associated with modified armc weekly numbers, weekly rpep and arig rates and mean numbers of consultations were analyzed using regression with autocorrelated errors to account for the regression residuals (arima procedure). to explore whether care provided by the armc might be influenced by experience in previous french endemic enzootic areas, we divided the country into three areas based on the french administrative regions: area 1, the former enzootic rabiesinfected-fox region from 1968 to 1998; area 2, a region that has always remained rabies-free, and area 3, the region where event #6 occurred ( figure 2 ). all analyses were performed using r (www.r-project.org) and sas software. after the reintroduction of rabies into france in 1968, the number of rabid animal cases increased to reach a maximum of 4,212 cases in 1989 [18] , followed rapidly by a maximum of 9,763 rpep prescribed for 15,948 patients consulting at armc recorded in 1990 ( figure 3 ). in 2001, france was declared free rabies reemergence and antirabies prophylaxis www.plosntds.org of rabies in non-flying terrestrial mammals based on oie criteria [19] and, as a consequence, the number of patients consulting armc and receiving rpep began to decline progressively to respective minima of 7,788 and 3,378 in 2003 ( figure 3 ). however, the numbers of patients consulting at armc and given rpep suddenly rose in 2004. therefore, 2000-2005 data were further investigated using arima modeling to describe in greater detail the trends observed. between 1 january 2000 (week 1) and 31 december (week 312) 2005, five rabid dogs illegally imported from morocco and one rabies-infected human from gabon were detected in france. during the period examined, the first event #1 dog (5 months old) was confirmed as being rabid in may 2001 (week 74) and the second, event #2 dog (3 months old) in september 2002 (week 139); they entered france from morocco, 2 months and 2 weeks before their deaths, respectively. the human case (event #3) was a 5-year-old boy, who traveled from gabon and died 2 months later, in october 2003 (week 199) [20] . event #4, #5 and #6 dogs were diagnosed as being rabid, respectively, in february 2004 (week 213), may 2004 (week 229), and august 2004 (week 243) [21] . event #6 was a 4-month-old puppy, illegally imported by car from morocco to bordeaux, france, via spain, who died of rabies in august 2004 (week 243); he was not officially vaccinated. between 1 january 2000 and 31 december 2005, 56,924 rabiesexposed individuals in france (all patients exposed abroad were excluded from the analysis) consulted in an armc, among whom 56,446 had valid exposure dates and bite/contact locations. among them, 50,930 had valid consultation dates and 56,406 had valid treatment information (figure 4 ). because the data presented 52-week seasonality, the time preceding event #1 was too short to be analyzed. in such a case, box and jenkins recommend using at least two seasonality periods to calibrate the model [14] . data analyses concerning events #1, #2, #4 and #5, corresponding to rabid dog importations, were simple and rapidly done, as these dogs had had no known contact with animals and humans other than their owners during their communicable risk periods. as a consequence, events #2, #4 and #5 were not reported in the major national newspapers and were not associated with any significant increase of armc activity. in contrast, events #3 and #6 were reported in 6 and 54 published articles retained for this study, respectively, and significantly affected the numbers of patients consulting at an armc ( figure 5 ). until event #3 (october 2003), the weekly number of patients consulting an armc declined significantly (slope = 20.34; p,0.0001), with 52-week seasonality that peaked during the summer ( figure 5 ). in october 2003, the weekly number of armc patients was significantly higher than the predicted number during the 6 weeks surrounding event #3 (weeks 198-203), with an estimated ncae of 355 (ir = 54.7%, 95% ci = 30.0-83.0). furthermore, event #3 was followed by a significant flattening of the decreasing slope of armc activity (20.23 versus 20.34; p = 0.0003). no rpep-or arig-rate modification associated with event #3 was observed. in the summer of 2004 (event #6), the weekly number of armc patients differed significantly from the predicted number during the 26 weeks surrounding it (weeks 238-263). the total 26week number of additional armc patient load was estimated at 2,928 (ir = 84.0%, 95% ci = 57.0-123.3) over the model predicted 3,486 ( figure 5 ). during that period, the observed mean rpep and arig rates were significantly higher than those recorded during the period preceding event #6, ir = 19.7% and 43.4%, respectively ( table 1) . the slopes of the armc-consultation decline after week 263 and before week 238 were estimated at 20.12 and 20.23, respectively; p,0.001. surprisingly, between weeks 264 and 312, the mean rpep rate remained persistently and significantly higher rabies reemergence and antirabies prophylaxis www.plosntds.org than before the reference period, as did the arig rate, which was more than two-fold higher than before week 237 ( table 1 ). the increased number of patients consulting at an armc in response to the newspaper articles concerning event #6 peaked at the same time as the media coverage in the three different french areas defined according to their rabies experience ( figure 6a ). in area 3, the exposure dates reported by armc patients corresponded to the risk period coinciding with the dog's movements and infectivity, whereas in areas 1 and 2, patients reported exposure dates more compatible with newspaper coverage than with the risk period ( figure 6b ). france progressively eliminated rabies in foxes and became rabies-free for indigenous non-flying terrestrial mammals in 2001 [19] . consequently, use of public health facilities dedicated to the disease decreased steadily from 1990 until 2003, suggesting a continuous impact of rabies elimination on related public health resources and expenses. however, the very mild decline of the 2000-2003 slope probably reflects the difficulties in convincing the public and adapting medical practice to the changing risk. although elimination of rabies in foxes reduced the number of rabid pets and other domestic animals, and thus exposure to rabies, pet bites continue. importation of rabid animals and infected travelers returning from abroad also regularly challenge the french public health organization of rabies control. therefore, the number of rpep prescriptions and the associated costs will not decline significantly until there is adequate assurance that the probability of a pet being rabid is sufficiently low that such therapy is not warranted, even when the pet's status cannot be verified [22, 23, 24] . regardless of potential french specificities, public health decision-makers are obliged to consider such potential events and their ensuing demand on medical community resources when attempting to predict and maintain the efficacy of rabiescontrol policies even in rabies-free countries [24] [25] [26] [27] [28] . among the six rabies events occurring during 2000-2005 in france, only two significantly affected armc activities and rpep rates. the human case imported from gabon in 2003 (event #3) was associated with enhanced armc activity during a brief period and also changed armc's declining activity, which had been observed since 2000. the boy's demise was reported 6 times in the newspapers, further confirming that ''death makes news'' for rare and acute diseases [29] . in contrast, the illegally imported . rabies-exposure notifications to armc and numbers of rpep prescribed to exposed patients in france, 1989 france, -2006 . these data are from the annual nrcr report (http://www.pasteur.fr/sante/clre/cadrecnr/rage/rage-actualites.html). doi:10.1371/journal.pntd.0000723.g003 rabies reemergence and antirabies prophylaxis www.plosntds.org rabid dog from morocco in august 2004 (event #6) had a significant and rapid impact on rabies public health resources. indeed, the critical shortage of prophylactic drugs resulted from the 84% ir of patients consulting at an armc with a 62.5% rpep rate for those patients over 26 weeks. this influx explains the bottleneck observed in armc. similarly, laboratory rabiesdiagnosis workload for animals increased by .40% during the same period (data not shown). to comply with the threatened shortage of rpep and arig due to the cumulative effect of enhanced patient influx and their more frequent prescriptions, a specific communication strategy was established for the armc network to provide information concerning the evolution of the epidemiological situation and to recall the indications of rpep. this information was disseminated via the websites of the nrcr, the ministry of health (moh), the national institute for health surveillance and the ministry of agriculture, which were regularly updated as of 28 august, fax on 2 september, and phone conferences on 3 and 9 september. to complete this plan, temporary licensing of a multidose vaccine (verorab, sanofi pasteur) was accorded and arig injections were postponed, as necessary, in accordance with who guidelines [12] . unfortunately, it was not feasible to quantitatively analyze the extent of that adaptation. however, rpep and arig never became completely unavailable. notably, the risk of a potential arig shortage in the event of an unplanned increase of demand or a limitation of supply is shared by many countries in europe and on other continents [30, 31] . compared to similar events occurring during 2000-2005 in france, event #6 has several particularities. while only restricted contacts with humans (owners, neighbors…) were suspected for cases #2, #4 and #5, the event #6 dog traveled through southwestern france during the communicable risk period, and had been roaming unleashed at three large summer music festivals, each with at least 10,000-20,000 participants [21] . according to immediate inquiries made by veterinary and medical services, this trajectory potentially led to extensive contacts between the rabid dog and humans and animals. therefore, the public health authorities' concern triggered extensive media alerts. first, the moh wanted to identify and contact each individual with confirmed contact with the event #6 dog. national and local authorities coordinated several news conferences and newspaper reports to inform the french rabies reemergence and antirabies prophylaxis www.plosntds.org population about the risk and recommendations concerning errant dogs in general, and how to react to potential exposure to a rabid dog. a european-wide alert was launched through the european warning and response system. second, beginning in early september 2004, this intensive communication frenzy of 54 newspaper articles heightened public awareness of the rabies risk. third, additional public concern might also have been heightened by controversies surrounding the crisis management. notably, event #6 occurred just before the annual opening of hunting season, in a strongly traditional hunting region. an initial decision was made to forbid hunting with dogs in the counties where the rabid dog had traveled during his infectious period. that restriction led to a passionate public debate, angering hunters and ending with hunting organizations successfully blocking the ban. fourth, public health authorities decided to eradicate freeroaming dogs. finally, press releases issued by the minister of rural affairs and the moh were contradictory concerning the implementation of mandatory antirabies vaccination of dogs and cats. the constant media attention drawn by these different players during event #6 may have contributed to enhancing the sense of rabies risk, thereby prompting people to associate dog bites with we only examined national newspaper stories available in factiva but not local newspaper reporting or television, radio and internet stories, and, thus, probably underestimated the global coverage of these episodes. in response to national newspaper coverage, people who are far from the event location can become concerned and start taking precautions as if they were in the affected area [3, 4, 32] . this phenomenon is particularly well illustrated by event #6, for which exposure dates reported by patients consulting at amrc in areas 1 and 2 corresponded to the period of newspaper coverage rather than to the risk-oftransmission period during the dog's movements. lastly, long-term modifications of armc activity and rpepand arig-prescription rates were observed. in particular, 2005 rpep and arig rates (arima study herein) and even those for 2006 had not yet returned to 2003 levels. this finding strongly suggests a persistent and unjustified heightened perception of the risk by individuals and physicians, even those specialized in rabies treatment, and this despite regular information provided by the nrcr to the armc network and a rapidly controlled situation with no recorded secondary animal and human cases during the following 2 years. in conclusion, event #6 and its associated national newspaper coverage profoundly perturbed health services, with excessive consulting at armc and durably increased antirabies drug rates for several months, along with more animal diagnostic testing. this crisis highlighted a lack of experienced manpower and insufficient vaccine stocks. outbreaks of emerging and/or deadly infections, like severe acute respiratory syndrome [34] [35] [36] [37] [38] , anthrax [39, 40] and rabies (herein), have shown that media messages dramatically influence both the public's and health-care workers' perceptions of the risk with potential implications for health-care resources. our observations underscore to what extent, under such circumstances, public health decision-makers have to anticipate the depth and scope of potential consequences of emerging or reemerging infectious diseases and their related press communications, and the need to prepare appropriate responses to keep the public health organization effective. it also illustrated that, despite communication efforts implemented by the french public health authorities and messages released through the armc network, long-term modifications of armc activities and prescriptions were observed, further emphasizing that a post-crisis communication strategy is essential. table s1 case-report form for human exposure to rabies used in france. since 2006, collection and dissemination of information are made by filling out questionnaires available at a centralized online site named voozanoo (http://www2.voozanoo.net/tikiindex.php?page = what%27s+voozanoo). found at: doi:10.1371/journal.pntd.0000723.s001 (0.07 mb doc) best practices in public health risk and crisis communication communicating the threat of emerging infections to the public six propositions on public participation and their relevance for risk communication the public's response to severe acute respiratory syndrome in toronto and the united states communicating the risks of a new, emerging pathogen: the case of cryptococcus gattii rabies and other lyssavirus diseases estimating the public health impact of rabies rabies in europe in 2005 expert consultation on rabies epidemiology and prophylaxis of rabies in humans in france: evaluation and perspectives of a twenty-five year surveillance programme a reliable diagnosis of human rabies based on analysis of skin biopsy specimens rabies vaccines what is an acceptable delay in rabies immune globulin administration when vaccine alone had been given previously time series analysis : forecasting and control revue méthodologique de quelques techniques spécifiques à l'analyse des séries temporelles en épidémiologie et santé publique a time series construction of an alert threshold with application to s. bobimorbificans in france the theory and pratice of econometrics que penser de la rage en 1990? bulletin epidémiologique de la fox rabies in france la rage : une maladie encore présente en france! an imported case of canine rabies in aquitaine: investigation and management of the contacts at risk rabies postexposure prophylaxis in returned injured travelers from france, australia, and new zealand: a retrospective study rabies postexposure prophylaxis potential cost savings with terrestrial rabies control economics of human and canine rabies elimination: guidelines for programme orientation cost effectiveness of rabies post exposure prophylaxis in the united states rabies control in the republic of the philippines: benefits and costs of elimination rabies exposures, post-exposure prophylaxis and deaths in a region of endemic canine rabies death makes news: the social impact of disease on newspaper coverage is there a need for anti-rabies vaccine and immunoglobulins rationing in europe appropriateness of rabies postexposure prophylaxis treatment for animal exposures the power of the pen: medical journalism and public awareness what are the roles and responsibilities of the media in disseminating health information media effects on students during sars outbreak the impact of the sars epidemic on the utilization of medical services: sars and the fear of sars sars epidemic in the press responding to global infectious disease outbreaks: lessons from sars on the role of risk perception, communication and management representations of sars in the british newspapers anthrax-related panic is more dangerous than the disease anthrax 2001: observations on the medical and public health response the authors thank all the armc personnel, who collected and send their data to the nrcr, for their contribution. we are grateful to janet jacobson for expert editing of the manuscript. conceived and designed the experiments: dg hb. performed the experiments: zl dg hb. analyzed the data: zl lw ab cb dg hb. contributed reagents/materials/analysis tools: lw mg ld yr dg hb. wrote the paper: zl lw ld dg hb. key: cord-335215-h9p4kmss authors: follet, jérôme; guyot, karine; leruste, hélène; follet-dumoulin, anne; hammouma-ghelboun, ourida; certad, gabriela; dei-cas, eduardo; halama, patrice title: cryptosporidium infection in a veal calf cohort in france: molecular characterization of species in a longitudinal study date: 2011-12-02 journal: vet res doi: 10.1186/1297-9716-42-116 sha: doc_id: 335215 cord_uid: h9p4kmss feces from 142 animals were collected on 15 farms in the region of brittany, france. each sample was directly collected from the rectum of the animal and identified with the ear tag number. animals were sampled three times, at 5, 15 and 22 weeks of age. after dna extraction from stool samples, nested pcr was performed to amplify partial 18s-rdna and 60 kda glycoprotein genes of cryptosporidium. the parasite was detected on all farms. one hundred out of 142 calves (70.4%) were found to be parasitized by cryptosporidium. amplified fragments were sequenced for cryptosporidium species identification and revealed the presence of c. parvum (43.8%), c. ryanae (28.5%), and c. bovis (27%). one animal was infected with cryptosporidium ubiquitum. the prevalence of these species was related to the age of the animal. c. parvum caused 86.7% of cryptosporidium infections in 5-week-old calves but only 1.7% in 15-week-old animals. the analysis of the results showed that animals could be infected successively by c. parvum, c. ryanae, and c. bovis for the study period. c. parvum gp60 genotyping identifies 6 iia subtypes of which 74.5% were represented by iiaa15g2r1. this work confirms previous studies in other countries showing that zoonotic c. parvum is the dominant species seen in young calves. cryptosporidium is a genus of protozoan parasites infecting a wide range of hosts [1] . all groups of vertebrates are susceptible to cryptosporidium infection worldwide. this parasite is the etiological agent of cryptosporidiosis, which is mainly characterized by diarrhea in humans and livestock. massive outbreaks of enteritis in people such as in milwaukee, wisconsin (usa) have increased public awareness of this parasite [2] . in humans, the prevalence and severity of infection increase in immunodeficient individuals such as aids patients. in immunocompetent patients, the disease is self-limited [3] . no drug therapy is yet available and the high resistance of oocysts to environmental conditions and chemical treatment make cryptosporidiosis difficult to control [4] . cattle have been considered to be a primary reservoir for cryptosporidium oocysts for zoonotic c. parvum [5] . these animals could be a risk factor via environmental contamination from their manure being spread on farmland or their grazing on watersheds [6] . on farms, transmission of cryptosporidium spp. can result from ingestion of contaminated food or water, by direct transmission from host to host, or through insect vectors [7] . in cattle, infection by cryptosporidium spp. was first reported in 1971 [8] . since vaccines have become commercially available against escherichia coli k99, rotavirus, and coronavirus, cryptosporidium has emerged as the main neonatal diarrheic agent in calves [9] . in farm animals, the economic impact is related to morbidity, mortality and growth retardation [10] . among the 24 species previously described (if the three fish species are accepted without complete genetic characterization) [1, [11] [12] [13] , c. parvum, c. bovis, c. ryanae and c. andersoni usually infect cattle. c. parvum has zoonotic potential and is a frequent cause of human cryptosporidiosis [14] . c. bovis and c. ryanae have not been found in humans and there is only one description of c. andersoni in a patient [15] . recent reports have described an age-related distribution of these aforementioned species in dairy cattle on the east coast of the united states [16] [17] [18] , india, china, georgia [19] , malaysia [20] , and denmark [21] . the most prevalent species were c. parvum in preweaned calves, c. ryanae and c. bovis in postweaned calves and c. andersoni in adult cows [16, 17] . in france, previous studies on the prevalence of cryptosporidium in cattle were based on microscopic determination [22] or coproantigen detection using elisa [23] . these studies on dairy calves reported a within herd prevalence of cryptosporidium without identifying species or the relation to the host's age. the present study was conducted in 15 farms in brittany, france to determine the prevalence of cryptosporidium in veal calves. we used genotyping and subtyping for the molecular study of cryptosporidium isolates. follow-up of the same animal allowed us to determine the outcome of the infection and the age distribution of cryptosporidium species. fifteen fattening units in brittany (france) were included in this work. they belonged to three industrial veal producers representative of integrators in france and did not present any known history of cryptosporidium infection. these farms were located in four administrative regions ( figure 1 ): côtes d'armor (ca1-ca3), morbihan (mo1), ile-et-vilaine (iv1-iv5), and mayenne (ma1-ma6). during the summer and autumn of 2007, all farms were visited three times and fecal samples were taken from 142 animals exhibiting diarrhea at the age of 5 weeks old. calves arrived in fattening units at the age of 2 weeks old and were confined in small groups from four to six animals per pen. because of a concomitant welfare study [24] , calves had to stay 2 to 3 weeks without any external stress despite the farmer's presence. at the age of 22 weeks old, calves were finally sent to the slaughterhouse. consequently, sampling was done at the ages of 5 weeks, 15 weeks, and 22 weeks (table 1 ). these points of sampling corresponded to the beginning, the middle and the end of the fattening period. fecal samples were collected and shipped by a veterinarian. collectors respected the following criteria: use of a single pair of latex gloves per animal, a single plastic sterile cup per animal, and collection of at least 5 g of feces per sample. feces were collected directly from the rectum of each animal and stored at 4°c in potassium dichromate (2.5% wt/vol) until processed. cups were capped, labeled with the animal's ear tag number, and accompanied by a form recording the date of sampling, the animal's sex, breed, identification number, and the mean age of the herd. after washing steps in water to eliminate potassium dichromate from the samples, dna was extracted according to the method previously described [25] without the cetyl trimethylammonium bromide (ctab) and polyvi-nylpyrrolidone (pvp) treatment steps. an 18s rna gene fragment was amplified by nested pcr according to xiao et al. [26] . the partial gp60 gene was amplified according to gatei et al., [27] . pcr products were analyzed on 2% agarose gel and visualized by ethidium bromide staining. to ensure purity and limit the presence of pcr inhibitors, all pcr-negative samples were reprocessed. samples were treated for oocyst purification by immunomagnetic separation (dynabeads ® anti-cryptosporidium, invitrogen ™, norway) according to the manufacturer's instructions. these samples were finally processed as previously for dna extraction and pcr amplification. cryptosporidium species identification pcr products were purified on an ultracel ym50 membrane (microcon, millipore, bedford, ma, usa) according to the manufacturer's instructions. dna sequencing reactions were performed using internal primers of the nested pcr with the abi prism big dye terminator cycle sequencing kit (applied biosystem, foster city, ca, usa). capillary electrophoresis was performed by genoscreen (lille, france). sequences were analyzed using blast at ncbi [28] . the prevalence of cryptosporidium infection on 15 farms from four administrative regions in brittany (france) was studied ( figure 1 ). all cryptosporidium-positive specimens generated the expected ssu-rna products in nested pcr and revealed that no farm was free of cryptosporidium. the molecular analysis of 422 fecal samples revealed that 147 (34.8%) were positive for cryptosporidium. as shown in table 1 , the overall prevalence of infected animals was 70.4% (100/142) and ranged from 10% on a farm in morbihan (mo1) to 100% on farms in ile-et-vilaine (iv1, iv3) and in mayenne (ma5). amongst the specimens sampled from 5-week-old and 15-week-old animals, cryptosporidium prevalence was 47.9% and 42.1%, respectively (range, 0%-87.5%). in 22-week-old calves, the prevalence decreased to 14.3% (range, 0%-37.5%). the prevalence of infection decreased as the age of the calves increased. for species identification, the 147 positive nested pcr products were sequenced. sequence analysis from 137 readable electrophoregrams revealed the presence of c. parvum, c. bovis, and c. ryanae. one additional cryptosporidium genotype showing 99% identity with cryptosporidium ubiquitum (eu827413) (previously identified as * a calf is considered to be positive if at least one out of the three samples is positive. **the number of animals is 9 because one calf died between the age of 5 and 15 weeks. cryptosporidium cervine genotype [13] ) was detected in one calf. this sequence was deposited in genbank under the accession number gu124629. sixty (43.8%) samples were identified as c. parvum as follows: forty-six sequences had 100% identity with the genbank af093490 nucleotide sequence, 11 had 100% identity with the af308600 nucleotide sequence and three had 99% identity compared to both references. these sequences were deposited in genbank under the accession numbers gu124615 to gu124617. for the other positive specimens, 39 (28.5%) were identified as c. ryanae (previously described as cryptosporidium deer-like genotype). thirtyone of these had 100% identity with the ay587166 sequence [17] and eight were 99% identical to this reference. these nucleotide sequences were deposited in gen-bank under the accession numbers gu124621 to gu124628. for the last positive samples, 37 (27%) had an identical nucleotide sequence with c. bovis (genbank accession number, ay120911) formerly known as the cryptosporidium bovine b genotype. within these sequences, 34 had 100% identity to the reference deposited in genbank, three sequences had 99% identity. these last sequences were deposited in genbank under the accession numbers gu124618 to gu124620. prevalence of c. parvum, c. ryanae, and c. bovis in relation to calf age the distribution of cryptosporidium species identified in animals at the age of 5, 15, and 22 weeks is shown in figure 2 . the prevalence of each species changed with the age of the calves. c. parvum prevalence was 86.7% in the 5-week-old calves and decreased to 1.7% in 15-week-old animals. this species was not identified in 22-week-old calves. c. ryanae and c. bovis were identified in 5-weekold calves in 4.4% and 1.5% of the specimens, respectively. the prevalence of these species in 15-week-old animals increased to 44.1% and 45.7%, respectively. this prevalence evolved to 50% and 45% in 22-week-old animals. the presence of one, two, or three species of cryptosporidium was determined in each animal (n = 91) for which the sequences were readable in all positive samples. three calves positive for c. parvum at the age of 5 weeks were excluded because cryptosporidium species could not be identified in all of the following samples collected in these animals. as shown in in the time lapse of this study, 34% of the animals (31/91) were found to excrete two different species of cryptosporidium successively. indeed, 13.2% (12/91) produced c. parvum and c. ryanae, 12.1% (11/91) excreted c. parvum and c. bovis, and 8.8% (8/91) excreted c. ryanae and c. bovis. finally, 2.2% (2/91) of the animals studied were detected to produce c. parvum, c. ryanae, and c. bovis. the subtyping analysis was performed on c. parvum positive specimens. from 60 targeted samples, 51 could be used for sequence analysis. as shown in table 3 , all alleles identified belong to the iia family. the most common subtype iiaa15g2r1 (100% identity with reference strain ab 514090) was found in 38 out of 51 samples (74.51%). six samples (11.76%) were typed as subtype iiaa17g1r1 (100% identity with reference strain gq983359), three samples (5.89%) as subtype iiaa16g3r1 (100% identity with reference strain dq192506) and two samples (3.92%) as subtype iiaa16g2r1 (100% identity with reference strain dq192505). finally one sample (1.96%) was subtyped as iiaa16g1r1 (100% identity with reference strain dq192504) and another one (1.96%) as subtype iiaa13g1r1 (100% identity with reference strain dq192502). discussion calves under 1 month of age are frequently infected with cryptosporidium sp [29] which results in economic loss [10] . in france, up to date, the prevalence of cryptosporidium in diarrheic calves has been studied only by elisa and microscopic strategies [22, 23, 30] . no data are available on a molecular basis to study cryptosporidium species in calf herds in that country. the present study based on 18s rdna and gp60 gene analysis is the first in france to include molecular characterization to describe the prevalence and the host age related susceptibility to different cryptosporidium species after a follow up of the same animal. our results showed that all fifteen farms were contaminated with cryptosporidium. the parasite prevalence on farms ranged from 10% to 100% of the sampled animals. this observation was in accordance with results in michigan (usa) where this parameter ranged from 0% to 100% [31] . the prevalence of 70.4% obtained in this work tended toward the upper end of the scale compared to other investigations done in france which ranged from 15.6% in beef herds [30] to 95% in suckling calves [23] and in other european countries where prevalence ranged from 3.4% to 96% [32, 33] . however, the sampling program did not allow the study of animals under 5 weeks of age. indeed, the animals arrived in these structures at the age of 2 to 3 weeks and farmers did not allow sampling before two complete resting weeks for each animal. therefore, our results could underestimate the real prevalence as huetink et al. showed that the percentage of parasite excreting animal declines after the third week of age [34] and that the first peak of prevalence is at the age of 15 days [17] . in our study, the higher prevalence of cryptosporidiosis was observed in calves 5 weeks old (47.9%) and the lowest (14.3%) in the 22-week-old animals. this observation shows that prevalence of cryptosporidium infection decreases with increasing age of the cattle in france as in many other countries [17, 19, [33] [34] [35] [36] [37] [38] . additionally, our data confirmed the presence in france of a host age-related susceptibility to the infection with different cryptosporidium species. c. parvum was predominantly detected in 5-week-old calves (86.7%) compared to c. ryanae or c. bovis detected in 4.4% and 1.5% of the positive samples respectively. it is noteworthy that these results are very similar to data obtained in ireland on calves under 30 days of age with 95%, 3.6%, and 1.3% of prevalence of the same species, respectively [39] and in the uk on animals over 3 weeks old with 93% c. parvum, 6% c. bovis, and 2% c. ryanae [40] . in contrast to previous studies [17, 41] , c. ryanae and c. bovis were found with similar prevalence predominantly in 15 week and 22 week old calves. this association between the age of the cattle and the cryptosporidium species identification has been supported by several studies [17, 19, 21, 38, 40] but different reports suggest that cryptosporidium species repartition regarding the age of the host could be due to a technical artifact. despite the fact that the methodological strategy based on pcr using genus specific primers and partial direct sequencing of the 18s rdna is commonly used to identify cryptosporidium species [42] , this molecular tool is limited in the case of mixed infections. feng et al., [19] suggested that the important shedding of c. parvum in preweaned calves had probably masked the concurrent infection of these animals by c. bovis or c. ryanae. furthermore, previous reports suggested that a dominant cryptosporidium species in a sample can be preferentially amplified by pcr [43, 44] . it is noteworthy that this situation of mixed cryptosporidium species infection in farm animals would be more prevalent than originally believed [45] [46] [47] . mixed cryptosporidium species could also explain sequencing difficulties encountered in this work. the simultaneous presence of several species in the same sample could lead to amplification and sequencing of different genetic fragments leading to unreadable superimposition of electrophoregrams. consequently, in our work based on the utilization of cryptosporidium generic primers, the amplification of a single fragment with a single sequence is not conclusive evidence that the sample contains only a single species. however, based on our results, it is possible to confirm the predominance of different species of cryptosporidium by group of age among the calves. particularly, our data showed that animals can be sequentially infected with c. parvum, c. ryanae and c. bovis as well as c. parvum, c. bovis and c. ryanae. this observation provides evidence that a previous infection with c. parvum did not protect calves against an infection with other cryptosporidium species. fayer et al. suggested that the peak of cryptosporidiosis prevalence in young calves could reflect the immaturity of the immune status [48] . it was also suggested that the low excretion of c. parvum oocysts in older calves might be related to the development of immunity that also protected the animal against a secondary challenge [49] . it has been reported that immunity arises in the first two weeks after infection [50] . interestingly, fayer et al. [51] described that calves previously challenged with c. parvum were able to excrete oocysts after a second challenge with c. bovis but not with c. parvum. the authors concluded that immunity to c. parvum was not extended to c. bovis. consistently, in our study, the presence in the same animal during sequential sampling of c. parvum, c. bovis and c. ryanae suggests that immunity against c. parvum and against c. bovis did not extend to c. ryanae. furthermore, the observation that one animal excreted sequentially c. parvum, c. ryanae and c. bovis suggests that immunity against c. ryanae did not extend to c. bovis as well. finally, the risk to human health posed by cryptosporidium infected cattle in france was assessed. the detection of c. ubiquitum (a rare infectious agent detected in humans [52] ), c. ryanae and c. bovis (which are mainly specific for cattle) led to consider that the 22-week-old calves are not likely a public health concern. however, the major detection of c. parvum, a prevalent zoonotic species, in 5-week-old calves was in agreement with the report of atwill et al., who considered that the contribution of cattle to human cryptosporidiosis is limited to calves under 2 months of age [53] . to determine c. parvum subtypes, the sequence analysis of a fragment of the gp60 gene was done. our results show that in the region of brittany, all identified c. parvum gp60 subtypes belonged to the iia family which was previously found in both animals and humans [42] . particularly, human infections with the iia subtype are commonly seen in areas with intensive animal production [54] . among the 48 gp60 subtypes formerly described in cattle [55] , only six were identified in this work, being iiaa15g2r1 the most commonly found. this subtype has been widely reported in calves and humans in different countries such as in portugal [54] , slovenia [56] and the netherlands [57] . this observation confirms previous works and suggests a zoonotic transmission of the parasite also in this region. it is noteworthy that the three predominant subtypes (iiaa15g2r1, iiaa17g1r1, and iiaa16g3r1) found in this work were also described in cattle with an equivalent distribution in the netherlands [57] and england [40] . thus, the subtype iiaa15g2r1 was found in 74.5% of the samples in this work, 68.9% in the netherlands and 68.6% in england. the iiaa17g1r1 was identified in 11.7% of the samples in this report, 10.8% in the netherlands and 13.8% in england. the iiaa16g3r1 determined in 5.9% of our samples, was characterized in 4.65% in the netherlands and 5.8% in england. it is remarkable that subtypes, iiaa16g2r1, iiaa16g1r1 and iiaa13g2r1 were equivalently underrepresented in these three countries. this observation could suggest that the proportion of a gp60 subtype would not be randomly represented in a population. finally, the zoonotic transmission assessment of c. parvum in france would require a comparative investigation of variable genetic loci both in human and animal samples. this is the first report on the molecular identification of cryptosporidium species or genotypes in veal calves in france. according to data reported previously in many countries, a sequential distribution of species is observed in cattle according to age. c. parvum was mainly observed in the youngest calves, while c. ryanae and c. bovis became predominant in stool specimens collected in older animals. in some cases, several cryptosporidium species were successively detected in the same calf, suggesting that the immune defense against c. parvum is not efficient against c. ryanae or c. bovis. finally, the major identification of the iiaa15g2r1 subtype in france suggests that 5-week old calves could be a reservoir for zoonotic parasites transmissible to humans. fayer r: taxonomy and species delimitation in cryptosporidium a massive outbreak in milwaukee of cryptosporidium infection transmitted through the public water supply the cell biology of cryptosporidium infection advances in the epidemiology, diagnosis and treatment of cryptosporidiosis zoonotic cryptosporidiosis in the uk -challenges for control cryptosporidium and giardia-zoonoses: fact or fiction? involvement of insects in the dissemination of cryptosporidium in the environment cryptosporidial infection in a calf cryptosporidiosis in neonatal calves: 277 cases (1986-1987) a review of the importance of cryptosporidiosis in farm animals apicomplexa: cryptosporidiidae) in sheep (ovis aries) re-description of cryptosporidium cuniculus inman and takeuchi, 1979 (apicomplexa: cryptosporidiidae): morphology, biology and phylogeny cryptosporidium ubiquitum n. sp. in animals and humans epidemiology of anthroponotic and zoonotic human cryptosporidiosis in england and wales genetic analysis of cryptosporidium from 2414 humans with diarrhoea in england between 1985 and prevalence of species and genotypes of cryptosporidium found in 1-2-year-old dairy cattle in the eastern united states prevalence and agerelated variation of cryptosporidium species and genotypes in dairy calves a longitudinal study of cryptosporidiosis in dairy cattle from birth to 2 years of age wide geographic distribution of cryptosporidium bovis and the deer-like genotype in bovines prevalence of and management factors contributing to cryptosporidium sp. infection in pre-weaned and post-weaned calves in johor maddox-hyttel c: molecular and phylogenetic characterization of cryptosporidium and giardia from pigs and cattle in denmark prevalence of cryptosporidium infection in calves in france role of cryptosporidium parvum as a pathogen in neonatal diarrhoea complex in suckling and dairy calves in france prevalence of gastrointestinal disorders recorded at postmortem inspection in white veal calves and associated risk factors molecular characterization of cryptosporidium isolates obtained from humans in france identification of species and sources of cryptosporidium oocysts in storm waters with a smallsubunit rrna-based diagnostic and genotyping tool development of a multilocus sequence typing tool for cryptosporidium hominis basic local alignment search tool prevalence of cryptosporidium and giardia infections in cattle in aragon (northeastern spain) pattern of diarrhoea in newborn beef calves in south-west france genetic diversity of cryptosporidium spp. in cattle in michigan: implications for understanding the transmission dynamics risk factors associated with cryptosporidium parvum infection in cattle cryptosporidium and giardia in different age groups of danish cattle and pigs-occurrence and management associated risk factors epidemiology of cryptosporidium spp. and giardia duodenalis on a dairy farm molecular characterization and assessment of zoonotic transmission of cryptosporidium from dairy cattle in west bengal prevalence of cryptosporidium parvum infection in punjab (india) and its association with diarrhea in neonatal dairy calves high prevalence of cryptosporidium bovis and the deer-like genotype in calves compared to mature cows in beef cow-calf operations molecular characterization of cryptosporidium spp. in native breeds of cattle in kaduna state genotypes and subtypes of cryptosporidium spp. in neonatal calves in northern ireland molecular epidemiology of cryptosporidium subtypes in cattle in england cryptosporidium genotype and subtype distribution in raw wastewater in shanghai, china: evidence for possible unique cryptosporidium hominis transmission molecular epidemiology of cryptosporidiosis: an update cryptosporidium parvum mixed genotypes detected by pcr-restriction fragment length polymorphism analysis identification of genotypically mixed cryptosporidium parvum populations in humans and calves new view on the agespecificity of pig cryptosporidium by species-specific primers for distinguishing cryptosporidium suis and cryptosporidium pig genotype ii a multiplex polymerase chain reaction assay to simultaneously distinguish cryptosporidium species of veterinary and public health concern in cattle characteristics of cryptosporidium transmission in preweaned dairy cattle in henan cryptosporidium parvum infection in bovine neonates: dynamic clinical, parasitic and immunologic patterns resistance of calves to cryptosporidium parvum: effects of age and previous exposure cryptosporidium parvum in calves: kinetics and immunoblot analysis of specific serum and local antibody responses (immunoglobulin a [iga], igg, and igm) after natural and experimental infections apicomplexa: cryptosporidiidae) in cattle (bos taurus) genotype and subtype analyses of cryptosporidium isolates from dairy calves and humans in ontario age, geographic, and temporal distribution of fecal shedding of cryptosporidium parvum oocysts in cow-calf herds distribution of cryptosporidium subtypes in humans and domestic and wild ruminants in portugal genetic richness and diversity in cryptosporidium hominis and c. parvum reveals major knowledge gaps and a need for the application of "next generation" technologies-research review genetic classification of cryptosporidium isolates from humans and calves in slovenia molecular epidemiology of cryptosporidium in humans and cattle in the netherlands submit your next manuscript to biomed central and take full advantage of: • convenient online submission • thorough peer review • no space constraints or color figure charges • immediate publication on acceptance • inclusion in pubmed, cas, scopus and google scholar • research which is freely available for redistribution this study was supported by the catholic university of lille through the "projet grande campagne ensemble innovons" genotyping program. we would like to thank the veal unit managers who participated in this study. authors' contributions jf and kg participated in the conception and design of the study, carried out the experiments and drafted the manuscript. hl designed the sampling strategy and collected samples on farms. jf, kg and afd designed the protocol for molecular assay and participated in the analysis result. ohg carried out molecular assays. edc, gc and ph participated in the coordination of the study and helped draft the manuscript. all authors read and approved the final manuscript. the authors declare that they have no competing interests. key: cord-344984-mg779mix authors: bitar, dounia; tarantola, arnaud; capek, isabelle; barboza, philippe; che, didier title: risques d’importation des maladies infectieuses exotiques en france métropolitaine : détection, alerte et réponse date: 2009-11-30 journal: bulletin de l'académie nationale de médecine doi: 10.1016/s0001-4079(19)32418-5 sha: doc_id: 344984 cord_uid: mg779mix summary the french public health institute is responsible for promoting and coordinating threats the detection and assessment of health risks, and for suggesting possible responses. transmissible diseases affecting both human and animal health are the focus of surveillance networks. early detection of potential infectious threats is based on the screening of “ alert signals “ identified through routine surveillance networks and other systems. the quality and accuracy of these signals is first verified, before assessing, through a multidisciplinary approach, the risk of introduction and dissemination. this article examines specific cases illustrating the process of detection, risk analysis and response, with respect to infectious threats that are endemic in tropical regions and have the potential to be imported into metropolitan france. for both novel pathogens and exotic diseases — which, not being endemic in france, are less well known — the analysis and response process must regularly be adapted to the latest epidemiological, clinical and biological findings, taking interactions between the pathogen, host, and environment into consideration. the need to improve reaction times and risk assessment is also discussed. dans le cadre de ses missions de surveillance et d'alerte, l'une des priorités de l'institut de veille sanitaire (invs) est d'anticiper l'introduction et la diffusion d'une menace de santé publique sur notre territoire afin d'apporter les réponses les plus appropriées [1] . à cette fin, la veille prospective, la surveillance et l'expertise sont associées dans une démarche multidisciplinaire. l'organisation de la réponse est facilitée lorsqu'un phénomène infectieux est identifié au-delà de nos frontières et que sa vitesse de diffusion est relativement lente. ce fut le cas avec le sras en 2003 : bien qu'initialement peu précises, des informations cliniques et épidémiologiques étaient disponibles dans les pays du sud-est asiatique avant l'alerte internationale par l'organisation mondiale de la santé (oms). cette circulation des informations en amont d'une confirmation par une source officielle a permis aux pays européens d'anticiper l'introduction de cas importés sur leur territoire et d'organiser la mise en place des mesures de prévention et de contrôle, dans un délai de temps suffisant [2] . de même, la réactivité de la réponse est améliorée lorsque des plans de lutte sont déjà élaborés et mis à disposition des différents acteurs. ainsi, les plans contre la pandémie grippale et les nombreux exercices conduits depuis plusieurs années aux échelons local, national et international [3] ont permis aux pays européens de mettre en place des mesures de réduction de risque et de contrôle dans un délai très bref, suite à la diffusion rapide du nouveau variant de grippe a(h1n1) depuis le continent américain en avril 2009. dans ces deux exemples, la stratégie de l'autorité sanitaire ne vise pas nécessairement à « éviter à tout prix » l'introduction de cas individuels sur le territoire car la faisabilité et l'efficience de cette stratégie sont limitées. il s'agit surtout d'empêcher, de retarder, voire d'atténuer un cycle de transmission autochtone d'un pathogène par le diagnos-tic, l'isolement et la prise en charge précoces des patients et de leur entourage. un autre scenario pouvant être envisagé est celui d'un patient porteur de fièvre hémorragique virale arrivant en france métropolitaine, pour lequel les mesures d'isolement seraient retardées faute de diagnostic précoce. enfin, dans l'hypothèse de pathogènes non identifiés ou mal connus, il importe d'évaluer très rapidement le degré de menace pour la santé publique, malgré des informations cliniques, biologiques et épidémiologiques initialement peu précises. dans cet article nous nous attacherons à décrire quelques exemples de détection et d'analyse de risque concernant des phénomènes infectieux endémiques dans la zone intertropicale et potentiellement importables en france. il faut toutefois préciser que ces exemples n'illustrent qu'une partie du champ de la surveillance des maladies infectieuses telle qu'elle est organisée en france. par exemple la surveillance des infections nosocomiales ou les spécificités de la surveillance en fonction des zones géographiques ne sont pas abordées. le dispositif de veille et de surveillance des maladies infectieuses est basé sur le recueil d'informations visant à décrire un problème de santé, à détecter des épidémies ou des augmentations anormales de cas, à alerter l'autorité sanitaire, à aider à la gestion et enfin à évaluer l'efficacité des mesures (figure 1). dans une optique de prévention et de contrôle des maladies infectieuses -que ces dernières soient importées ou qu'elles surviennent sur le territoire métropolitain -cette surveillance ne peut être dissociée de celle concernant les phénomènes émergents. ces derniers sont en effet définis comme '' des phénomènes infectieux ou présumés comme tels, inattendus (en référence à leurs propriétés intrinsèques ou aux connaissances de leur biologie), touchant l'homme, l'animal ou les deux. il peut s'agir d'entités cliniques d'origine infectieuse nouvellement apparues ou identifiées, d'entités pathologiques infectieuses connues dont l'incidence augmente dans un espace ou dans un groupe de population donné ou d'une modification qualitative et/ou quantitative des caractéristiques de l'agent, de la maladie, de la population touchée et de son environnement. dans une optique d'anticipation, il peut s'agir d'une maladie identifiée dont les conditions d'expansion deviennent favorables. habituellement, une incertitude réelle ou perçue quant au potentiel évolutif, la maîtrise du phénomène et l'impact en santé publique humaine et/ou animale est présente '' [4] . les signaux d'alerte repérés par la veille sanitaire peuvent inclure des évènements non préalablement suivis ou la modification d'indicateurs provenant de systèmes de [5, 6] . l'un des objectifs des systèmes de surveillance organisés est de détecter des épidémies ou des phénomènes inhabituels. la surveillance des maladies à déclaration obligatoire et celle issue des centres nationaux de référence permettent le recueil de données cliniques, biologiques et épidémiologiques concernant des entités cliniques précises. le choléra ou les fièvres virales hémorragiques sont des exemples de pathologies non endémiques sur le territoire français métropolitain, soumises à la déclaration obligatoire, de pronostic potentiellement sévère, pour lesquelles des mesures rapidement mises en oeuvre peuvent prévenir la diffusion secondaire. la « surveillance syndromique » complète le dispositif en recueillant de manière systématique et continue des données d'activité médicale regroupées en syndromes : consultations aux urgences ou admissions hospitalières codées selon la classification internationale des maladies, pour lesquelles le tableau clinique détaillé et l'étiologie ne sont pas disponibles. une augmentation inhabituelle du nombre d'évènements par rapport à des valeurs de référence représente un signal qui doit être exploré. ces systèmes organisés ne permettent pas de détecter des phénomènes rares, peu connus ou insuffisamment caractérisés. pour ces derniers, il est demandé aux soignants, biologistes et acteurs de santé publique de '' signaler tout syndrome infectieux dont la fréquence et/ou les circonstances de survenue et/ou la présentation clinique et/ou la gravité sont jugées inhabituelles ''. le signalement peut être effectué par téléphone pour plus de réactivité comme illustré dans l'encadré avec l'exemple de la fièvre de la vallée du rift (fvr). un autre axe de la surveillance est celui de la veille bibliographique basée sur l'étude des publications scientifiques. dans une optique d'anticipation, cette veille scientifique permet par exemple d'analyser le risque potentiellement lié à des modifications de l'environnement comme le changement climatique, de s'informer sur les éventuelles adaptations des arthropodes vecteurs ou sur les mutations de certains agents pathogènes, de surveiller l'évolution et la rapidité de diffusion des résistances aux anti-infectieux, etc. l'anticipation s'appuie également sur une combinaison de réseaux d'information et d'alerte qui constituent la veille internationale et qui couvrent notamment l'importation de maladies infectieuses exotiques. les dispositifs les plus fréquemment utilisés incluent la veille issue des media, particulièrement structurée par le système canadien gphin d'accès payant [7] et le réseau promed, plus largement accessible [8] . des réseaux d'accès limité complètent ce dispositif, notamment le « global outbreak alert and response network (goarn) » de l'oms et le « early warning and response system (ewrs) » de la commission européenne, géré par le '' european centre for diseases control '' (ecdc) [9, 10] . ces réseaux sécurisés permettent aux institutions partenaires de s'informer mutuellement en temps réel et de lancer des alertes de portée régionale ou internationale. ils s'appuient sur le nouveau règlement sanitaire international de 2005 dont le but est « d'aider la communauté internationale à éviter les risques aigus pour la santé publique susceptibles de se propager au-delà des frontières et de constituer une menace dans le monde entier, en prenant les mesures qui s'imposent » [11] . en 2007, le centre national de référence des fièvres hémorragiques virales (institut pasteur) signalait par téléphone à l'invs un diagnostic de fvr chez un enfant habitant aux comores et transféré pour une hospitalisation à mayotte. une enquête téléphonique immédiate auprès des cliniciens de mayotte permettait de retracer le parcours et les dates de séjour de l'enfant aux comores. au cours des 48 heures suivantes l'équipe de la veille internationale complétait l'investigation en interrogeant les partenaires sur la survenue d'épizooties, de vagues d'avortements chez le bétail, de cas suspects ou confirmés aux comores et à madagascar. cette enquête mettant à jour la circulation du virus dans l'archipel des comores, une alerte a été immédiatement émise. les partenaires au niveau national et local (invs, cire, ddass, cliniciens, biologistes, etc.) ont été mobilisés pour la détection rapide des cas et la mise en place de mesures de contrôle. un dispositif de détection des cas à mayotte a été mis en place, complété par une intensification de la surveillance syndromique. en parallèle une analyse rétrospective des prélèvements a été effectuée, à la recherche d'anticorps anti-fvr parmi des patients ayant présenté des symptômes similaires mais pour lesquels le résultat initial était négatif pour la dengue, le chikungunya, la leptospirose et le paludisme. la grande richesse et la diversité des systèmes de surveillance assure une complémentarité des informations. certains systèmes sont plus sensibles au risque de générer de nombreuses fausses alertes. d'autres sont plus spécifiques et basés sur la confirmation étiologique mais peuvent engendrer un retard avant l'alerte. il importe donc de vérifier de manière rapide, rigoureuse et systématique les signaux réceptionnés à l'invs. cette étape est d'autant plus importante que la source d'information est peu spécifique, comme la veille issue des media. les éléments de validation portent sur plusieurs critères : la source de données et leur degré de fiabilité, les informations initiales sur la description clinique des cas (sévérité, tableau clinique compatible avec le diagnostic évoqué, éléments de confirmation biologique ou d'orientation étiologique, etc.), les données épidémiologiques (modalités de diffusion du germe), l'existence éventuelle de phénomènes similaires rapportés dans d'autres zones géographiques ou dans un passé proche, etc. la possibilité d'une origine bioterroriste doit être évoquée à ce stade. un nouvel évènement infectieux, une nouvelle souche bactérienne identifiée ne signifient pas nécessairement un risque pour la santé publique. ce dernier est évalué en fonction de nombreux critères : données cliniques et épidémiologiques (taux d'attaque, incidence, gravité, létalité, population touchée, classes d'âge touchées) ; nature de l'agent et modes de transmission supposés ou avérés (une transmission interhumaine par voie respiratoire faisant suspecter un phénomène hautement contagieux) ; interactions animal-homme (maladies vectorielles, zoonoses) ; facteurs environnementaux (changement climatique, pullulation vectorielle) ; modifications des comportements humains (flux migratoires, échanges commerciaux licite, ou non, entre pays) ; capacité épidémique et risque d'émergence ou d'extension (notamment en fonction de la sévérité du tableau, des difficultés de diagnostic biologique et de traitement spécifique) ; perception sociale et politique du risque ; enjeux économiques ; risque de diffusion internationale. les disciplines impliquées dans l'analyse reflètent la diversité de ces critères d'analyse. elles associent cliniciens, biologistes, épidémiologistes, vétérinaires, entomologistes, sociologues, etc. ainsi que les gestionnaires, afin de s'assurer de la faisabilité et de la cohérence de la réponse. l'expertise intègre de plus en plus souvent les modèles mathématiques qui permettent par exemple d'estimer le risque d'importation du sras ou de la dengue en europe [12, 13] ou d'évaluer l'impact des mesures de contrôle aux frontières pour la réduction de la diffusion d'une pandémie grippale [14] [15] [16] . l'analyse de risque peut être effectuée conjointement par plusieurs pays frontaliers ou sous la coordination et l'impulsion des institutions internationales : le risque d'implantation du chikungunya en france métropolitaine et sur le pourtour méditerranéen a fait l'objet de diverses expertises nationales et européennes [17] [18] [19] [20] . nous proposons ci-dessous un exemple d'analyse multidisciplinaire du risque d'introduction et implantation de la fvr en france métropolitaine. -modalités de l'expertise l'invs a mis en place dès 2000 un groupe d'experts multidisciplinaires (cliniciens, épidémiologistes, biologistes, vétérinaires, responsables de santé publique, chercheurs, gestionnaires) chargé de définir les priorités de surveillance concernant les zoonoses non-alimentaires [22] . ce groupe définit les zoonoses à surveiller en priorité en fonction de l'importance de la maladie humaine (incidence ou prévalence, sévérité, mortalité, potentiel épidémique, modes de transmission, existence de mesures de prévention et de contrôle), de l'importance de la maladie ou du portage chez l'animal (mammifère ou insecte) et du contexte environnemental qui peut évoluer au cours du temps. entre 2006 et 2008, le groupe d'experts a notamment été interrogé sur le risque de survenue de fvr en métropole. -analyse de risque pour la france métropolitaine (2008) l'incidence de la maladie est nulle en 2008 mais le risque d'importation et d'implantation (i.e. d'émergence) existe en raison de l'épizootie dans des zones géographiques ayant des liens particuliers avec la france métropolitaine, de la présence de vecteurs compétents et du rôle potentiellement aggravant du réchauffement climatique. il n'y a pas, en théorie, d'importation de bétail en métropole à partir de ces zones. le risque serait essentiellement lié à une arrivée de voyageurs infectés et virémiques, pouvant être à l'origine d'une chaîne de transmission en cas de piqûre par l'un des vecteurs présents. sur la base de ces critères, les experts ont classé la fvr comme hautement prioritaire en matière de surveillance humaine (niveau 4 sur une échelle de 1 à 4). en parallèle il importe de renforcer et adapter la prévention et le contrôle (niveau 2) et de développer des recherches pour une meilleure connaissance clinique et épidémiologique (niveau 3). par ailleurs, la surveillance et le contrôle dans le milieu animal sont également jugés prioritaires car l'infection peut provoquer des avortements spontanés et la mort de jeunes animaux d'élevage avec un impact économique important. les connaissances concernant le risque de fvr doivent être approfondies avec les entomologistes, les vecteurs potentiels étant nombreux (rapport en cours). les résultats de l'expertise ont été soumis à l'autorité sanitaire (direction générale de la santé) et à l'agence française de sécurité sanitaire des aliments (afssa) (lien http://www.afssa.fr/cgi-bin/countdocs.cgi). la première étape de la réponse a consisté à intensifier la surveillance biologique et à renforcer les capacités de détection des cas dans l'archipel comorien, avec une mise à dispostion de fonds. concernant la métropole, l'afssa devrait évaluer le risque d'importation pour le volet animal et recommander des mesures pour éviter l'implantation. ce travail est en cours. une alerte est émise lorsqu'une menace pour la santé publique est retenue comme plausible ; cette alerte doit s'accompagner d'informations sur les mesures de réduction de risque ou de contrôle. un exemple désormais classique est celui du sras au début de l'alerte mondiale, lorsque le germe était encore inconnu [2] . en cas d'urgence, des réunions téléphoniques sont rapidement organisées avec les experts des différents secteurs concernés sous la coordination de la direction générale de la santé. des conduites à tenir peuvent être proposées. par exemple, un guide de gestion autour des cas importés de choléra a été élaboré entre institutions (lien 1 ). ce dans le cas de menaces avérées par leur sévérité ou leur diffusion, mais insuffisamment caractérisées, l'évaluation, l'alerte et la réponse doivent être conduites rapidement malgré les nombreuses incertitudes. ainsi, la mise en quarantaine de personnes asymptomatiques exposées à un cas de sras n'est plus considérée comme justifiée car le risque de diffusion à l'entourage est infime avant la survenue de symptômes [23] , mais cette notion n'était pas clairement établie dans les premiers temps, justifiant que des mesures drastiques aient été mises en place. le dispositif que nous avons décrit doit également fournir des informations permettant l'évaluation de l'impact et l'adaptation des mesures mises en oeuvre. la mise à jour régulière des plans et conduites à tenir est une preuve de ces ajustements effectués en fonction des nouvelles connaissances épidémiologiques, biologiques, cliniques dans les domaines humain et animal, et en particulier dans le domaine des vecteurs. la mise à jour des connaissances par la veille scientifique est complétée par des retours d'expérience organisés au niveau national ou international. toutefois, ces retours d'expériences sont rarement publiés [2, [24] [25] [26] ). enfin, l'ensemble du dispositif de surveillance concernant un pathogène spécifique peut être évalué [27] . du fait de la grande diversité des sources d'information, les épidémiologistes chargés de la veille en maladies infectieuses font face à un volume croissant de signaux d'alerte, issus d'outils et de réseaux de veille très diversifiés. il importe d'observer avec un regard critique ces outils visant à détecter et à retarder l'implantation des agents importés. selon une étude récente [28] , parmi les milliers de signaux émis par le réseau promed entre 2006 et 2007 sur une période d'un an, vingt-sept signaux avaient été sélectionnés par l'institut de santé publique néerlandais en raison d'un risque potentiel d'importation et de diffusion pour leur pays. deux signaux étaient exclus car non validés et une menace possible était retenue et analysée de manière approfondie pour cinq des vingt-cinq signaux retenus. toutefois aucune de ces cinq menaces possibles n'a finalement donné lieu à une alerte ou à la mise en place de mesures de contrôle immédiates. d'autres auteurs [29, 30] soulignent également les limites de ces outils de veille dont la valeur ajoutée serait marginale en termes d'alerte et de réponse pour les pays européens ; ils indiquent néanmoins que ces signaux restent utiles et nécessaires pour connaître le « bruit de fond ». la combinaison de différents systèmes de surveillance apportant une meilleure sensibilité, des projets de recherche opérationnelle visant à améliorer les modalités de détection et à définir des critères plus pertinents de sélection des signaux d'alerte pourraient éventuellement permettre d'améliorer la spécificité. en parallèle, le renforcement des capacités de diagnostic biologique et des capacités de réponse est essentiel. enfin, le renforcement de la surveillance doit impérativement englober les pays de la zone intertropicale : le risque relatif de survenue et de diffusion d'une émergence infectieuse est en effet particulièrement élevé dans ces pays alors que les efforts de veille et d'anticipation sont généralement développés dans les pays plus riches [31] . à daniel eilstein et christine saura (institut de veille sanitaire) pour leur relecture attentive du document surveillance des maladies infectieuses : principes et organisation en france en l'épidémie de sras en 2003 en france. rapport sur la gestion épidémiologique du sras par l'invs different approaches to gathering epidemic intelligence in europe emergence des maladies infectieuses animales et humaines. inra productions animales the role of evolution in the emergence of infectious diseases gowtage-sequeria s. -host range and emerging and reemerging pathogens the global public health intelligence network and early warning outbreak detection: a canadian contribution to global public health promed-mail: an early warning system for emerging diseases emergent pathogens, international surveillance and international health regulations the early warning and response system for communicable diseases in the eu: an overview from implementing the international health regulations (2005) in europe desenclos j.c. -an approach to estimate the number of sars cases imported by international air travel -assessing the risk of importing dengue and chikungunya viruses to the european union international travels and fever screening during epidemics: a literature review on the effectiveness and potential use of non-contact infrared thermometers delaying the international spread of pandemic influenza entry screening for severe acute respiratory syndrome (sars) or influenza: policy evaluation invs -cas importés de chikungunya et de dengue en france métropolitaine: bilan de la surveillance à partir des données de laboratoire imported cases of chikungunya in metropolitan france: update to chikungunya risk assessment for europe: recommendations for action définition de priorités et actions réalisées dans le domaine des zoonoses non alimentaires severe acute respiratory syndrome: an update rapport de la mission d'évaluation et d'expertise de la veille sanitaire en france fièvres virales hémorragiques. bulletin épidemiologique hebdomadaire bulletind'alerteetdesurveillanceantilles-guyane2008 -la surveillance du virus west nile en france -the value of promed-mail for the early warning committee in the netherlands: more specific approach recommended epidemic intelligence in the european union: strengthening the ties surveillance sans frontieres global trends in emerging infectious diseases quelle est la place des centres nationaux de référence (cnr) dans l'élaboration de l'alerte sanitaire lorsqu'une maladie est inconnue en france ? en effet, face à un agent inconnu, il est difficile pour les cliniciens et biologistes de pouvoir identifier d'emblée vers quel cnr s'adresser en première intention. le renvoi des prélèvements ou souches vers un autre cnr lorsqu'un diagnostic alternatif est évoqué peut provoquer différents dysfonctionnements : délais prolongés d'acheminement, risque d'égarement des prélèvements ou mauvais rendement de prélèvements répartis dans plusieurs laboratoires, etc. l'autorité sanitaire a mis en place un laboratoire de niveau p3 à même d'effectuer un premier triage '' à l'aveugle '' sur des prélèvements de nature inconnue : la cellule d'intervention biologique d'urgence (cibu) basée à l'institut pasteur. ce laboratoire a pour mission d'intervenir lorsque la classe d'agent biologique n'est pas encore identifiée. l'intervention de la cibu est déclenchée par l'autorité sanitaire après analyse du risque. dès que l'orientation le permet, les cnr correspondants sont sollicités les plans d'action devant une alerte sanitaire sont-ils harmonisés en europe ? les modalités et déclinaisons des interventions restent adaptées à chaque pays, selon la situation et l'organisation du système sanitaire des pays respectifs. les plans '' pandémie grippale '' des pays européens ont été inspirés d'une trameélaborée par l'oms et adaptée par les différents pays de la communauté. par la suite de nombreux exercices inter-pays ont été organisés sous la coordination de l'union européenne, comme le '' common ground exercise '' conduit en 2005 (suivi d'autres exercices). ces exemples de mise en commun ne concernent pas exclusivement la pandémie : les alertes bioterroristes ou celles liées à des agents inconnus font également l'objet de discussions la réponse face à une alerte est également coordonnéee ou concertée, autant que faire se peut : -pour une alerte prévisible (exemple : pandémie grippale) dès le début de l'alerte l'union européenne a organisé des réunions téléphoniques de coordination et concertation, accompagnées d'échanges sur le site internet de l'ewrs. ces échanges permettent tout d'abord une mise à jour des connaissances et un partage sur les arguments ayant conduit chaque pays à faire un choix stratégique particulier. par la suite une mise en commun des données est organisée. ainsi pour la grippe h1n1, l'invs fournit chaque semaine à l'ecdc un bilan des cas de grippe, selon un format standard. l'ecdc se charge ensuite du transfert de ces données agrégées à l'oms pour les bilans mondiaux key: cord-285187-1h5tjs0r authors: kuchenbuch, mathieu; d’onofrio, gianluca; wirrell, elaine; jiang, yuwu; dupont, sophie; grinspan, zachary m.; auvin, stephane; wilmshurst, jo m.; arzimanoglou, alexis; cross, j. helen; specchio, nicola; nabbout, rima title: an accelerated shift in the use of remote systems in epilepsy due to the covid-19 pandemic date: 2020-08-31 journal: epilepsy behav doi: 10.1016/j.yebeh.2020.107376 sha: doc_id: 285187 cord_uid: 1h5tjs0r purpose: the purpose of the study was to describe epileptologists' opinion on the increased use of remote systems implemented during the covid-19 pandemic across clinics, education, and scientific meetings activities. methods: between april and may 2020, we conducted a cross-sectional, electronic survey on remote systems use before and during the covid-19 pandemic through the european reference center for rare and complex epilepsies (epicare) network, the international and the french leagues against epilepsy, and the international and the french child neurology associations. after descriptive statistical analysis, we compared the results of france, china, and italy. results: one hundred and seventy-two respondents from 35 countries completed the survey. prior to the covid-19 pandemic, 63.4% had experienced remote systems for clinical care. during the pandemic, the use of remote clinics, either institutional or personal, significantly increased (p < 10(−4)). eighty-three percent used remote systems with video, either institutional (75%) or personal (25%). during the pandemic, 84.6% of respondents involved in academic activities transformed their courses to online teaching. from february to july 2020, few scientific meetings relevant to epileptologists and routinely attended was adapted to virtual meeting (median: 1 [25th–75th percentile: 0–2]). responders were quite satisfied with remote systems in all three activity domains. interestingly, before the covid-19 pandemic, remote systems were significantly more frequently used in china for clinical activity compared with france or italy. this difference became less marked during the pandemic. conclusion: the covid-19 pandemic has dramatically altered how academic epileptologists carry out their core missions of clinical care, medical education, and scientific discovery and dissemination. close attention to the impact of these changes is merited. pandemics can lead to government regulations that limit social contact, decreased access to healthcare resources, and increased anxiety and fearall can disrupt the care path of patients with chronic diseases and decrease of face-to-face visits. in 2003, a study on the outbreak of severe acute respiratory syndrome (sars) in china showed that the loss of contact with medical care providers led to an increase in the withdrawal of antiseizure medications resulting in an increase in seizure frequency [1] . the current covid-19 pandemic is an important challenge for the management of patients with epilepsy worldwide. remote patient management systems [2] (in use since the 1990s and now integral to several national digital health strategies [3] [4] [5] [6] ) are a valuable tool during a pandemic to continue medical follow-up. they include different types of communications such as phone calls, one-way video links, and on live interactive communication. in addition, the epilepsy medical community is involved in educational activity and promoting knowledge dissemination through courses and scientific congresses. these activities also rely on face-to-face interactions and are also likely affected by the covid-19 pandemic. the aim of this study was to assess the impact of the covid-19 pandemic on the acute use of remote systems in clinics, education, and scientific meetings in the field of epilepsy and to explore the users' satisfaction and the perspectives of future use. we conducted a cross-sectional, electronic survey of epileptologists, neurologists, and pediatric neurologists mainly involved in the epilepsy field to determine the use of remote work during the covid-19 pandemic (supplementary data). to reach a wider public, this survey was addressed to adult and child neurologists specialized in epilepsy care through the european reference network for rare and complex epilepsies (epicare), international league against epilepsy (ilae), international child neurology association (icna), the french league against epilepsy (lfce), and the french society of child neurology (sfnp). the survey was comprised of 60 questions divided into four sections: demographic and general information followed by remote work for clinical practice, education, and scientific meetings and symposia (for details, see supplementary data). items assessed practice before and during the covid-19 pandemic. the first two sections were mandatory (demographics and clinical practice). we used different types of questions: closed (n = 60), semiopen (n = 12) , and open (n = 12). some questions used semiquantitative scales such as the likert scale. descriptive statistics included mean ± standard deviation for normal data, and median [25th-75th percentile] for non-normal data. in the event of missing data, percentages were calculated per number of responses obtained, item by item. frequency of use of remote system was scored as follows: never = 0, used it once = 0.5, few = 1, monthly = 2, weekly = 3, and daily = 4. wilcoxon signed-rank test were used to compare the frequency of the institutional and personal remote system use before and during the covid-19 pandemic and the frequency of use of these two systems during the same period. open-ended questions on free text allowed us to obtain qualitative data to illustrate respondents' feelings about their satisfaction with remote systems. we constructed a coding frame to analyze free-text data about satisfaction of remote clinic, online teaching, and virtual meeting. we subdivided into level categories to evaluate positive and negative aspects with some subcategory: cost, time, interaction, and target public. two authors (mk and rn) discussed the coding and interpretation of results. finally, we compared findings among the three countries with the highest number of respondents (france, china, and italy). quantitative or semiquantitative data were compared using kruskal-wallis h test followed, in case of significance (p b 0.05), by a dwass-steel-critchlow-fligner procedure. for qualitative data, we used chi 2 tests. a p-value b0.05 was considered as statistically significant, and a p-value b0.1 as a tendency. the statistical analyses were performed using r software [7] . this study was approved by the ethics committee of our institution necker hospital, aphp. participants were entirely free to participate and their consent was implicit. between april 6 and may 13, 2020, 172 respondents in 35 countries from 5 continents completed the survey from all over the word (table 1 , fig. 1 ). responders were involved in caring for children with epilepsy (n = 111, 64.5%), adults (n = 48, 27.9%), or both (n = 13, 7.6%). one hundred and fifty (87.2%) worked in a public hospital. all had a clinical practice, 91 were involved in clinical research (52.9%), and 27 in basic research activities (15.7%). most of the participants were from europe (n = 121, 70.4%). a containment policy due to the covid-19 pandemic was decreed in the countries of 166 participants (96.5%). indeed, 129 participants (75%) belonged to the 15 most impacted countries of the world in this period [8] . the sections concerning remote work for education and scientific meetings were completed by 160 participants (93% of all respondents). the questionnaire completion rate was 97% (302/10,150). prior to the covid-19 pandemic, 109 responders (63.4%) had already experienced using a personal (84/109, 77.1%) or an institutional (89/109, 81.7%) remote system: 64/109 for patient direct care (58.7%), 43 for education of trainees (39.4%), 31 for clinical case discussions within other institutions (28.4%), 40 for research (36.7%), and 13 for clinical case discussions within their own institutions (11.9%). for 76 on the 172 responders (44.2%), this experience was at least monthly using institutional (n = 64/76, 84.2%) or personal (n = 51/76, 67.1%) remote systems ( fig. 2a) . there was no statistical difference between the frequency of use of institutional versus personal remote system (p = 0.2). the three main personal systems used were skype® (n = 49 of the 89 using personal remote system, 55.1%), zoom® (n = 43, 48.3%), and webex® (n = 15, 16.9%). the means frequently used to contact remote respondents in an emergency were telephone calls (n = 134, 78%) and e-mails from families (126, 73%). other means (letters from families (37, 21.5%) and letters (71, 4.3%), telephone calls (74, 43%), and e-mails (88, 51.2%) from the attending physician) were less frequently used. during the covid-19 pandemic, the use of remote systems increased, both institutional systems (89 to 139) and personal systems fig. 2b ). however, contrary to the pre-pandemic period, the use of institutional remote systems was significantly higher than that of personal systems (p = 0.001). only one respondent from china did not have to reschedule any face-to-face clinics compared with 138 who rescheduled most or all their clinics (80.3%). one hundred and sixty-two respondents (94.2%) replaced faceto-face visits by various ways of remote connections with the families or the patients. this involved all clinics for 32 (19.8%), most of them for 98 (60.5%) and only a few for 32/162 respondents (19.8%). sixty-eight (42%) used phone calls without any remote specific connection with or without video, and 135 a remote system with video connection (83.3%). this system was institutional for 101 (74.8%) either regularly available (50/101, 49.5%) or developed for this pandemic (51, 50.5%). a personal remote system was used by 34 (25.2%). the duration of remote clinics was considered as identical as face-to-face ones for 50 respondents (31.1%), shorter for 77 (47.8%, including much shorter for 13), and longer for 34 (21.1%, including much longer for 6). regarding antiseizure medication changes, 50.9% of respondents tended to make fewer amendments (n = 83), 45.4% same (n = 74), and 3.7% more (n = 6). electroencephalogram (eeg) were less frequently requested for 65.6% (n = 107), without changing the frequency requested for 30.1% (n = 49), and more frequently requested for 4.3% (n = 7). blood test were less frequently requested for 52.8% (n = 86), without changing the frequency requested for 42.3% (n = 69), and more frequently requested for 4.9% (n = 8). respondents reported an increase in email and phone contacts by patients and their families (for 116, 67.4% and 104, 60.5% of respondents, respectively) but also by primary care physicians (for 45, 26.2% and 64, 37.2%, respectively for email and phone). one hundred and thirty-four respondents (134/160, 83.8%) were involved in educational activities. before the covid-19 pandemic, 131 (97.8%) had face-to-face lectures or small group teaching courses, and 76 (56.7%) had been involved in online teaching. educational activities were impacted by the covid-19 pandemic for 117 respondents (87.3%). eighty-two percent had at least a part of their activities canceled (n = 96/117, 82%), postponed (n = 43, 36.8%), or transformed to online teaching (99, 84.6%). respondents' courses were either interactive (24/99, 24.2%), video recorded (17/99, 17.2%), or both (52/99, 52.6%). for 40/99 respondents, all courses were transformed to online teaching (40.4%). before the covid-19 pandemic, 62.9% (83/132) of the respondents had participated in remote scientific meetings, 57.6%, (72/125) in workshops, 48.2% (53/110) in clinical studies meetings, and 37.1% (39/105) in research symposia. few had such experience with national 9.8% (13/132) or international 9.1% (11/121) congresses. during the period from february to july 2020, responders had planned 4 [3] [4] [5] [6] [7] meetings. only a few were transformed to remote (median: 1[0-1.5]) giving the opportunity to eighty-nine responders (67.4%) who participated in at least one meeting transformed to remote. the rest of the meetings were canceled or postponed. sixty-one percent of respondents were satisfied by their remote clinics (99/162, including 17 very satisfied), 56.7% by their online teaching (55/99, including 8 very satisfied), and 45.2% by remote meetings (57/126) (fig. 3) . feelings regarding family and patient satisfaction with the remote clinic were positive for 72.2% of the respondents (118/162, of which 18 were very positive) and 51.5% regarding students and online teaching (51/99, of which 6 were very positive). almost onequarter of responders reported dissatisfaction with remote work, mostly for remote education (22.9%, n = 22/99), remote meetings (18%, 22/126), and remote clinics (23/162, 14.3%). respondents indicated they would likely continue greater use of remote work for remote clinics, education, and meetings after the covid-19 pandemic, in 81.2% (121/149), 62.9% (61/97), and 54.7% (87/159), respectively (fig. 4) . free text allowed us to have more qualitative data on the reasons to maintain remote working in their different activity axis. indeed, in their opinion, remote clinics had the advantage of decreasing time and cost for families and patients travel and consequently of work absenteeism. this was highlighted for follow-up visits but not for new patients having their first evaluation. for first visits, respondents declared a clear need for a face-to-face visit. saving time, adapting to the availability of students, and increasing the target audience due to the absence of the need to travel to attend the course seemed to be the positive factors identified by respondents regarding remote education. however, they identified several negative factors including a decrease in interactions, especially the immediate students' feedback reactions. workshops with a small number of participants was reported as particularly adapted to remote systems allowing a gain in term regarding travel, time, and cost. however, respondents agreed that national and international meetings are more adapted to in-person meetings as their major goal in addition to disseminate knowledge is to favor personal interactions and consolidate personal friendships and contacts to enhance collaboration and exchange of ideas. the pandemic began in december 2019 in china, late february 2020 in italy, and early march in france. these countries were all placed under quarantine (from 23 january to 8 april 2020 in china, from 9 march to 4 may in italy and from 16 march to 11 may in france). the peak of pandemic-related deaths occurred between february 14, 2020 [9] . comparison of data from france (n = 60), italy (n = 16), and china (n = 24) showed no significant differences in terms of age, gender, and practice (pediatric, adult, or both, public or private, epilepsy center or not). belonging to a healthcare network was statistically different between countries (p b 10 −4 ). indeed, only four chinese respondents (16.7%) belonged to a patient care network, whereas there were 50% in france (30 responders) and 87.5% in italy (14) . before the covid-19 pandemic, the rate of respondents who had experienced remote working systems was higher in china than in the two other countries (91.6% versus 61.7% for france and 25% for italy, p = 10 −4 ). in the same way, the number of respondents with an institutional remote work system was higher in china (65.5% versus 45% for france and 12.5% for italy, p b 10 −4 ). however, the rate of respondents who had a personal remote work system was not statistically different (china: 45.8%, france: 40% and italy: 18.7%, p = ns). the frequency (scored from 0: never to 4: daily) of use of institutional remote systems was significantly different between the three countries the proportion of respondents using official remote systems or phone calls without video for remote clinics was not statistically different between france, italy, and china (official remote system: china: 66.6%, france: 62.1%, italy: 66.6%, p = ns and phone call without video: china: 33.3%, france: 43%, and italy: 13%, p = ns). however, in china, remote personal systems were more often used to manage patients than in other countries (57.1%, italy: 20%, france: 5.2%, p b 10 −4 ). concerning educational activities, the proportion of respondents involved was not statistically different (china: 95.2%, france: 75.9%, and italy: 68.8%, p = ns). before the covid-19 pandemic, the proportion of chinese respondents who had already experience online teaching was significantly higher (82.6% versus 40.5% for france and 36.4% for italy, p = 0.003). the respondents who had their teaching activities impacted by covid-19 pandemic were 95% for china, 87.8% for france, and 72.8% for italy (p = ns). during the pandemic, all respondents in china replaced at least part of their course with online teaching (20/ 20) compared with 56% in france and 60% in italy (p = 0.002) in particular using interactive online teaching (china: 78.3%, france: 31%, and italy: 27%, p = 0.001). concerning remote meetings, a large majority of respondents had already used this system without any statistical difference between countries (china: 90%, france: 87.5%, and italy: 85.7%, p = ns). for satisfaction scores (from very unsatisfied: −2 to very satisfied: +2 with a neutral position: 0), only the impression on families' and patients' satisfaction for remote clinics had a tendency to be higher in france compared with china (1 for france and 1 [0-1] for china, p = 0.06, italy: 1 [0. ). all other satisfaction scores showed no significant difference. the covid-19 pandemic blockage has significantly strengthened the use of remote access technologies in medicine. our study showed that pandemic has increased the shift from classical to remote communication for epilepsy practitioners in all the fields of their activity, namely clinical activity, teaching, and scientific meetings. the satisfaction was acceptable, and almost all responders agreed on a possible future use of remote systems for some of the scientific and educational meetings or for occasional remote clinics excluding first visit. our study demonstrated that during the covid-19 pandemic, there has been a reduction of face-to-face visits with a replacement for most by remote clinics. in similar situations, remote systems had already been identified as a possible alternative to face-to-face visits, for example, during ebola or sars epidemics [10] . in the same way, our study showed an increase of remote clinic frequency use during the pandemic compared with the pre-pandemic period. prior work on remote systems in epilepsy has shown notable benefits. a pilot study in canada compared remote systems to face-to-face clinics showing a decrease of costs of 92.5% ($35.85 versus $466) with a satisfaction for patients of 90% and only 8% preferring a face-to-face next visit in both groups [11] . the main barriers to remote clinics are the need for clinical examination, technical support, and reimbursement [12] . in our survey, respondents identified the same advantages and barriers, the first visit being the most challenging. in another study comparing the impression of new patients on remote visits with face-to-face visits, patients' perceptions of the neurologist's understanding, their ability to say what they wanted, their confidence in the neurologist, and the usefulness of the visit were similar [13] . however, they stated more difficulties in describing their symptoms and concerns about confidentiality. in our study, 34 respondents used personal remote system for remote clinics. this raises concerns about privacy and protection. of note is that the explosion of remote working systems due to covid-19 attracted hackers [14] . one attack, called "zoom bombing", consists of an unwanted intrusion, causing disruption and possibly disclosure of medical confidentiality. in order to regulate the security of health data during remote clinics, countries have established strict rules such as the health insurance portability and accountability act (hipaa) in the usa [15] and general data protection regulation in eu [16] . most of the free-access personal remote systems in our study are not hipaa compliant. this point should be better addressed by health authorities in future development of remote clinics. until 2015, attendance in medical classes was correlated with passing the examination [17] [18] [19] . since 2015, however, some studies have found no clear correlation [20, 21] . for example, 4th year medical school students have more absences than 2nd year students due to conferences, meetings, and residency interviews, but unlike personal absences, this type of absence is not significantly associated with lower academic test scores [22] . this is likely due to the improvement of means of communication that have enabled the students to fill in the gaps. in a recent study using a combined approach between online teaching and face-to-face interactive medical course, online teaching attendance was higher than face-to-face, and the exam score was correlated to online teaching attendance. ninety-eight percent were satisfied with this teaching, and 93% wished to extend it to the entire second cycle [23] . this is a good illustration of the change of perspective that is taking place in undergraduate and postgraduate university education. factors associated with a good adherence to online teaching are mainly the quality of the technical system, support system, learner and instructor, and the perceived usefulness [24, 25] . the advantages and disadvantages identified by the providers in our study were in line with the literature, i.e., on the one hand, a greater flexibility, an increase of the dissemination of knowledge, a decrease of travel cost and time, and better accessibility, on the other hand, less peer-to-peer exchange and feedback difficulties, including nonverbal communication [26] . in the symposia and meetings, the same advantages and disadvantages as with teaching were identified, but the proportion of respondents who recommended this method for the future were lower than for clinics and teaching. the use of remote systems seemed to be more adapted for research networks and workshops than congresses. but during the covid-19 pandemic, the european academy of neurology replaced its congress by a virtual meeting free-of-charge. with more than 40,000 participants, they claimed this to be "the biggest neurology meeting ever" [27] . a virtual congress allows for lower prices, time savings, and a greater dissemination of knowledge both to and from all over the world. however, the respondents interviewed stressed the importance of face-to-face for the development of collaborative projects. our questionnaire highlighted, before the covid-19 epidemic, a stronger experience of remote systems in china compared with france and italy. this may be due to previous epidemic crisis in china, a larger geographic area of china compared with france and italy, and a lower density of neurologists and child neurologists (0.1 and 0.02 per 100,000 persons for neurologists and child neurologists, respectively in south-east asia region versus 6.6 and 0.8 per 100,000 persons in europe [28] ). indeed, prior to the covid-19 pandemic, some studies and reviews identified remote clinics in the field of epilepsy as an opportunity in rural regions and in resource-poor setting where the access to a specialist is an important barrier to epilepsy diagnosis and treatment [29] [30] [31] [32] [33] [34] [35] [36] . however, the covid-19 pandemic seemed to have accelerated the shift towards the implementation of remote clinics and had enabled france and italy to fill the gap with a strong development of remote patient management tools. the sample of this study was small, but respondents completed the survey just after the covid-19 in china and during the covid-19 pandemic and lock down in other countries, giving to this survey a value of "almost" real-time responses. responses were from many countries around the world thanks to the involvement of international societies. we cannot rule out the presence of a selection bias since this questionnaire was sent online. however, we believe that this study can present a picture about practitioners' opinion on remote work in epilepsy and help to develop future perspectives. in addition, a significant proportion of respondents in our sample focus on pediatric care. the use of remote clinics in this population is probably easier than in adults. indeed, parents may be able to successfully complete a visit on their child's behalf but adults with cognitive impairment or other limitations may have more difficulty negotiating the technical requirements of such a visit. finally, we did not request a detailed description of the applied methods of online teaching. the survey aimed to have answers on the three activity fields of the respondents without adding much details relatively long survey. the covid-19 pandemic has increased the shift from classical to remote communication for epilepsy practitioners in all the fields of their activity, namely clinical activity, teaching, and scientific meetings. the advances of these methods of communication have allowed a rapid adaptation to confinement policies using their flexibility and their accessibility. this allowed a maintained link between practitioners and patients, professors and students, and between groups and colleagues. the satisfaction was acceptable, and almost all responders agreed on a possible future for remote work, for some of the scientific and educational meetings or for occasional teleconsultations. in addition, the positive ecological impact of such approaches might be interesting in addition to the economic impact on health and academic costs. it is likely that in "the world after covid", the shift process to the implementation of these new modes of communication is moving forward, although the balance between face-to-face and remote work has yet to be determined in our different fields of activities, and the long-term benefit of such shift to virtual interaction should be evaluated. m. kuchenbuch, g d'onofrio, y. jiang, zm grinspan, j wilmshurst and r nabbout have any conflict of interest to disclose. s dupont has received honoria from eisai, ucb, gw, novartis, advicenne and shire. e wirrell has acted as an investigator for gw pharma and zogenix and has received consulting fees from biocodex and biomarin. s auvin has served as consultant or received honoraria for lectures from arvelle therapeutics, biocodex, eisai, gw pharma, novartis, nutricia, ucb pharma, zogenyx. he has been investigator for clinical trials for advicenne pharma, eisai, ucb pharma and zogenyx. a arzimanoglou has served as consultant, received honoraria for lectures from arvelle therapeutics, eisai, gw pharma, ucb pharma and zogenix. on behalf of his instiitution he has been investigator for clinical trials sponsored by eisai, gw, ucb pharma and zogenix. jh cross has acted as an investigator for studies with gw pharma, zogenix, vitaflo and marinius. she has been a speaker and on advisory boards for gw pharma, zogenix, and nutricia; all remuneration has been paid to her department. her research is supported by the national institute of health research (nihr) biomedical research centre at great ormond street hospital, nihr, epsrc, gosh charity, eruk, the waterloo foundation. n specchio has acted as an investigator for studies with zogenix, marinus, biomarin, and livanova, and has received consulting fees from zogenix, biomarin, arvelle, livanova. the impact of sars on epilepsy: the experience of drug withdrawal in epileptic patients advantages and limitations of teleneurology scotland ' s digital health & care strategy n.d danish ministry of health south african national department of health. national digital health strategy for south africa rdct. a language and environment for statistical computing. r found stat comput world health organization. coronavirus disease (covid-19) situation report-111 coronavirus covid-19 global cases by johns hopkins coronavirus resource center telemedicine: potential applications in epidemic situations feasibility of epilepsy follow-up care through telemedicine: a pilot study on the patient's perspective telemedicine and epilepsy care-a canada-wide survey randomised controlled trial of telemedicine for new neurological outpatient referrals hackers' new target during pandemic: video conference calls n health insurance portability and accountability act of 1996 regulation on the protection of natural persons with regard to the processing of personal data and on the free movement of such data, and repealing directive 95/46/ec (data protection directive) class attendance in college: a meta-analytic review of the relationship of class attendance with grades and student characteristics student attendance and academic performance in undergraduate obstetrics/gynecology clinical rotations absenteeism among medical and health science undergraduate students at hawassa university, ethiopia relationship between classroom attendance and examination performance in a second-year medical pathophysiology class does class attendance matter? results from a second-year medical school dermatology cohort study the impact of student absences on grade outcomes in a neurology clerkship setting assessment of a newly-implemented blended teaching of intensive care and emergency medicine at paris-diderot university evaluating e-learning systems success: an empirical study e-learning success determinants: brazilian empirical study emerging themes in e-learning: a review from the stakeholders' perspective ean congress highlights challenges -predict, prevent, repair atlas -country resources for neurological disorders tele-neurology in sub-saharan africa: a systematic review of the literature can technology help reduce risk of harm in patients with epilepsy? pediatric teleneurology: a model of epilepsy care for rural populations epilepsy field workers, a smartphone application and telephone telemedicine: safe and effective epilepsy care in rural nepal telemedicine for epilepsy: a useful contribution telemedicine for epilepsy support in resource-poor settings. front public heal managing epilepsy by telemedicine in resource-poor settings telemedicine in epilepsy: how can we improve care, teaching, and awareness? epilepsy behav the authors wish to thank the practitioners who participated in this study and the networks without whom this study could not have been carried out, namely the international league against epilepsy, the international child neurology association, the french league against epilepsy, the french society of pediatric neurology, and the ern epicare network. rima nabbout was supported by state funding from the agence nationale de la recherche under "investissements d'avenir" program (anr-10-iahu-01) and the "fondation bettencourt schueller". this research was supported by the agence nationale de la recherche under "investissements d'avenir" program (anr-10-iahu-01) and the "fondation bettencourt schueller", paris, france. supplementary data to this article can be found online at https://doi. org/10.1016/j.yebeh.2020.107376. key: cord-331509-p19dg1jw authors: bigault, lionel; brown, paul; bernard, cécilia; blanchard, yannick; grasland, béatrice title: porcine epidemic diarrhea virus: viral rna detection and quantification using a validated one-step real time rt-pcr date: 2020-05-31 journal: j virol methods doi: 10.1016/j.jviromet.2020.113906 sha: doc_id: 331509 cord_uid: p19dg1jw since 2014, porcine epidemic diarrhea virus (pedv) has reemerged in europe. rt-pcr methods have been described for the detection of pedv, but none have been validated according to a norm. in this study we described the development and validation of a sybr™ green one-step rt-qpcr according to the french norm nf u47-600, for the detection and quantification of pedv viral rna. the method was validated from sample preparation (feces or jejunum) through to nucleic acid extraction and rt-qpcr detection. specificity and sensitivity, limit of detection (lod), limit of quantification (lq), linearity, intra and inter assay variability were evaluated using transcribed rna and fecal and jejunum matrices spiked with virus. the analytical and diagnostic specificities and sensitivities of this rt-qpcr were 100% in this study. a lod of 50 genome copies/5 µl of extract from fecal matrices spiked with virus or rna transcript and 100 genome copies/5 µl of extract from jejunum matrices spiked with virus were obtained. the lower lq (llq) was 100 genome copies/5 µl and the upper lq (ulq) 10(8) copies/5 µl. this method is the first, validated according a norm for pedv and may serve as a global reference method to harmonize detection and quantification of pedv viral rna in both field and experimental settings. porcine epidemic diarrhea (ped) was first described in europe in 1971. it is characterized by watery diarrhea, vomiting, dehydration, and is most notable in young piglets. the etiologic agent, porcine epidemic diarrhea virus (pedv) which was first identified by electron microscopy (em) in 1977 (chasey and cartwright, 1978; debouck and pensaert, 1980) is now characterized as an enveloped virus with a single stranded positive sense rna genome, member of the order nidovirales, suborder cornidovirinae, family coronaviridae, subfamily orthocoronavirinae, genus alphacoronavirus, subgenus pedacovirus (walker et al., 2019) . in the 1980's, pedv was detected for the first time in asia whilst in europe it was endemic. during the 90's only few sporadic cases were reported in europe and most of these were reported in italy were it remains endemic (martelli et al., 2008) . during the last two decades new pedv strains have appeared in china and some of these strains have caused extremely severe outbreaks characterized by a morbidity of 100% and a mortality of 80-100% on suckling piglets (sun et al., 2012) . this has led to the naming of pedv as either s-non-indel or s-indel genotypes. in general the more virulent viruses belong to the s-non-indel group. in the last decade both s-non-indel and s-indel viruses have emerged in the usa with serious consequences for the industry. throughout europe, the predominant types are now closely related to the viruses circulating in asia and north and central america (boniotti et al., 2016) . furthermore, all viruses reported in europe since 2014 belong to the s-indel group (grasland et al., 2015; stadler et al., 2015; steinrigl et al., 2015; theuns et al., 2015) except for one in the ukraine (dastjerdi et al., 2015) . this data highlights the importance of pedv diversity across several continents. in france, since 2014, ped caused by s-non-indel is a notifiable disease. for territory monitoring purpose, all pedv suspicions have to be notified to french ministry of agriculture and the pedv genotype has to be confirmed by the national reference laboratory at the french agency for food, environmental and occupational health safety (anses). until today, no official method has been validated for the detection and quantification of the pedv viral rna. since the 2000s, real-time pcr emerged as a tool of choice for the detection and quantification of viral rna and has multiple benefits: i) these tests are highly specific ii) are easily standardized compared to "classical" virology procedures, iii) are much less time consuming, and iv) are highly reproducible. several rt-pcrs have been described for the detection of pedv rna (kim et al., 2007; miller et al., 2016) . for a rapid, accurate and reliable diagnosis of ped in the veterinary laboratory, a method for the detection of pedv viral rna has been developed and more importantly validated according to the "association francaise de normalisation" (afnor) french nf u47-600 norm entitled "requirement and recommendation for the implementation, development and validation of pcr in animal health" (afnor, 2015a; afnor, 2015b) . this validated sybr tm green one-step rt-qpcr was based on a previously published taqman® probe real time rt-qpcr (kim et al., 2007) and targeted the same zones of sequence in the conserved n open reading frame (orf) as this had previously allowed for broad range detection and the capability to differentiate between the closely related virus transmissible gastro-enteric virus (tgev). the method developed in the current study under nf u47-600, unlike other molecular tests developed for pedv, evaluates the whole process from sample preparation through to the detection and quantification by rt-qpcr. this method should help harmonize detection and quantification of viral rna from pedv belonging to both s-non-indel and s-indel strains in both field and experimental settings. all commercial methods were performed according to the manufacturers' recommendations unless otherwise stated. an alignment of 192 pedv n orf sequences that were available on the data base at the time of the study (2014) was made using mafft (katoh and standley, 2013) and the probabilities of the nucleotides at the priming zones defined by kim et al. (2007) (pednf : 5'-cgcaaagactgaacccactaattt-3', and pednr : 5'-ttgcctctgttgttactt-ggagat-3') were calculated using r (wagih, 2017) (fig. 1 ). based on these probabilities forward primer mpednf (5'-cgcaaagactgaacccactaa-3') and reverse primer pednr were chosen (fig. 1) . these primers were subsequently checked against n orfs of the s-indel and s-non-indel pedv strains circulating in europe (dastjerdi et al., 2015; grasland et al., 2015; hanke et al., 2017; martelli et al., 2008; stadler et al., 2015; steinrigl et al., 2015; theuns et al., 2015) . original cv777, the pedv reference strain isolated in 1977, was collected from perfused jejunum performed in 1981 and kept at -80°c. this stock was named wtcv777. wtcv777 was propagated in cell culture as previously described (hofmann and wyler, 1988) and was named cccv777. a stock of cccv777 was produced as follows: 20 x 175 mm 2 confluent monolayer of vero cells (atcc® ccl-81) were infected each with 500 µl of 6.8x10 4 tcid50 of cccv777 in infection media; emem (thermofisher scientific, france) supplemented with 0.3% tryptone phosphate broth, 0.02% yeast extract, 1% penicillin/streptomycin and 10µg/ml trypsin. after 24 hours of infection, cells were subjected to three freeze thaw cycles and the culture medium was clarified by centrifugation at 10000g for 10 minutes. a total volume of 1l of supernatant was then centrifuged for four hours at 20000g to pellet the virus. the pellet was then resuspended in 100ml of pbs. the infectious viral titer of cccv777 was determined by immuneperoxidase monolayer assay according kärber's method (kärber, 1931) . the virus stock solution was titrated by immuno-peroxidase monolayer assay to 1.2x10 7 tcid50/ml. four other pedv strains were used: three french field strains (pedv/fr/001/2014 genbank accession number (gb acc) kr011756, pedv/fr/001/2017 and pedv/fr/001/2019 gb acc mn056942), and one american strain (pedv/usa/2014/iowa gb acc mf373643, kindly provided by dr p.gauger from iowa state university). nine other 'non-pedv' rna viruses were also used: one pig alpha-coronavirus (porcine respiratory coronavirus, prcv), and two gamma-coronaviruses (infectious bronchitis virus (ibv) gb acc fj904713), turkey coronavirus (tcov) gb acc kr822424) as well as other pig viruses: a pig artevirus (porcine reproductive and respiratory syndrome virus (prrsv), gb acc ky366411), a pestivirus (classical swine fever virus (csfv)), three pig ortomyxoviruses (swine influenza viruses h1ni, h1n2, h3n2), and two swine dna virus, one circovirus (porcine circovirus type 2 (pcv2) gb acc af201311), and an asfavirus (african swine fever virus (asfv) bankit1774827 anses-mada68322). jejunum and fecal samples were collected from both specific pathogen free (spf) pigs confirmed negative for coronavirus rna by deep sequencing and from pedv infected pigs positive for pedv rna. the pedv positive samples had been collected during previous experimental studies (gallien et al., 2018a; gallien et al., 2018b; gallien et al., 2019) . spf samples were used as negative controls or were spiked with pedv produced in vitro as described in section 2.2. spiked spf samples were used for the validation of the method and are later referred to as 'infectious reference materials'. for each jejunum sample, 200mg were homogenized in 1ml of phosphate buffered saline (pbs) (merck, france) with 4mm stainless steel beads in a tissuelyserii (qiagen, france). samples were then clarified by centrifugation at 10000g for 10 minutes. for each fecal sample, 1ml was diluted in 9ml of pbs and vortexed for 5 minutes before clarification by centrifugation as describe above. to determine the limit of quantification (lq) of the pcr and produce standard for quantification, a rna transcript was produced by in vitro transcription of the pedv wtcv777 n orf sequence. wtcv777 rna was extracted using trizol (thermofisher scientific, france). viral rna extract was subjected to reverse transcription using hexanucleotide primers and superscript iii reverse transcriptase (thermofisher scientific, france). reverse transcription was performed at 55°c for 1 hour followed by enzyme inactivation at 70°c for 15 minutes. to amplify the n orf, 5µl of rt were subjected to pcr amplification in 50µl reaction containing 400nm of primers ogvb160-f (gtcggatccactttatggcttct) and ogvb160-r (gtcctcgagatt gtttaatttccterror! reference source not found.), 2.5 units of platinum taq hifi (invitrogen, france), 5µl of 10x high fidelity pcr buffer, and mgso4 at a final concentration of 2mm. the pcr was performed as follows: 95°c for 2 minutes for initial denaturation, 5 cycles of 95°c for 15 seconds, 30 seconds at 55°c decreasing by 2.5°c per cycle and then 68°c for 2 minutes, follow by 40 cycles of 95°c for 15 seconds, 60°c for 30 seconds and 68°c for 2 minutes. amplified pedv n cdna was separated on 2% agarose gel and extracted using montage gel extraction kit (millipore, france). 100ng of extracted product were cloned in pcr4-topo vector (invitrogen, france). plasmid dna was prepared using nucleospin® plasmid kit (macherey nagel, france). in vitro transcription was performed with maxiscript tm t7 transcription kit (thermofisher scientific, france) using 1µg of precipitated spei linearized n orfs plasmid. rna was purified with agencourt® rnaclean xp kit (beckmancoulter, france), and quantified using qubit® fluorometer (life technology, france, saint aubin). stock of in vitro transcribed rna was stored at -80°c. number of molecular copies was calculated according the following formula: ) × 6.023 × 10 23 rna transcript was diluted to 10 9 molecules/5µl, aliquoted in 100µl, supplemented with 20µl of rnastable® (m, france) and dried in speedvac® vacuum concentrator (thermoelectron, france). the standard transcript was resuspended in 1ml in deionized nuclease-free water and then log10 serially diluted from 10 8 to 10 2 copies/5µl and stored at -80°c. all rna extractions were performed using rneasy® mini kit (qiagen, france) with the following modifications. 120µl of sample mixture containing 100µl of sample, 10µl of an external exogenous control (eec) and 10µl of proteinase k were used as opposed to 100µl of sample alone as recommended by the kit. rna was eluted with 50µl of nuclease-free water and stored at -80°c until use. eec used in this study was viral rna genome (mengovirus). reactions were carried out in an applied biosystems 7500 real-time pcr system, with power sybr tm green rna-to-ct tm 1-step kit (applied biosystems, saint aubin, france). the final pcr mix volume was composed of 12.5µl of master mix (2x), 0.2µl of enzyme mix, 5µl of rna template, primers mpednf and pednr at 300nm or 600 nm, h2o to final volume of 25µl. rt-pcr cycles were as follows: reverse transcription at 48°c for 30 minutes, followed by 95°c for 10 minutes, then 40 cycles of 95°c for 15 seconds, 60°c for 1 minute, and a final melting curve analysis step as defined by the applied 7500 software v2.3. all sample amplifications with a melting temperature corresponding to the standard with a viral rna concentration equal to, or above to the limit of detection (lod) were considered positive. all of the following tests were performed using primers at 300nm. j o u r n a l p r e -p r o o f the analytical sensitivity and specificity were determined as described in the nf u47-600 norm. all nucleic acid extractions from viruses listed in 2.2 were tested. five strains of pedv were tested for inclusivity, and eleven other virus for exclusivity, among which, four coronaviruses, five other rna viruses, and two dna virus, all known as pathogens in pigs. the diagnostic sensitivity and specificity were determined as described in the nf u47-600 according to nf u47-600, lod is the last dilution of reference material that allows a detection of the target with a confidence level of 95%. n rna transcript dilutions were tested for the lod of the pcr. six points of a two-fold dilution series ranging from 400 to 12.5 genome copies/5µl were analyzed in eight replicates. three independent assays were performed for rna transcripts (lodpcr). to determine the lod of the method, spf jejunum and fecal samples spiked with cccv777 from 10 6 to 10 -2 tcid50/ml, were tested in two independent assays on a hundredfold serial dilution ranging from 10 8 to 10 2 and 50 n transcripts equivalent/5µl, as infectious reference materials (lodjejunum or lodfeces). lod's were determined by probit calculation (finney and stevens, 1948) . according to nf u47-600, lq is defined as the lowest (lower lq, llq) and highest level (upper lq, ulq) between which, for each dilution, the statistical bias is under or equal to 0.25log10. the bias is the difference between the measured value and the theoretical value calculated by linear regression on all dilutions. uncertainty is calculated as the variance of calculated point plus the medium bias value. the statistical bias is defined as the medium of uncertainty. for the lq, seven points of a ten-fold serial dilution of n rna transcript were tested (10 8 to10 2 ). ten independent assays were performed on four independent serial dilutions. the lq for organic matrices were calculated on results obtained for the lod assessment (hundredfold dilution from 10 8 to 10 2 ). pcr efficiency was evaluated by plotting the ct against an expected rna copy number in respect to the tcid50/ml (data not shown) for infectious reference material or by qubit j o u r n a l p r e -p r o o f quantification for rna transcript. in agreement with the nf u47-600 norm, an efficiency of 75 -125% was accepted. the forward primer of kim et al. (kim et al., 2007) (pednf) had perfect base pairing with 7 of the 192 (3.6%) n orfs sequences. the forward primer designed in the current study (mpednf) which did not contain the last three bases of kim et al. (2007) had perfect base pairing with 188 of 192 (97,9%) and of those that did not match at 100 % only one had a mismatch at the last 3' position ( fig. 1 a) . sequence of the reverse primer (pednr) had perfect base pairing with 123 of 192 sequence (64.1%) and those that did not match at 100 % did not have any mismatches in the last three nucleotides of the 3' end ( fig. 1 b) . concerning the alignments with the european strains available after may 2014, pednf had perfect base pairing with 7 of 56 n orfs sequences (fig. 1 c) . mpednf had perfect base pairing with 54 of 56 sequences, those sequences that did not match at 100 % only contained one mismatch and these were localized close to the 5' end ( fig. 1 c) . pednr had perfect base pairing with 55 of 56 sequences and only one single miss-match with the remaining sequence at the 5' end. amongst the different viruses strains listed in 2.2, only the pedv strains (cv777, american field strain, and three french field strains) were positive. wtcv777 (ct = 20 ), cccv777 (ct = 12), all with a tm of 79.5 ± 0.5°c which is the expected tm for the pedv sequence amplicon according to the in vitro transcription control. all the other viruses were negative. the analytical specificity and sensibility were both 100%. efficiency of the method, calculated by linear regression, was 91.04% ± 1.31(0.01) for rna transcripts, 93.51% ± 3.97(0.04) for spiked jejunum and 99.36% ± 5.12(0.05) for spiked feces. different concentrations of primers had no effect on the efficiency of the method (data not shown), however melting curve analysis showed the presence of primer dimers at 600nm and not at 300nm (figure 2 ). the lod was determined at 50 copies/5µl for the rna transcript, 50 copies/5µl (0.5x10 0.01 tcid50/ml for the spiked feces and 100 copies/5µl (10 0.01 tcid50/ml) for spiked jejunum (table 1) . for every selected rna dilution tested, from 10 8 to 10 2 copies/5µl, bias enlarged of uncertainty were included in the norm limits (-0.5 to 0.5) and statistical bias (mean of uncertainty) were < 0.25 log10 ( table 2 ). the ulqs and llq were 10 8 and 10 2 copies/5µl respectively for all matrices. calculations were done when a minimum of 23 out of 24 results were positive for the lod and for all replicates for lq. all coefficients of variation (cv) were below the 0.1 limit given by the norm nf u47-600 with 0.004 -0.032, 0.002 -0.035, 0.0004 -0.018, for rna transcript, jejunum and feces intra-assay cvs respectively and 0.022 -0.064, 0.008 -0.031, 0.007 -0.031 for rna transcript, jejunum and feces inter-assay cvs respectively (table 1) . the diagnostic sensitivity was 100% at two and fourteen dpi, pedv viral rna were detected in all true positive pigs. the diagnostic specificity was 100% as all non pedv infected pigs were found negative all along all experiments. pedv is of global importance to the pig industry with many different strains and genotypes existing in different continents. after 2013 and the introduction of both s-indel and s-non-indel strains to north america and the resulting huge economic losses, the french ministry for agriculture classified ped caused by the s-non-indel virulent strains as a notifiable disease. thus there was a need for a reliable method for rapid, accurate and specific detection and quantification of a broad range of pedv strains and one that was completely validated according to french norm nf u47-600. many methods have been developed and used for pedv detection and quantification as previously reviewed (diel et al., 2016) such as direct viral isolation, but it is laborious, time consuming, and requires a reliable model for all possible strains. furthermore, many pedv strains cannot be isolated in vitro. many immuno-assay tests have been developed to detect viral proteins (ifa, blotting, elisa) but all these methods are time consuming, have a low sensitivity and reaction, and are subject to cross reactivity decreasing the specificity. for these reasons the current study focused on developing and validating a specific and rapid diagnostic test for the detection of pedv viral rna. basing this test on a taqman® multiplex rt-qpcr, published by kim et al. (2007) , we developed and validated a sybr tm green one-step rt-qpcr method. the development and validation of the complete method, including the steps of sample preparation, rna extraction, and rt-qpcr, were done according to the french standard nf u47-600. this norm is an adaptation to the french context of the manual of diagnostic tests and vaccines for terrestrial animals (international office of epizootics, 2018) and respects the criteria stated by the world organization for animal health (oie). these standards describe the validation criteria for a pcr method in animal health and allows the characteristics not only of rt-qpcr to be determined, but also of the complete method, including sample preparation and extraction. for this, fecal and jejunum samples were used as this material has previously been described as the best matrices for detection of pedv rna in animals (gallien et al., 2018a) . validating the complete method in this way means that the method is applicable for both experimental and diagnostic purposes. in the current study the primers used by kim et al. in 2007 were refined by in silico analysis. n orf alignments of the priming site showed that the pednf forward primer of kim et al. (2007) had mismatches with several different pedv n orfs and that the last three nucleotides at the 3' end only matched with 3.6% of the sequences. removing these three nucleotides in primer mpednf allowed a 100% match with 97.9% of international sequences and with 96.4% of european strains. the method using the new coupled primers demonstrated sufficient sensitivity to detect all tested pedv strains (historical, s-indel and s-non-indel strains). although sybr tm green pcrs are characteristically less specific than probe based pcrs, the specificity of the method was 100% against all viral types tested. primer dimer formation, which are problematic for fluorescent dye based methods as they interfere dramatically with quantification, were eliminated by optimizing the primer concentration to 300nm. during validation, the sample preparation and rna extraction step were optimized by the addition of a proteinase k treatment step which allowed the statistical bias to be maintained in acceptable limits (<0.25log10). the statistical bias obtain with the proteinase k treatment confirms a correct reproducibility at all quantification points, and guarantees a near or equivalent lod (50 and 100 copies/5µl for feces and jejunum) for the different matrices than for the transcribed rna (50 copies/5µl). in addition, the detection limit determined in this study (10 0.01 tcid50/ml) is very similar to other rt-qpcrs (10 0.03 tcid50/ml) (miller et al., 2016) . in conclusion, many pcrs have been developed to detect and monitor the presence of pedv, but, as yet to the authors' knowledge none have been developed with a complete validation according to a norm such as the french nf u47-600. this fully validated method is the first of its kind for pedv and should help harmonize detection and quantification of pedv viral rna in both field and experimental settings. nucleotide probabilities at each position are shown as coloured text above the alignments. red text in the alignment sequences represent a mismatch. sequences of primers are shown above the alignment (pednf, mpednf or pednr). pednr is shown as reverse complement. each line represents a hybridization sequence, the number of strains presenting this sequence is indicated to the left of the sequence. j o u r n a l p r e -p r o o f afnor nf u47-600-1 méthodes d'analyse en santé animale -pcr (réaction de polymérisation en chaîne) -partie 1 : exigences et recommandations pour la mise en oeuvre de la pcr en santé animale afnor nf u47-600-2 méthodes d'analyse en santé animale -pcr (réaction de polymérisation en chaîne) -partie 2 : exigences et recommandations pour le développement et la validation de la pcr en santé animale porcine epidemic diarrhea virus and discovery of a recombinant swine enteric coronavirus virus-like particles associated with porcine epidemic diarrhoea porcine epidemic diarrhea virus among farmed pigs experimental infection of pigs with a new porcine enteric coronavirus, cv 777 porcine epidemic diarrhea virus: an overview of current virological and serological diagnostic methods a table for the calculation of working probits and weights in probit analysis better horizontal transmission of a us non-indel strain compared with a french indel strain of porcine epidemic diarrhoea virus evidence of porcine epidemic diarrhea virus (pedv) shedding in semen from infected specific pathogen-free boars limited shedding of an s-indel strain of porcine epidemic diarrhea virus (pedv) in semen and questions regarding the infectivity of the detected virus complete genome sequence of a porcine epidemic diarrhea s gene indel strain isolated in france porcine epidemic diarrhea in europe: in-detail analyses of disease dynamics and molecular epidemiology propagation of the virus of porcine epidemic diarrhea in cell culture manual of diagnostic tests and vaccines for terrestrial animals : (mammals, birds and bees) beitrag zur kollektiven behandlung pharmakologisher reihenversuche mafft multiple sequence alignment software version 7: improvements in performance and usability multiplex real-time rt-pcr for the simultaneous detection and quantification of transmissible gastroenteritis virus and porcine epidemic diarrhea virus epidemic of diarrhoea caused by porcine epidemic diarrhoea virus in italy evaluation of two real-time polymerase chain reaction assays for porcine epidemic diarrhea virus (pedv) to assess pedv transmission in growing pigs first detection, clinical presentation and phylogenetic characterization of porcine epidemic diarrhea virus in austria outbreak of porcine epidemic diarrhea in suckling piglets complete genome sequence of a porcine epidemic diarrhea virus from a novel outbreak in belgium ggseqlogo: a versatile r package for drawing sequence logos changes to virus taxonomy and the international code of virus classification and nomenclature ratified by the international committee on taxonomy of viruses the authors wish to thanks ms. cherbonnel-pansart for her help with afnor validation methodology, phd. le guyader for the furniture of the mengovirus and phd p.gauger for the s-indel strain furniture. this work was partially funded by "direction générale de l'alimentation" of the french ministry of agriculture (project n°2014-145). key: cord-320636-mvtux07x authors: pullano, g.; di domenico, l.; sabbatini, c. e.; valdano, e.; turbelin, c.; debin, m.; guerrisi, c.; kengne-kuetche, c.; souty, c.; hanslik, t.; blanchon, t.; boeì�lle, p.-y.; figoni, j.; vaux, s.; campese, c.; bernard-stoecklin, s.; colizza, v. title: underdetection of covid-19 cases in france in the exit phase following lockdown date: 2020-08-12 journal: nan doi: 10.1101/2020.08.10.20171744 sha: doc_id: 320636 cord_uid: mvtux07x a novel testing policy was implemented in may in france to systematically screen potential covid-19 infections and suppress local outbreaks while lifting lockdown restrictions. 20,736 virologically-confirmed cases were reported in mainland france from may 13, 2020 (week 20, end of lockdown) to june 28 (week 26). accounting for missing data and the delay from symptom onset to confirmation test, this corresponds to 7,258 [95% ci 7,160-7,336] cases with symptom onset during this period, a likely underestimation of the real number. using age-stratified transmission models parameterized to behavioral data and calibrated to regional hospital admissions, we estimated that 69,115 [58,072-77,449] covid-19 symptomatic cases occurred, suggesting that 9 out of 10 cases with symptoms were not ascertained. median detection rate increased from 7% [6-9]% to 31% [28-35]% over time, with regional estimates varying from 11% (grand est) to 78% (normandy) by the end of june. healthcare-seeking behavior in covid-19 suspect cases remained low (31%) throughout the period. model projections for the incidence of symptomatic cases (4.5 [3.9-5.0] per 100,000) were compatible with estimates integrating participatory and virological surveillance data, assuming all suspect cases consulted. encouraging healthcare-seeking behavior and awareness in suspect cases is critical to improve detection. substantially more aggressive and efficient testing with easier access is required to act as a pandemic-fighting tool. these elements should be considered in light of the currently observed resurgence of cases in france and other european countries. as countries in western europe gradually relaxed lockdown restrictions, robust surveillance and detection systems became critical to monitor the epidemic situation and maintain activity at low levels 1 . the need is to rapidly identify and isolate cases to prevent onward transmission in the community and avoid substantial resurgence of cases. in france, the surveillance strategy implemented by authorities to exit lockdown on may 11, 2020 was multifold 2,3 and based on an expanded case definition for covid-19 suspect cases to guide clinical diagnosis 4 ; recommendations to the general population to seek healthcare even in presence of mild symptoms; prescription of diagnostic tests to suspect cases by general practitioners for systematic and comprehensive testing; isolation of confirmed cases and tracing of their contacts. the specific characteristics of covid-19 epidemic, however, hinder the identification of cases 5 . large proportions of asymptomatic infectious individuals 6 , and presence of mild or paucisymptomatic infections that easily go unobserved 7, 8 present serious challenges to detection and control [8] [9] [10] . this may potentially result in substantial underestimates of the real number of covid-19 cases in the country. here we estimated the rate of detection of covid-19 symptomatic cases in france in may-june 2020 after lockdown, through the use of virological and participatory syndromic surveillance data coupled with mathematical transmission models calibrated to regional hospitalizations. the study focused on mainland france where the epidemic situation was comparable across regions, and excluded corsica reporting a very limited epidemic activity and overseas territories characterized by increasing transmission 11 . covid-19 epidemic management in france in the post-lockdown phase involved the creation of a centralized database collecting data on virological testing (si-dep, information system for testing) to provide indicators to monitor the epidemic over time 2, 12 . 20,736 virologically-confirmed cases were reported from may 13 (week 20) to june 28 (week 26) in mainland france. after imputation of missing data (see methods), an estimated 9,326 [95%ci 9,234-9,403] cases with symptoms resulted in the study period (figure 1) . the average delay from symptom onset to testing decreased from 20.7 days in week 20 (w20) to 7.1 days in w26. accounting for this delay (see methods), we estimated that 7,258 [95% ci 7,160-7,336] confirmed symptomatic cases had onset in the study period, showing a decreasing trend over time (1, 663 in w20, 892 in w26). the test positivity rate decreased in the first weeks and stabilized around 1.2% (average over w24-w26). a digital participatory system was additionally considered for covid-19 syndromic surveillance in the general population 11 , including those who do not consult a doctor. called covidnet.fr, it was adapted from the platform grippenet.fr (dedicated to influenza-like-illness surveillance since 2011 13, 14 ) to respond to the covid-19 health crisis in early 2020. it is based on a set of volunteers who weekly self-declare their symptoms, along with socio-demographic information. based on symptoms declared by approximately 7,500 participants each week, the estimated incidence of covid-19 suspect cases decreased from about 1% to 0.8% over time (figure 1) , according to the expanded suspect case definition recommended by the high council of public health for testing 4 (methods). 162 out of 524 suspect cases (31%) consulted a doctor in the study period. among them, 89 (55%) received a prescription for a test, resulting in screening for 50 individuals (56% of those given the prescription). week of symptom onset for symptomatic confirmed cases was estimated based on patients' declarations (see panel b) through a gamma distribution fitted to the data with a maximum likelihood approach. missing data about presence/absence of symptoms and declaration of onset were imputed by region and by week, by sampling from a multinomial distribution according to the observed breakdown among cases with complete information (see methods). test positivity rate was computed on cases with complete information. data for weeks 20-26 were consolidated in w27. (b) breakdown of virologically-confirmed symptomatic cases by week of testing according to declared onset of symptoms, along with estimated mean time from onset to testing. (c) incidence of covid-19 suspect cases (estimates by week and 3-week moving average (thick line)), along with percentage of those seeking healthcare, estimated from participatory surveillance system covidnet.fr. (d) number of covid-19 suspect cases of the participatory cohort seeking healthcare, and among them those receiving a prescription, and performing a virological test given the prescription. covidnet.fr estimates were adjusted on age and sex of participants. (e) estimated change in individuals attending their workplace locations over time and by region based on google location history data 15 we used stochastic discrete age-stratified epidemic models 17,18 based on demography, age profile 19 , and social contact data 20 of the 12 regions of mainland france, to account for age-specific contact activity and role in covid-19 transmission. disease progression is specific to covid-19 17, 18 and parameterized with current knowledge to include presymptomatic transmission 21 , asymptomatic 6 and symptomatic infections with different degrees of severity (paucisymptomatic, with mild symptoms, with severe symptoms requiring hospitalization) 8, 22 . the model was shown to capture the transmission dynamics of the epidemic in île-de-france and was used to assess the impact of lockdown and exit strategies 17, 18 . full details are reported in the methods section. intervention measures were modeled as modifications of the contact matrices, accounting for a reduction of the number of contacts engaged in specific settings, and were informed from empirical data. lockdown data came from refs. 17, 18 . the exit phase was modeled considering region-specific attendance at school based on . cc-by-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. (which was not certified by peer review) the copyright holder for this preprint this version posted august 12, 2020. . https://doi.org/10.1101/2020.08. 10.20171744 doi: medrxiv preprint ministry of education's data 23 , partial maintenance of telework according to estimated presence in workplaces from mobile phones location history data 15 (figure 1) , reduction in adoption of physical distancing over time based on survey data 16 (figure 1) , partial reopening of activities, senior protection 17 . a sensitivity analysis was performed on the reopening of activities and senior protection, as data were missing for an accurate parameterization. testing and isolation of detected cases were implemented by considering a 90% reduction of contacts for the number of virologically-confirmed covid-19 cases 17, 18 . region-specific models were calibrated to regional hospital admission data (figure 2 ) through a maximum likelihood approach in the phase before lockdown, during lockdown, and in the exit phase. further details are reported in the methods section. (a-c) hospital admissions over time, data (points) and simulations (median and 95% probability range), for île-de-france (a), pays de la loire (b), normandie (c). hospital admission data up to w27 (consolidated in w28) were used to calibrate the models. (d-f) projected number of new symptomatic cases over time (median and 95% probability range) and estimated number of virologically-confirmed symptomatic cases by week of onset (points), for the same regions above. the estimated detection probability of symptomatic cases (%) is also shown (red points, median and 95% probability ranges, right y axis). projected number of cases decreased over time in all regions, in agreement with the decreasing tendency reported in hospital admissions in the study period (figure 2) . overall, 69,115 [58,072-77,449, 95% probability range] new infections were predicted in mainland france in weeks 20-26 (from 22,882 [18, 221 ] in w20 to 2,922 [2,530-3,248] in w26). île-de-france was the region with the largest predicted number of cases (8,126 [4,848-10,305 ] to 944 [712-1,088] from w20 to w26), followed by grand est and hauts-de-france (table 1) . projections were substantially higher than virologically-confirmed cases (figures 2 and 3) . the estimated detection rate for symptomatic infections in mainland france in the period w20-w26 was 11% [9-13%], suggesting that 9 out of 10 new cases with symptoms were not identified by the surveillance system. estimated detection rate increased over time (7% [6] [7] [8] [9] % in w20, 31% [28] [29] [30] [31] [32] [33] [34] [35] % in w26). by the end of june, . cc-by-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted august 12, 2020. . https://doi.org/10.1101/2020.08.10.20171744 doi: medrxiv preprint 9 regions had a median detection above or equal 25% (figure 3) , and 2 regions detected a number of cases within the probability ranges of model projections ( table 1) . all regions except pays de la loire displayed increasing trends in the estimated detection rate. we compared the projected incidence of covid-19 symptomatic cases in w26 (4.5 [3.9-5.0] per 100,000) with the value obtained from confirmed cases (1.38 per 100,000) and two estimates based on covidnet.fr data (figure 3 ). the first estimate applies the measured test positivity rate to the number of self-reported covid-19 suspect cases (estimate #1, yielding 8.6 [6.2-11.5] per 100,000); the second additionally assumes that only 55% would be confirmed as suspect case by a physician and prescribed a test (according to covidnet.fr data on consulting participants, estimate #2, yielding 4.7 [3.4-6.3] per 100,000). our projections are in line with plausible estimates from covidnet.fr. sensitivity analysis showed that findings were robust against elements of the contact matrices that could not be informed by empirical data, and against current epidemiological uncertainties. including in the analysis also asymptomatic cases led to higher detection rates, 43% [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] % in w26 compared to 31% [28] [29] [30] [31] [32] [33] [34] [35] % for symptomatic cases only. this however assumes that asymptomatic cases were detected by the virological surveillance system in the week of infection, as no additional information was available to adjust for the possible delay. . cc-by-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted august 12, 2020. . https://doi.org/10.1101/2020.08.10.20171744 doi: medrxiv preprint table 1 . number of virologically-confirmed symptomatic cases, number of projected symptomatic cases, estimated detection rate, estimated trend in detection rate, population per region. regions are ranked by decreasing number of confirmed cases in w20. the trend is estimated comparing the average of the estimated detection rate in the weeks of june (w23-26) with the average in the weeks of may (w20-w22). despite a test positivity rate in mainland france well below who recommendations (5%) 24 , a substantial proportion of symptomatic cases (9 out of 10) remained undetected in the first 7 weeks following the end of lockdown. more than 60,000 symptomatic infections were not ascertained by the surveillance system from may 11 to june 28, 2020, according to our estimates. surveillance improved substantially over time. detection rate was estimated to be 7% [6] [7] [8] [9] % at the national level in mid-may, in line with estimates for the same period from a seroprevalence study in geneva, switzerland 25 . by the end of june, it increased to 31% [28] [29] [30] [31] [32] [33] [34] [35] %, leaving about 2/3 of cases with symptoms undetected. two regions (occitanie, normandy) reported cases compatible with model projections. these figures suggest that the new surveillance framework was progressively strengthened with increasing resources and capacity over time. detection became also faster, with a 66% reduction of the delay from symptom onset to testing. at the same time, increasing performance benefited from a concurrent decrease of the epidemic activity in all regions. despite this positive trend, our findings highlight a critical need for improvement. some regions remained with limited diagnostic exhaustiveness by the end of june. this is particularly concerning in those regions predicted to have a large number of weekly infections, such as île-de-france where approximately only 3 out of 10 cases with symptoms were detected by the end of june, and grand est (1 out of 10). novel recommendations since the end of lockdown require that all patients with symptoms suggestive of covid-19 (as well as contacts of a confirmed case) be screened for sars-cov-2 2 . almost all cases (92% since may 25) clinically diagnosed by sentinel general practitioners as possible covid-19 cases were prescribed a test 11 . however, only 31% of individuals with covid-19-like symptoms consulted a doctor in the study period according to participatory surveillance data. overall, these figures suggest that a large number of symptomatic covid-19 cases were not screened because they did not seek medical care despite recommendations. a similar evidence emerged from a large-scale serological study in spain where only . cc-by-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted august 12, 2020. . https://doi.org/10.1101/2020.08.10.20171744 doi: medrxiv preprint between 16% and 20% of symptomatic participants with antibodies against sars-cov-2 reported a previous virological screening 26 . by combining estimates from virological and participatory surveillance, together with measured rates for test recommendations by general practitioners (e.g., due to misclassification of selfreported symptoms), we extrapolated an incidence of symptomatic cases from crowdsourced data that is compatible with model projections. this finding further supports consultation for all covid-19 suspect cases. large-scale communication campaigns should reinforce recommendations to raise awareness in the population and strongly encourage healthcare-seeking behavior especially in patients with mild symptoms. at the same time, investigations to identify reasons for not consulting could be quickly performed through the participatory surveillance system. red tape might have contributed to low testing rates. prescription for a test was deemed compulsory in the new testing policy to prevent misuse of diagnostic resources 2 , however the path involving consultation, prescription, and lab appointment may have discouraged mildly affected individuals not requiring medical assistance. to facilitate access, some local initiatives emerged recently that increase the number of drivethrough testing facilities, mail test vouchers to promote massive screening in certain regions (e.g. in île-de-france 27 ), offer temporary mobile testing facilities (buses, pavilions) to increase proximity with the population 28 . these initiatives are particularly relevant for counteracting socio-economic inequalities in access to information and care in populations vulnerable to covid-19 29 and may be established in the longterm. given the non-uniform detection rate estimated within the country, learning from specific regional realities may aid to improve detection. the recent change in screening policy no longer requiring a prescription for testing 30 could further improve access. screening rates remained overall well below the objective fixed by authorities for the post-lockdown phase (average weekly number of tests in may-june was 250,000 vs. target of 700,000), and the delay from onset to screening was still very long (7 days) by the end of june, despite substantial reduction over time. the large demand for testing currently observed in certain regions, mainly as a result of imminent travels and protocols imposed by certain countries and air companies, is reportedly causing long waiting lists at overwhelmed testing sites 31 . given pre-symptomatic transmission, notification to contacts should be almost immediate to allow the effective interruption of transmission chains 21 . for testing to be an actionable tool for surveillance and, most importantly, for control of covid-19 transmission, screening rates should be radically increased and delays suppressed. the risk would otherwise be a rapid and uncontrolled resurgence of cases with potential transmission in the community 10 , as currently reported in some french areas (e.g. mayenne district in pays de la loire region) 32 and countries in europe 33 . such risk is expected to increase if the reported relaxation in preventive behaviors persists, due to adhesion fatigue 16 . aggressive and efficient testing will become increasingly more important in the fall months, as other respiratory viruses, such as influenza, rsv, rhinoviruses, will start to circulate and influenza-like-illness incidence levels will become comparable with those of covid-19. reviewing the testing strategy over summer, while at low covid-19 epidemic activity, is an important opportunity to strengthen french response system for next season. models were region-based and did not consider a possible coupling between regional epidemics caused by mobility. this choice was supported by stringent movement restrictions during lockdown 34 , and by the limited mobility increase in may-june 35 , before important inter-regional displacements took place at the start of summer holidays in july. foreign importations were neglected 9,36 as france reopened its borders with eu member states on june 15, and the schengen area remained closed till july. covidnet.fr cohort is not representative of the general population 14 , however the agreement found with sentinel incidence trends for influenza-like-illness suggests that these limitations have little effect once results are adjusted for lack of representativeness 13 . underdetection may also proceed from the imperfect characteristics of rt-pcr (reverse transcription-polymerase chain reaction) tests used to identify infected cases 37 . some cases tested for sars-. cc-by-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted august 12, 2020. . https://doi.org/10.1101/2020.08.10.20171744 doi: medrxiv preprint cov-2 could have been falsely negative. this would affect the analysis presented in the manuscript and would be in line with our conclusion that a large part of cases may have been undetected. asymptomatic infections were not considered in the main analysis, as we lacked information on the likely time of infection. the median duration from first to last positive nasopharyngeal swab was estimated to be 19 days in asymptomatic patients in china, with the longest duration at 45 days 38 . no such analysis has been performed in france yet. assuming that asymptomatic infections were rapidly identified through contact tracing yielded higher detection rates than estimated for symptomatic cases only. this is due to a proportionally higher fraction of asymptomatic cases among the confirmed ones. though limited by the underlying assumption, this result further strengthens the main conclusion that detection of symptomatic index cases is the key aspect that requires fundamental improvement. our findings identify critical needs of improvement to increase case ascertainment in france and the performance of the response system to monitor and control covid-19 epidemic. substantially more aggressive and efficient testing needs to be achieved to act as a pandemic-fighting tool. these elements should be considered in light of the resurgence of cases currently observed in some regions in france and in other countries with similar response systems. virological surveillance data and analysis. the centralized database si-dep for virological surveillance 12 collects detailed information on patients tested in france, including (i) date of test, (ii) result of test (positive or negative), (iii) location (region), (iv) absence or presence of symptoms, (v) self-declared delay between onset to test in presence of symptoms. the delay is provided with the following breakdown: onset date occurring 0-1 day before date of test, 2-4 days before, 5-7 days before, 8-15 days before, or >15 days before. the si-dep database provided complete information for 13,887 (62%) out of 23,053 laboratory-confirmed covid-19 cases tested between week 20 (may 11-may 17) and week 27 (june 29-july 5), with an increasing trend of complete information over time. the study referred to the period from w20 to w26; data of w27 were used to account for the delays. data were consolidated in w27. for cases with complete information, we estimated the number of symptomatic laboratory-confirmed covid-19 cases by date of onset, using the information on the date of test and the time-interval of onset-to-test delay declared by the patient. we fitted a gamma distribution to these data with a maximum likelihood approach, obtaining a shape parameter equal to 0.64, and expected value of delay equal to 8 days. given a symptomatic confirmed case tested on a specific date, we assigned the onset date by sampling the onset-to-testing delay from the fitted distribution, conditional to the fact that the delay lies in the corresponding time-interval declared by the patient. to account for the changes in the distribution of self-declared delays over time, we also fitted the distribution to three periods of time, obtaining no significant difference. cases with missing data were imputed by sampling from a multinomial distribution with probabilities equal to the rate of occurrence of the labels reported for cases with complete information. imputation was performed by region and by week. onset date was then estimated for the imputed symptomatic cases. participatory surveillance data and analysis. covidnet.fr is a participatory online system for the surveillance of covid-19, available at www.covidnet.fr. it was adapted from grippenet.fr to respond to the covid-19 health crisis in march 2020. grippenet.fr is a participatory system for the surveillance of influenzalike-illness available in france since 2011 through a collaboration between inserm, sorbonne universite and sante publique france 13, 14, 39 , supplementing sentinel surveillance. the system is based on a dedicated website to conduct syndromic surveillance through self-reported symptoms volunteered by participants resident in france. data are collected on a weekly basis; participants also provide detailed profile information at enrollment. in addition to tracking influenza-like-illness incidence 13, 39 , grippenet.fr was used to estimate . cc-by-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted august 12, 2020. . https://doi.org/10.1101/2020.08.10.20171744 doi: medrxiv preprint vaccine coverage in specific subgroups 40 individual perceptions toward vaccination 41 and health-seeking behavior 42 . it was also used to assess behaviors and perceptions related to other diseases beyond influenza 43 . participants are on average older and include a larger proportion of women compared to the general population 14, 44 . participating population is however representative in terms of health indicators such as diabetes and asthma conditions. despite these discrepancies, trends of estimated influenza-like-illness incidence from grippenet.fr reports compared well with those of the national sentinel system 13, 39 . all analyses were adjusted by age and sex of participants. to monitor covid-19 suspect cases in the general population, we used the expanded case definition recommended by the high council of public health for systematic testing and described in their 20 april 2020 notice 4 : • (sudden onset of symptoms or sudden onset of fever) and (fever or chills) and (cough or shortness of breath or (chest pain and age > 5 years old)) • or (sudden onset of symptoms or (sudden onset of fever and fever)) and o (age > 5 years old and (feeling tired or exhausted or muscle/joint pain or headache or (loss of smell without runny/blocked nose) or loss of taste) o or ((age ≥ 80 years old or age < 18 years old) and diarrhea) o or (age < 3 months old and (fever without other symptoms))). figure 3 reports two estimates obtained from covidnet.fr cohort data for the incidence of symptomatic cases in w26. they are computed as follows: • estimate #1 = (covidnet.fr estimated incidence in w26) * (test positivity rate from si-dep in w26) • estimate #2 = (covidnet.fr estimated incidence in w26) * (estimated proportion screened and confirmed as covid-19 suspect case by a physician, and prescribed a test; estimates from covidnet.fr) * (test positivity rate from si-dep in w26) ethics statement. grippenet.fr/covidnet.fr was reviewed and approved by the french advisory committee for research on information treatment in the health sector (i.e. cctirs, authorization 11.565), and by the french national commission on informatics and liberty (i.e. cnil, authorization dr-2012-024) -the authorities ruling on all matters related to ethics, data, and privacy in the country. transmission models summary. we used a stochastic discrete age-structured epidemic model for each region based on demographic, contact 20 , and age profile data of french regions 19 . four age classes were considered: [0-11), [11] [12] [13] [14] [15] [16] [17] [18] [19] , , and 65+ years old. transmission dynamics follows a compartmental scheme specific for covid-19, where individuals were divided into susceptible, exposed, infectious, and hospitalized. we did not consider further progression from hospitalization (e.g. admission to icu, recovery, death 17 ) as it was not needed for the objective of the study. the infectious phase is divided into two steps: a prodromic phase ( ) and a phase where individuals may remain either asymptomatic ( , with probability =40% 6 ) or develop symptoms. in the latter case, we distinguished between different degrees of severity of symptoms, ranging from paucisymptomatic ( ), to infectious individuals with mild ( ) or severe ( ) symptoms. prodromic, asymptomatic and paucisymptomatic individuals have a reduced transmissibility = 0.55, as estimated in ref. 7 . a reduced susceptibility was considered for younger children and adolescents, along with a reduced relative transmissibility of younger children, following available evidence from household studies, contact tracing investigations, and modeling works [45] [46] [47] [48] [49] [50] . a sensitivity analysis was performed on relative susceptibility and transmissibility of younger children, and on the proportion of asymptomatic infections. models calibration. models were calibrated regionally to daily hospital admission data through a maximum likelihood approach. the likelihood function is of the form . cc-by-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. (which was not certified by peer review) the copyright holder for this preprint this version posted august 12, 2020. is the time window considered for the fit. calibration involved three steps, each one corresponding to a different epidemic situation: pre-lockdown 17 , during lockdown 17 , post-lockdown. covid-19) in the eu/eea and the uk -tenth update prise en charge par les médicins de ville des patients atteints de covid-19 en phase de déconfinement testing for covid-19: a way to lift confinement restrictions signes cliniques d'orientation diagnostique du covid-19 factors that make an infectious disease outbreak controllable suppression of a sars-cov-2 outbreak in the italian municipality of vo' substantial undocumented infection facilitates the rapid dissemination of novel coronavirus (sars-cov2) estimates of the severity of coronavirus disease 2019: a model-based analysis tracing and analysis of 288 early sars-cov-2 infections outside china: a modeling study reconstruction of the full transmission dynamics of covid-19 in wuhan données relatives aux résultats des tests virologiques covid-19 the potential value of crowdsourced surveillance systems in supplementing sentinel influenza networks: the case of france evaluating the feasibility and participants' representativeness of an online nationwide surveillance system for influenza in france covid-19 community mobil covid-19 : une enquête pour suivre l'évolution des comportements et de la santé mentale pendant l'épidémie. /etudes-et-enquetes/covid-19-une-enquête-pour-suivre-l'évolution-des-comportements-etde-la-sante-mentale-pendant-l-epidemie impact of lockdown on covid-19 epidemic in île-de-france and possible exit strategies expected impact of reopening schools after lockdown on covid-19 epidemic in île-de-france pyramide des âges 2020 -france et france métropolitaine the french connection: the first large population-based contact survey in france relevant for the spread of infectious diseases quantifying sars-cov-2 transmission suggests epidemic control with digital contact tracing déconfinement phase 2 : point de situation au 28 mai. ministère educ. natl public health criteria to adjust public health and social measures in the context of covid-19 seroprevalence of anti-sars-cov-2 igg antibodies in geneva, switzerland (serocov-pop): a population-based study prevalence of sars-cov-2 in spain (ene-covid): a nationwide, population-based seroepidemiological study covid-19 : une campagne de tests à grande échelle débute dans 32 communes d'ile-de-france covid-19 : dépistage gratuit dans un labo itinérant coronavirus highlights inequality in france's poorest department covid-19 : les tests de dépistages sont possibles sans prescription médicale tests covid-19 : les délais d'attente s'allongent dans les labos surchargés. lci covid-19 country overviews-week 28 population mobility reductions during covid-19 epidemic in france under lockdown mobility patterns in paris before, during, and after lockdown assessing the impact of coordinated covid-19 exit strategies across europe pitfalls in sars-cov-2 pcr diagnostics clinical and immunological assessment of asymptomatic sars-cov-2 infections participatory syndromic surveillance of influenza in europe influenza during pregnancy: incidence, vaccination coverage and attitudes toward vaccination in the french web-based cohort g-grippenet opinion about seasonal influenza vaccination among the general population 3 years after the a(h1n1)pdm2009 influenza pandemic healthcare-seeking behaviour in case of influenza-like illness in the french general population and factors associated with a gp consultation: an observational prospective study population perception of mandatory childhood vaccination programme before its implementation the representativeness of a european multi-center network for influenza-like-illness participatory surveillance changes in contact patterns shape the dynamics of the covid-19 outbreak in china age-dependent effects in the transmission and control of covid-19 epidemics coronavirus infections in children including covid-19: an overview of the epidemiology, clinical features, diagnosis, treatment and prevention options in children a case series of children with 2019 novel coronavirus infection: clinical and epidemiological features sars-cov-2 infection in primary schools in northern france: a retrospective cohort study in an area of high transmission cluster of covid-19 in northern france: a retrospective closed cohort study. medrxiv this study was partially supported by anr project dataredux (anr-19-ce46-0008-03) and evalcovid-19 (anr-20-706 covi-0007); eu h2020 grants mood (h2020-874850) and recover (h2020-101003589); reacting covid-19 modeling and surveillance grants. we thank pascal crepey, camille pelat, edouard chatignoux, juliette paireau, daniel levy-bruhl for useful discussions. we also thank all participants of covidnet.fr system. key: cord-275827-r86ygqmy authors: lapeyre-mestre, maryse; boucher, alexandra; daveluy, amelie; gibaja, valerie; jouanjus, emilie; mallaret, michel; peyrière, helene; micallef, joëlle title: addictovigilance contribution during covid-19 epidemic and lockdown in france date: 2020-06-23 journal: therapie doi: 10.1016/j.therap.2020.06.006 sha: doc_id: 275827 cord_uid: r86ygqmy abstract addictovigilance is a safety monitoring targeted at substances with potential for abuse and dependence. this vigilance was involved during the period of covid-19 epidemic due to the significant changes in access to drugs and psychological disruption caused by the pandemic and lockdown. this article aims to present the different steps implemented by the french addictovigilance network in collaboration with the french health authorities from march to may 2020, including monitoring of potential harmful events, and scientific communication. the first events were identified through the continuity of the networking between the french addictovigilance centres and their partners: community pharmacies, general practitioners, specialized structures and emergency wards. as soon as the lockdown began, first cases of overdoses (lethal or not) were reported with opioids, mainly with methadone, and other opioids (heroin, oxycodone, tramadol or antitussive codeine). lockdown-related noteworthy events consisted in clinical cases or other relevant information for which lockdown clearly played an important role : among the many substances identified at least once, pregabalin, benzodiazepines, cannabis, cocaine and nitrous oxide were the most significant in terms of prevalence, seriousness or particularly specific to the lockdown context. despite significant decrease in the activity and travel limited to vital needs, community pharmacies continued to identify falsified prescriptions in this period, highlighting an increase in suspicious requests for pregabalin, codeine and tramadol. in parallel, the french addictovigilance network continued its communications efforts in the period, issuing a newsletter on tramadol, a press release on methadone and naloxone, and participating in the covid-19 frequently asked questions (faqs) of the french society of pharmacology and therapeutic website (https://sfpt-fr.org/covid19 ). covid-19 epidemic has been an important challenge for addictovigilance, and has proved that this monitoring is highly essential for alerting health professionals and health authorities to points of vigilance in the field of psychoactive substances. any safety monitoring system is part of a global approach aimed at identifying emergence or spread of a health risk. this health security approach involves the early detection of signals and their most rapid integration into an action system allowing an adapted, effective and early intervention to preserve the health of populations. in the context of pharmacovigilance and drug safety, new or unexpected adverse drug reactions should be detected as early as possible in order to further inform and secure the use of the drug, giving the general population and health professionals the opportunity of evidence-based information about these risks. in the context of covid-19 epidemic, the french regional pharmacovigilance centres network ensured this mission [1] , with an assessment maintained in its continuity, based on a pharmacological and medical characterization of cases, shared with a population-based approach integrating pharmacoepidemiological methods when possible, contributing to optimizing the level of evidence. sharing and collaboration, both within and beyond the french pharmacology and therapeutics scientific community, was integral within these special weeks and beyond (see frequently asked questions [faq] at https://sfpt-fr.org/covid19 ) [2] [3] [4] . in the addictovigilance context, the field is even wider and more heterogeneous [5] [6] [7] . in the first weeks of epidemic spread, most of the interrogations were related to the disease itself and to drugs with supposed antiviral properties or interactions with the immune system. concerns about substances of abuse appeared as soon as lockdown occurred in france on march 17, 2020 . this article the french addictovigilance network was set up in the 1990s, in order to benefit from a proactive vigilance system targeted at substances with potential for abuse and dependence (except tobacco and alcohol), and to participate in a proactive and coordinated manner in the activities of the world health organization (who) expert committee on drug dependence [8, 9] . this vigilance is based on spontaneous notification by healthcare professionals of any serious case of misuse, abuse and drug dependence involving psychoactive substances, regardless of their nature or status [5, 10] . in addition to this passive monitoring subject to under-reporting, other sources of information have been developed to improve vigilance: systematic data collection on falsified prescriptions from pharmacies ("ordonnances suspectes indicateur d'abus possible", osiap survey) [11, 12] and on secure prescription forms for narcotic drug prescriptions ( "antalgiques stupéfiants et ordonnances sécurisées", asos survey) [13] , systematic data collection from patients seen in addiction specialized structures ("observatoire des produits psychotropes illicites ou détournés de leur usage médicamenteux", oppidum survey) [14] , analysis of toxicological data on chemical submission [15] or on deaths in a medico-legal framework ("décès en relation avec l'abus de médicaments et de substances", drames survey) [16] . addictovigilance can broaden the assessment of the potential for abuse and dangerousness of substances by specific analyses on large databases from the national health data warehouse [17, 18] , or on ad hoc field studies [19] [20] [21] [22] [23] [24] . the identification of a potential signal from one or more of the sources described above makes it possible to anticipate an emerging problem and to assess its magnitude using a multi-source approach (fig. 1) [5, 10] . on march 16, in his first address on the extend of the epidemic in france, the french president announced, in a message broadcast to the nation, the implementation of travel restrictions, lockdown, and a state of emergency involving the redeployment of the entire healthcare sector to prioritize covid-19 care from the following day. among the different measures launched by the government, several ones were intended to ensure continuous access for care, while limiting outing to what was strictly necessary (urgent medical care). in these conditions, both public and private medical sectors (general and specialized practitioners, nurses, most of other health professionals) decreased their activities, together with addiction specialized structures, in order to insure social distancing and prophylactic barrier measures to reduce the risks of viral contamination. for example, in many areas in france, several first line harm reduction structures [25] and addiction specialized centres modified their way of functioning, with limited access hours, redeployment of nurses and doctors for covid care, remote consultations, etc., all these changes leading to a degraded operating mode. some other structures may have also closed their doors, in particular those offering conviviality space with coffee and food for homeless and vulnerable isolated people, because of the impossibility to ensure social distancing. the rules for renewing prescriptions have been modified by several decrees (the first being published in the official journal on march 14, 2020 [26] ), in order to prevent the health risks related to the abrupt interruption of chronic exposure to drugs, in a context of a reduced availability of prescribers during covid-19 epidemic. pharmacists were invited to issue even if the period of validity of a renewable prescription has expired, within the framework of the initially planned dosage, a number of boxes per prescription line guaranteeing the continuation of treatment, for a period not exceeding one month. these measures include specific provisions concerning medicinal products liable to be abused or misused, such as anxiolytic or hypnotic drugs, opioid maintenance drugs and other narcotic drugs or drugs falling under the regulations of narcotics. along the successive decisions of the president and government over time, these decrees were intended to be prolonged during the period of the national state of health emergency. table 1 summarizes the different situations concerning psychotropic and narcotic drugs (at the date of may 31, 2020). from march 17, 2020, some important problems rapidly appeared: because of the strict lockdown and repeated controls for any outing or trip, border shutdown for all extra european countries, but also with our immediate neighbours, drug trafficking has been drastically impacted, raising fears of an increase of episodes of withdrawal syndromes in the population of drug users. opioid maintenance treatment (omt) should be considered as an essential treatment during the covid-19 pandemic, as significant risks to the community exist with an interruption of the stable provision of opioid treatment. difficulties for omt drug provision have been expected with permanent changes of the prescribing and dispensing rules for narcotic drugs in the first days of lockdown, leading some patients to stock large amount of methadone at home. there may be an increased risk of opioid j o u r n a l p r e -p r o o f 6 overdose arising from i) erratic access to omt, ii) erratic access to illicit opioid supplies and iii) increased access to takeaway doses of methadone, which would have required the systematic prescription for take-home naloxone supplies [27] . there was also a growing concern about the risk of overdose with methadone (or of accidental exposure because of lockdown and provision of takeaway methadone at home), as methadone was already the first substance involved in drug abuse-related deaths before the disease outbreak, with an increasing trend in the last years [28] . unfortunately, despite drug approval for forms of naloxone directly available without medical prescription in 2017, the level of use of takeaway naloxone from specialized structures or community pharmacies remains very low [27] . psychological disturbances may occur due to the lockdown, with an increasing risk of misuse and abuse of psychoactive drugs in the population of drug users (including patients on omt particularly vulnerable to these disruptions), but also in the general population [29] . distress may result in some people increasing their substance use and subsequently require treatment (for example, alcohol use may increase). changes in illicit drug supply may occur due to a range of complex interacting factors, with an increased demand for services. alternately, some people who use drugs may be less likely to request services during the pandemic, with an escalation of substance use during a time of distress. some not evidence-based and potentially deleterious "guidelines" were launched in order to anticipate withdrawals, with several dangerous recipes for substitution or techniques to make provisions of narcotic drugs. such practices may bring new patterns of problematic use, including access to new psychoactive substances sold on the internet, with free home delivery services for using up stocks of illicit drugs. finally, in relation with the covid-19 itself, concerns arose about risk of drug-drug interactions and qt prolongation with methadone potentially combined with chloroquine and hydroxychloroquine or azithromycin. actually, when infected by sars-cov-2, older people, men and those with medical comorbidities (chronic pulmonary disease, cardiovascular disease, cerebrovascular disease, diabetes and a compromised immune system) present a much higher likelihood of acute respiratory distress, renal failure and death. due to the respiratory and pulmonary tropism of sars-cov-2, people who smoke or vape tobacco or cannabis products were expected to be more at risk of pulmonary complications. immune-suppressed people, for example, due to hiv infection or other chronic medication conditions, are also at increased risk for sars-cov-2 infection. consequently, drug users with these conditions may be a subgroup more at risk. we described the different events and facts collected and observed from mid-march to may 31, 2020. in the first days of lockdown, several concerns emerged in the field. the first events were identified through the continuity of the networking between the french addictovigilance centres and their partners (i.e. community pharmacies, general practitioners, specialized structures and emergency wards). by the second week of lockdown, several cases of methadone overdoses for people at home were reported, and falsified prescription forms to obtain hydroxychloroquine and azithromycin were also identified as osiap by different pharmacies on the french territory. these early signals have been transmitted to national health authorities, leading to the implementation of a weekly specific monitoring of noteworthy cases or events related to the covid-19, related to the lockdown, and of all falsified or abnormal prescription forms reported through the osiap survey during the period. this weekly monitoring was closely done between the french addictovigilance network and the ansm [30] . the lessons of this weekly monitoring by conference calls and shared minutes of the meeting are presented in the following paragraphs. the fig. 2 summarizes the highlights of this monitoring. detecting and identifying signals are a cornerstone for addictovigilance actors: they need to be able to label a piece of information received as a signal [5, 10, 20, 25, [31] [32] [33] . signals suggesting a public health risk are collected and analysed in continuous manner in a surveillance process implemented by watchdog or public health structures, in a perspective of alert, anticipation and early action. in this framework, a signal is defined as a piece of information concerning a health phenomenon or exposure to a risk or hazard, which requires investigation in order to validate it and decide whether or not it should be considered as an alert. the signals observed in addictovigilance may be related to human cases (unusual deaths, symptoms or syndromes grouped in clusters); to psychoactive substances or associations thereof likely to have serious health consequences (presence of adjuvants, degree of purity, novelty of the substance or its usage) and to new ways of administration or new settings of use. monitoring such noteworthy events is an important issue in addictovigilance. simad-covid was the national periodical assessment with the aim to proactively monitor and share occurrence of fatal and non-fatal overdoses due to opioid medications (methadone, opioid analgesics) or opioid substances (heroin) or other illicit drugs (cocaine). as soon as the lockdown began, first cases of overdoses were reported with opioids, mainly with methadone, and to a less extent, with heroine and other opioid analgesics (oxycodone, tramadol) or cough syrups containing codeine. until may 31 methadone was the most reported drug among overdoses. interestingly, several characteristics of methadone overdose have emerged: i) accessibility of methadone by storage from family/friends at home was often reported ii) occurrence of overdose among opioid naïve subjects (never previously exposed to opioids or return to use after cessation) iii) occurrence among vulnerable subjects (homelessness, migrants, patients with psychiatric comorbidities) iv) methadone used outside its labelling in france, for anxiolytic or analgesic purposes iv) take-home naloxone was exceptionally used in the period. it is important to note that during this period the price for street methadone remained relatively low, suggesting continued accessibility during the lockdown period compared to illicit drugs. heroin overdoses were also observed in several areas, often among previous heroin users (around 40-45 years old) leading to severe opioid toxidromes (acute renal failure, rhabdomyolysis, haemodialysis). the same trend was observed with cocaine leading to cardiogenic complications including a patient with covid-19 myocarditis. overdoses were reported among young adults after tramadol use alone or associated with other drugs (cannabis) or after concomitant codeine and promethazine use (purple drank). lockdown related noteworthy events "simad confinement"" consisted in clinical cases or other relevant information for which lockdown clearly played an important role, and concerned all other substances, whatever their nature (medications, illicit drugs, diverted drugs). during the lockdown period and until may 31, 2020, 231 reports were collected by the 13 french addictovigilance centres all over the country, including oversea territories. among the many substances identified at least once in these reports, pregabalin, benzodiazepines (including z drugs), cannabis, cocaine and nitrous oxide (n2o) were the most significant in terms of prevalence, seriousness or particularly specific to the lockdown context. -first signals of abuse of pregabalin (a gabapentinoid close to gabapentin, approved for the treatment of neuropathic pain, epilepsy and generalized anxiety disorder) were reported in france from 2011 with falsified prescriptions, medical nomadism and diverted use for psychoactive effect [34] [35] [36] . the french addictovigilance monitoring of pregabalin has shown, at the end of 2018, a dramatic increase in the number of cases of abuse, with the emergence of a population of young abusers. during the whole lockdown period and then afterwards, reports came from medical doctors who were urgently requested for prescription of pregabalin by young people, often minors, including migrants. this pregabalin addiction was not clearly identified before by these health professionals, since in the recent past reports came only from community pharmacists reporting abnormal prescription of lyrica ® . during the period, several cases of overdose were reported with pregabalin, including one requiring hospitalization with dyspnoea and hallucinations in a 17-year-old male. -benzodiazepines and z-drugs were expected to be highly consumed during the beginning of the lockdown in france, because of social isolation or psychological troubles due to the lockdown with the potential increase of marital conflicts and domestic violence. no withdrawal syndrome was reported (renewal of prescriptions was possible along the period), but abuse or misuse (with alcohol or other psychoactive substances) were reported. clonazepam alprazolam, oxazepam and zolpidem were the most frequently reported. -several reports concerned n2o indicating persistent diverted use during the lockdown due to i) a shortage of other substances in some areas and ii) a need to consume due to inactivity. on the other hand, difficulties to easily obtain large quantities of n2o cartridges led a 24-year-old male to abuse cocaine because of his craving. during the lockdown, it would appear that home deliveries have been made easier with internet orders. neurological complications with sensory-motor axonal polyneuropathies were also observed in the period, highlighting the spread of this new phenomenon of n2o addiction that has appeared in recent months [37] . -unexpectedly, reports concerning cocaine were numerous (more over than with heroine or cannabis), while supply constraints could be considered as the same as for other illicit substances. actually, this accessibility varied according to the regions, with cocaine easily available in some ones and with a wide disparity in cocaine concentration. the above described case of switching n2o to cocaine illustrates this greater availability of cocaine, with modified supply chains (home delivery instead of buying on the street from dealers). -cannabis supply was expected to be more difficult during lockdown. some patients reported withdrawal symptoms due to supply difficulties or an increase in prices, while others abused cannabis in a context of anxiety related to the outbreak. cases of accidental poisoning in children under 2 years of age who have accidentally ingested cannabis have been also reported. in addition to these most frequent substances, other reports confirm that after a short period of waiting, the drug trade has adapted to lockdown, and cases of abuse, misuse or deleterious consequences of use were reported with synthetic cathinone 3-mmc (n = 7), amphetamines (n = 8), lsd, ketamine and ghb (n = 1 each). finally, even if the number of reports seems quite low, it should be borne in mind that there is often a delay in reporting (cases that have occurred since lockdown break have not been reported by may 31, 2020) and that under-reporting in this area is very significant [38] . the two first reports collected through the osiap survey concerned out of date and falsified hydroxychloroquine prescription forms (presented during the first week of lockdown), in the context of media coverage about its hypothetic efficacy on sars-cov-2 [1, 2] . this first signal has been forwarded to the ansm at the end of march. from this date, all suspected falsified prescription forms identified by community pharmacies and reported to the 13 addictovigilance centres were centrally analysed weekly and compared to the information collected at the same period in 2019. as a reminder, osiap is one of the national program implemented by the french addictovigilance network in the 1990s to record all falsified prescriptions presented to a network of community pharmacies located all over the country [11, 12] . this monitoring program has been useful to identify addictovigilance signals or characterize the abuse potential of prescription drugs [32, [39] [40] [41] . usually, osiap are periodically collected each year (in may and november) on a voluntary basis by sentinel pharmacies [12] . outside these proactive collection periods, osiap are continuously reported by community pharmacies, regional health authorities or medical/pharmacy councils. the osiap intensive data collection planned for may 2020 was cancelled due to the lockdown. between march 16 and may 31, 2020, 178 falsified prescription forms were reported by community pharmacies to the french addictovigilance network, in a context of a significant decrease in the activity and travel limited to vital needs. this frequency must be considered with caution, as falsified prescriptions are often reported with a significant delay each year. in comparison, 634 prescription forms were collected in the same period in 2019, including the intensive data collection in may [12, 42] . fig. 3 presents the main frequently reported drugs during the covid-19 monitoring by weeks, compared to the same period in 2019 (estimated through the information available on may 31, 2020). during this period, the most frequently reported drugs were pregabalin, antitussive codeine syrup and analgesic codeine and tramadol. pregabalin and codeine syrups were mainly requested by a population of young males. this profile was similar to that observed in the covid and the lockdown noteworthy events, highlighting the emergence of a little-known population to health professionals [43] [44] [45] . the french addictovigilance network has published a national newsletter on addictovigilance news for several years ("bulletin d'addictovigilance"), which was issued four times in 2019 (january, april, september and october) and once in 2020 (january). table 2 summarizes the different topics discussed in these newsletters, which highlight the emergence or confirmation of addictovigilance signals in the recent months. in retrospect, the majority of bulletins have addressed substances that had been a problem during lockdown. throughout the lockdown and then, communication by the french addictovigilance network remained active with release of new national communications. the last issue of the national addictovigilance bulletin was entitled: "limitation of the prescription period of tramadol: how did we get there". this bulletin presented a summary of the data collected in france on tramadol between 2013 and 2018 and summarized the key elements which have led in particular to limit the duration of prescription of this drug. from april 15, 2020, the maximum prescription period for analgesics containing tramadol has been reduced from 12 to 3 months. continuation of treatment beyond 3 months will require a new prescription. following the results of the national addictovigilance monitoring of methadone, the french addictovigilance network has published a press release on the need to maintain access to methadone during the lockdown period, while ensuring the safety of its use. methadone is a mu opioid receptor agonist indicated for the substitution of opioid dependence. in france, for at least the past ten years, it has been the most frequently retrieved substance during the toxicological analyses of those involved in deaths linked to the excessive use of psychoactive substances (drames survey). the lockdown period may increase the risks linked to exposure to this drug in naïve-opioid subjects including children and those around them not treated with methadone. it should be remembered that the potentially lethal dose of methadone ingestion in a person who has never used opioids is estimated at 1 mg/kg body weight. the press release focused on the risk of overdose, due to the larger dispensed quantities, methadone "storage", consumption of larger quantities of methadone or other respiratory depressants (alcohol, benzodiazepines, other opioids, etc.), resort to illegal obtaining, risk of overdose in the event of resumption of methadone after a few days off, risk of serious poisoning in children or naïve subjects. the press release also highlighted the risk of qt prolongation increased because high doses of methadone itself and because of combination with drugs or substances which also modify qt: domperidone, macrolides (erythromycin, clarithromycin, etc.), antidepressants (citalopram, escitalopram), antihistamines (hydroxyzine), antipsychotics (haloperidol, quetiapine), as well as drugs currently tested against covid-19 in hospitals (hydroxychloroquine, azithromycin, lopinavir/ritonavir) or other psychoactive substances such as cocaine. in order to minimize these risks, the press release insisted on warning about purchase of these drugs outside the pharmaceutical circuit, and on the need to report treatment with methadone in case of hospitalization for sars-cov-2 suspicion. the press release also insisted on the urgent need to increase the distribution of naloxone to methadone consumers (see brochure about where and how find naloxone; fig. 4 ). on march 16, 2020, the french society of pharmacology and therapeutics has launched a national faqs website at https://sfpt-fr.org/covid19, focused on the proper use of drugs during the covid-19 pandemic [4] . the french addictovigilance network has joined the scientific council and has participated to document the responses to each question related to addictovigilance. one topic of the faqs was about opioid maintenance treatment, because drugs approved in this indication (methadone and buprenorphine) should be considered as essential medications during the covid-19 pandemic, and significant risks to the community exist with an interruption of the stable provision of opioid treatment. another topic was related to the accessibility of naloxone take home in france. another topic gave information on the risk to switch to other substances (cannabidiol or gabapentin) to manage cannabis withdrawal or to switch to opioid analgesics outside medical management for non-cancer pain [46] conclusion covid-19 epidemic has been an important challenge for addictovigilance. only part of the events that took place during this period have been reported to the french addictovigilance network, and it is likely that in the coming weeks or months the number of overdoses or deaths related to substance abuse will be higher than described in this article. this is of particular concern for methadone, heroin and pregabalin, but also for cocaine and nitrous oxide which seem to be more accessible than expected in this period. this addictovigilance monitoring has proved to be indispensable for warning health professionals at the local and regional level in order to limit the risk for users, and for alerting health authorities at the national level to points of vigilance in the field of psychoactive substances. adverse drug reactions of hydroxychloroquine: analysis of french prepandemic sars-cov2 pharmacovigilance data off-label" use of hydroxychloroquine, azithromycin, lopinavir-ritonavir and chloroquine in covid-19: a survey of cardiac adverse drug reactions by the french network of pharmacovigilance centers french society of pharmacology t. non-steroidal anti-inflammatory drugs, pharmacology, and covid-19 infection genesis of an emergency public drug information website by the french society of pharmacology and therapeutics during the covid-19 pandemic signal identification in addictovigilance: the functioning of the french system social media mining for toxicovigilance: automatic monitoring of prescription medication abuse from twitter comment on: an insight into z-drug abuse and dependence: an examination of reports to the european medicines agency database of suspected adverse drug reactions from psychoactive medicines to addictovigilance in french public health code the french system of evaluation of dependence: establishment in a legal system safety signal detection by the french addictovigilance network: innovative methods of investigation, examples and usefulness for public health medical prescriptions falsified by the patients: a 12-year national monitoring to assess prescription drug diversion network of centers for e, information p. survey of forged prescriptions to investigate risk of psychoactive medications abuse in france: results of osiap survey tamperresistant prescription forms for narcotics in france: should we generalize them? surveillance system on drug abuse: interest of the french national oppidum program of french addictovigilance network french network of centers for e, information on p. chemical submission: results of 4-year french inquiry décès directement liés aux drogues interest of large electronic health care databases in addictovigilance: lessons from 15 years of pharmacoepidemiological contribution ten-year trend of opioid and non-opioid analgesic use in the french adult population a capture-recapture method for estimating the incidence of off-label prescriptions: the example of baclofen for alcohol use disorder in france identification and tracking of addictovigilance signals in general practice: which interactions between the general practitioners and the french addictovigilance network? parachuting psychoactive substances: pharmacokinetic clues for harm reduction medical complications of psychoactive substances with abuse risks: detection and assessment by the network of french addictovigilance centres use of new psychoactive substances to mimic prescription drugs: the trend in france identifying life-threatening admissions for drug dependence or abuse (iliadda): derivation and validation of a model les caarud, lieux privilégiés d'émergence de signaux pour l'addictovigilance arrêté du 19 mars 2020 complétant l'arrêté du 14 mars 2020 portant diverses mesures relatives à la lutte contre la propagation du virus covid-19 intérêt de la mise à disposition de la naloxone auprès des usagers de drogues pour le traitement d'urgence de surdosage d'opioïdes améliorer la balance bénéfices/risques de la méthadone en respectant ses spécificités pharmacologiques psychopathological consequences of confinement pharmacovigilance et addictovigilance dans le contexte du covid-19 : une surveillance renforcée detection of signals of abuse and dependence applying disproportionality analysis early signal of diverted use of tropicamide eye drops in france pregabalin use disorder and secondary nicotine dependence in a woman with no substance abuse history patterns of gabapentin and pregabalin use and misuse: results of a population-based cohort study in france drug abuse monitoring: which pharmacoepidemiological resources at the european level? warning on increased serious health complications related to non-medical use of nitrous oxide use of multiple sources and capture-recapture method to estimate the frequency of hospitalizations related to drug abuse evidence of clonazepam abuse liability: results of the tools developed by the french centers for evaluation and information on pharmacodependence (ceip) network slow-release oral morphine sulfate abuse: results of the postmarketing surveillance systems for psychoactive prescription drug abuse in france example of an investigation of an "emergent" phenomenon in addiction vigilance: the case of methylphenidate medical prescriptions falsified by the patients: a 12-year national monitoring to assess prescription drug diversion pharmaciens d'officine, étudiants en pharmacie et demandes de médicaments à base de codéine : étude observationnelle disproportionality analysis for the assessment of abuse and dependence potential of pregabalin in the french pharmacovigilance database detecting the diverted use of psychoactive drugs by adolescents and young adults: a pilot study site de l'association française des centres d'addictovigilance the french addictovigilance network would like to acknowledge all persons in the 13 addictovigilance centres who participated in the active monitoring during this period (all health professionals who reported cases during the period, and persons in charge of psychoactive drugs at the ansm (aldine fabreguettes, emilie monzon, charlotte pion, nathalie richard). authors have no competing interest to declare key: cord-339820-x8r27w14 authors: guan, l.; prieur, c.; zhang, l.; georges, d.; bellemain, p. title: transport effect of covid-19 pandemic in france date: 2020-07-29 journal: nan doi: 10.1101/2020.07.27.20161430 sha: doc_id: 339820 cord_uid: x8r27w14 an extension of the classical pandemic sird model is considered for the regional spread of covid-19 in france under lockdown strategies. this compartment model divides the infected and the recovered individuals into undetected and detected compartments respectively. by fitting the extended model to the real detected data during the lockdown, an optimization algorithm is used to derive the optimal parameters, the initial condition and the epidemics start date of regions in france. considering all the age classes together, a network model of the pandemic transport between regions in france is presented on the basis of the regional extended model and is simulated to reveal the transport effect of covid-19 pandemic after lockdown. using the the measured values of displacement of people mobilizing between each city, the pandemic network of all cities in france is simulated by using the same model and method as the pandemic network of regions. finally, a discussion on an integro-differential equation is given and a new model for the network pandemic model of each age class is provided. up to now, covid-19 has widely spread over the world and is much more contagious than expected. the outbreak of covid-19 has resulted in a huge pressure of hospital capacity and a massive death of population in the world. quarantine and lockdown measures have been taken in many countries to con5 trol the spread of the infection, and has proved the amazingly effectiveness of these measures for the outbreak of covid-19, in particular in china (see [1] ). quarantine is a rather old technique to prevent the spread of diseases. it is used at the individual level to constrain the movement of all the population and encourage them stay at home. lockdown measures reduce the pandemic trans-10 mission by increasing social distance and limiting the contacts and mobility of people, e.g. with cancellation of public gatherings, the closure of public transportation, the closure of borders. covid-19 may yield a very large number of asymptomatic infected individuals, as mentioned in [2] and [3] . therefore, most countries have implemented indiscriminate lockdown. but the long time 15 of duration of lockdown can cause inestimable financial costs, many job losses, and particularly psychological panic of people and social instability of some countries. as declared by some governments (see [4] ), testing is crucial to exit lockdown, mitigate the health harm and decrease the economic expensation. in this paper, 20 we consider two classes of active detection. the first one is the short range test: molecular or polymerase chain reaction (pcr) test, that is used to detect whether one person has been infected in the past. the second test is the long range test: serology or immunity test, that allows to determine whether one person is immune to covid-19 now. this test is used to identify the individuals 25 that cannot be infected again. for our research on covid-19, we aim to evaluate the effect of lockdown 2 all rights reserved. no reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted july 29, 2020. . https://doi.org/10.1101/2020.07.27.20161430 doi: medrxiv preprint within a given geographical scale in france, such as the largest cities, or urban agglomerations, or french departments, or one of the 13 metropolitan regions (to go from the finest geographical scale to the largest one). the estimations 30 of effect are also considered on different age-classes, such as early childhood, scholar childhood, working class groups, or the elderly. besides, we propose to understand the effect of partial lockdown or other confinement strategies depending on some geographical perimeters or some age groups (as the one that lyon experienced very recently, see [5] ) 35 in the context of covid-19, there have been many papers that focus on estimating the effect of lockdown strategies on the spread of the pandemic (e.g. [6] and [7] ). in [8] , the lockdown effect is estimated using stochastic approximation, expectation maximization and an estimation of basic reproductive numbers. in this work, we aim at evaluating the dynamics of the pandemic after the lockdown 40 by looking on the transport effect. in this paper, one contribution is that an extension of the typical sird pandemic model is presented for characterizing the regional spread of covid-19 in france before and after the lockdown strategies. taking into account the detection ratios of infected and immune persons, this extended compartment 45 model integrates all the related features of the transmission of covid-19 in the regional level. in order to estimate the effect of lockdown strategies and understand the evolution of the undetected compartments for each region in france, an optimization algorithm is used to derive the optimal parameters for regions by fitting the extended model to real reported data during the lockdown. 50 based on regional model analysis before and after the lockdown, we present a network model to characterize the pandemic transmission between regions in france after lockdown and evaluate the transport effect of covid-19 pandemic, when considering all age classes together. the most interesting point is the chosen exponential transmission rate function β(t), in order to incorpo-55 rate the complex effect of lockdown and unlockdown strategies and the delay of incubation. this paper is organized as follows. in section 2, the extended model is de-pandemic network of all cities in france. in the 'discussion' section, considering the age classification, an integro-differential model is presented for the pandemic network transmission, at any geographical scale, and for any set of age classes. in this paper, the scenario we consider is a large safe population into which 70 a low level of infectious agent is introduced and a closed population with neither birth, nor natural death, nor migration. there is one basic model of modelling pandemic transmission which is well known as susceptible-infected-recovereddead (sird) model in [9] . this mathematical compartmental model is described where s(t) is the number of susceptible people at time t, i(t) is the number of infected people at time t, r(t) is the number of recovered people at time t, d(t) is the number of deaths due to pandemic until time t, with constant parameters: β is transmission rate per infected, δ is the removal or recovery rate, α is the disease mortality rate. the compartment variables s(t), i(t), r(t), d(t) satisfy 80 at any time instant t, here n is the total number of population of the considered area. from the differential equations (1)-(4), it is obvious that at any time instant t, the total rate βi(t) of transmission from entire susceptible compartment to infected compartment is proportional to the infected i; the infected individuals 85 recover at a constant rate δ; the infected go to death compartment at a constant rate α. in fact, with the exception of the detected well-known data, there are some undetected data that cannot be measured but are significantly important for the analysis of the evolution of covid-19 in france under lockdown policy. moreover they are useful to provide efficient social policies, such as optimal management of limited healthcare resources, the ideal decision of the duration and level of lockdown or re-lockdown, and so on. inspired by [10] , the basic sird model is extended to a more sophisticated has recovered from the pandemic and is immune. the flow diagram of this model is sketched out in figure 1 . 5 all rights reserved. no reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted july 29, 2020. . https://doi.org/10.1101/2020.07.27.20161430 doi: medrxiv preprint the evolution of each compartments is modelled by the following equations, di with 6 all rights reserved. no reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted july 29, 2020. . https://doi.org/10.1101/2020.07.27.20161430 doi: medrxiv preprint the other parameters in equation (6)-(13) are defined as follows: • γ ir is the daily individual transition rate from i to r, and γ ir = (1 − all rights reserved. no reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted july 29, 2020. . https://doi.org/10.1101/2020.07.27.20161430 doi: medrxiv preprint • γ ih is the daily individual transition rate from i to h, and γ ih = (1 − • γ iu is the daily individual transition rate from i to u , and γ iu = (1 − • γ hr is the daily individual transition rate from h to r, and γ hr = (which was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted july 29, 2020. the infection transmission rate β(t) is the rate of the pandemic transmission from an undetected infected person to susceptible individuals at time instant t. as in [12] , in order to combine the complex effects of lockdown strategy, a time-dependent exponentially decreasing function can be used to model the 170 transmission rate β(t), with constant parameters β 0 , µ and κ. note that β(t) is constant during the initial stage of implementing effective lockdown strategies such as social distance, quarantine, healthcare, and mask worn. the transmission rate exponentially decreases at rate µ after these lockdown strategies take effect. the transmission 175 rate β(t) can be illustrated in figure 2 . as one of the most critical epidemiological parameters, the basic reproductive ratio r 0 defines the average number of secondary cases an average primary case produces in a totally susceptible population (see [13] ). as for the model in [10] , for the considered model in this paper, only the i − individuals transmit 180 9 all rights reserved. no reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted july 29, 2020. . https://doi.org/10.1101/2020.07.27.20161430 doi: medrxiv preprint the disease to the susceptible individuals during the early phase of outbreak. , the initial number of susceptible individuals exceeds the critical threshold to allow the pandemic to spread. thus the initial basic reproductive rate is when the transmission rate β(t) and s(t) evolve as time goes by, one dynamic reproductive rate that depends on time is introduced and known as effective reproduction number r(t) in [14] . in this model, it is defined as, for similarly, when r(t) < 1, the number of secondary cases infected by a 190 primary undetected infected case on day t, dies out over time, leading to a delay in the number of infected individuals. but when r(t) > 1, the number of undetected infected individuals grows over time. therefore, by the control of the transmission rate β(t) that can constrain r(t) to be less than 1, the number of infected individuals grows slowly to ease the pressure on medical resources. when s(t) is bellow a threshold, the epidemic goes to extinction (see e.g., [15] ). the required level of vaccination to eradicate the infection is also attained from the effective reproduction number. the compartmental model introduced in figure 1 exhibits a large number of unknown parameters (20 if we consider λ 2 = 0). the uncertainty on these 200 parameters can not be neglected. as an example, let us propagate uncertainty at the scale of the region auvergne-rhône-alpes. the vector of unknown parameters is: we take into account the uncertainties on these parameters by considering that each parameter is uniformly distributed with bounds consistent with typical 205 10 all rights reserved. no reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted july 29, 2020. . reported values (see, e.g., [10] and references therein). lower and upper bounds for each parameter are reported in table 1 hereafter. all rights reserved. no reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted july 29, 2020. . https://doi.org/10.1101/2020.07.27.20161430 doi: medrxiv preprint figure 3 shows that the prior uncertainty is pretty high, since for example the difference between the 75 % and the 25 % quantiles for the number of people in hospital is more than 50000 at the end of the lockdown period. on figure 4 we propagate the parameter uncertainty on the maximum number of people in intensive care units, on the date at which this maximum value is attained and on the total number of reported cases. note that the total number of reported cases is obtained from the daily number of reported cases, dr, which is driven by the following equation: the we see fpr example on these boxplots that the median for the maximum number 225 of people in intensive care is more than 8000 with the iqr greater than 20000. in view of the importance of uncertainties propagated from the model parameters to the quantities of interest (e.g., number of infected people at hospitals), it appears necessary to calibrate the model. our calibration procedure is described in the next section. in this section, regional scales of france are considered and all age classes are summed to calibrate the parameters of the pandemic model (6)-(13) during 12 all rights reserved. no reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted july 29, 2020. . https://doi.org/10.1101/2020.07.27.20161430 doi: medrxiv preprint the following weighted least square cost function is minimized for parameters optimization: where p is a vector which consists of calibrated parameters; z meas (t i ) is the measured values of the corresponding observed state vector z(p, t i ) at time t i , i = 1, . . . , n, with n the number of days considered for calibration. this optimization problem is solved using levenberg-marquardt algorithm (see [17] ). since it is a local algorithm, we adopt, as in [11, chapter 6] , a multi-start ap(which was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted july 29, 2020. is calibrated on daily data for h, u , d and r + on the lockdown period 2020-250 03-18 to 2020-05-11 from two data sources: the first one is a public and governmental data source [19] and the second one is a dedicated national platform with a privileged access [20] . the time step is chosen as ten percentage of one day for the numerical discretization. a general solver for ordinary differential equations is used to in order to characterize the dynamics of the pandemic transmission processes 260 during the confinement, the epidemiological model (6)-(13) was described in the 14 all rights reserved. no reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted july 29, 2020. . https://doi.org/10.1101/2020.07.27.20161430 doi: medrxiv preprint 15 all rights reserved. no reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted july 29, 2020. . https://doi.org/10.1101/2020.07.27.20161430 doi: medrxiv preprint 16 all rights reserved. no reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted july 29, 2020. previous section. we now consider the government action of unlockdown after confinement, there is a pandemic transmission effect between each region in france. considering n a age groups, the following pandemic network model of 17 all rights reserved. no reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted july 29, 2020. 18 all rights reserved. no reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted july 29, 2020. where transmission rate β ijk (t) depends only on (i, j), and is piecewise continuous depending on the scenario: lockdown or no-lockdown, for all t; for age group j, l kij (t) is the proportion of individuals moving from region k to region i in the age class j; the other parameters depend on the location, and also on the age group j; σ(j, k, t) is periodic (space dependent period t j,k ), satisfies 270 t j,k 0 σ(j, k, t)dt = 0, and takes value in the interval [−1, 1]; c i is the set of all regions that have pandemic transmission with region i. as the fast periodic switching policy in [21] , we consider the inverse of the (same) exponential function of infection transmission rate β(t) in (18) to denote β ijk (t). even though the end of confinement, the social strategies still go on, so a 275 continuous function β(t) is used for the whole transmission process of covid-19 from the start date of infection, with the end time of lockdown t end . the transmission rate β(t) for the whole transmission process is illustrated in (which was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted july 29, 2020. (which was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted july 29, 2020. table 8 : third part of components l ki in the mobility matrix l. all rights reserved. no reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted july 29, 2020. . https://doi.org/10.1101/2020.07.27.20161430 doi: medrxiv preprint is 11th of may in france. in this section, we use the parameter identification method developed in where l ki (t) is the proportion of individuals moving from city k to city i, and is derived from the real data of insee, and c i is the set of all cities that have 22 all rights reserved. no reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted july 29, 2020. . https://doi.org/10.1101/2020.07.27.20161430 doi: medrxiv preprint 23 all rights reserved. no reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted july 29, 2020. . https://doi.org/10.1101/2020.07.27.20161430 doi: medrxiv preprint 24 all rights reserved. no reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted july 29, 2020. . https://doi.org/10.1101/2020.07.27.20161430 doi: medrxiv preprint pandemic transmission with city i. all the other parameters are chosen as the 305 ones of the region to which each city belongs. to simulate this system of 8 * 36.000 differential equations, we now specify initial conditions. to simplify, the epidemic start date of each city is taken as the same as the epidemic start date of the region to which it belongs, and the initial condition for the undetected infected individuals i − 0 for the capital of (which was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted july 29, 2020. . https://doi.org/10.1101/2020.07.27.20161430 doi: medrxiv preprint figure 11 : the maps of the transport effect between cities in france (undetected infected plus detected infected from 0% (blue) to 2% (magenta) of the population for each commune): the date for the map on the left is 2020-05-01 and the one for the map on the right is 2020-06-01. figure 12 : the maps of the transport effect between cities in france(undetected infected plus detected infected from 0% (blue) to 2% (magenta) of the population for each commune): the date for the map on the left is 2020-07-01 and the one for the map on the right is 2020-08-01. all rights reserved. no reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted july 29, 2020. . https://doi.org/10.1101/2020.07.27.20161430 doi: medrxiv preprint which could be included in the modelling of the transmission rate β(t). 6. discussion and a new integro-differential model in this section, the general form of an integro-differential model capable of integrating different age classes and areas is introduced to discuss the transport effect of covid-19 in france after lockdown. by "areas" we mean a given 325 geographical scale as the set of 13 metropolitan regions (as considered in section 4), or the set or all 101 french departments, or all cities (as considered in section 5), or other geographical areas. for each age class a ∈ ages in area x ∈ areas, we consider the following integro-differential equations, for any time t ≥ 0 after confinement, ∂ t x(a, x, t) = f a (x(., x, t)) + areas σ(a, x, y, t)(λ in (a, x, y, t) − λ out (a, x, y, t))x(a, y, t)dy • areas, the set of different areas of population under study, depending on the considered geographical scale. as an example, considering all 27 all rights reserved. no reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted july 29, 2020. • x(a, x, t) ∈ r 8 is the 8-vector consisting of compartments of the age class a, in the area x, at time t; • for all age class a, f a (x(., x, t)) is the pandemic transmission dynamics for age class a from all other age classes in the area x at time t. without considering the age effect, it is given by the right-hand side of systems (6)(13) . inspired by the contact matrix approach developed in e.g. [23, chapter 3, page 76], by considering multiple age classes, the transmission term is the following integral where β a,b,x (t) is the contact function between age classes a and b, in the 28 all rights reserved. no reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted july 29, 2020. . https://doi.org/10.1101/2020.07.27.20161430 doi: medrxiv preprint area x, and at time t. therefore the function f a is given by where all parameters depend on the age class a and the area x; • λ in (a, x, y, t) ∈ r is the density of people coming (in) area x from area y ∈ areas at time t, for age class a; • λ out (a, x, y, t) ∈ r is the density of people going to (out) area y ∈ areas from area x at time t, for age class a; • σ(a, x, y, t) is the lockdown function for the age class a, between the areas x and y at time t. as an example, before the 11th of may, it was forbidden to travel for more than 100km in france. such a policy could depend on the age classes and on the areas, e.g., to control so called "clusters" of covid-19; • areas σ(a, x, y, t)λ in (a, x, y, t)x(a, y, t)dy provides the total number of people coming into area x from all the other areas. 29 all rights reserved. no reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted july 29, 2020. . https://doi.org/10.1101/2020.07.27.20161430 doi: medrxiv preprint equation (40) describes the network dynamics of covid-19 pandemic after lockdown and the transport effect on different age class on the basis of the regional pandemic transmission dynamics during lockdown. the proposed 360 structure makes it easier to understand different forms of the kernel. the interest of this model is that it could be adapated to any geographical scales, and to all age classes. for a control point of view, the most important term is σ(a, x, y, t) which defines the lockdown policy that defines the mobility between areas x and y at time t for the age class a. many control problems could be 365 studied for this model, as optimal control to reduced the pandemic effect, or to minimize the mortality in particular. it is of great importance for the mobility dynamics of the pandemic. beyond that, inspired by advection-diffusion modelling of population dynamics (as considered in [24]), it is natural to model the displacement inside a 370 given area by a diffusion term (see [25] ). the corresponding model is formulated as follows: ∂ t x(a, x, t) = f a (x(., x, t)) + d(a, x, t)∂ xx x(a, x, t) + areas (λ in (a, x, y, t) − λ out (a, x, y, t))x(b, y, t)dy +f ext (a, x, t), where the diffusion coefficient d(a, x, t) is a function that depends on age class a, areas x and time t. this 2-order partial differential equation predicts that for age class a in the 375 area x, how diffusion causes the number of individuals in the different compartments, especially undetected infectives and deaths, to change with respect to time t after lockdown. as long as one susceptible person is infected after directly or indirectly contacting disease carriers in the area x, diffusion takes place. when the number of infectious individuals in a local area is low compared 380 to the surrounding areas, the pandemic will diffuse in from the surroundings, so the number of infectives in this area will increase. conversely, the pandemic will diffuse out and the number of infectives will increase in the surrounding areas. the process of diffusion is influenced by distance, nearby individuals or 30 all rights reserved. no reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted july 29, 2020. . https://doi.org/10.1101/2020.07.27.20161430 doi: medrxiv preprint areas have higher probability of contact than remote individuals or areas. 385 finally, gender differentiation or other properties may be taken into account to characterize types of populations and to study the optimal lockdown control of pandemic dynamics based on our previous work. it is worth stressing that, in the long run, optimal lockdown strategies should consider the balance between the lower number of deaths and minimum healthcare and social costs. in this paper, we investigated an extended model of the classical sird pandemic model to characterize the regional transmission of covid-19 after lockdown in france. incorporating the time delays arising from incubation, testing and the complex effects of government measures, an exponential function of 395 the transmission rate β(t) was presented for the regional model. by fitting the regional model to the real data, the optimal parameters of this regional model for each region in france were derived. based on the previous results of the extended model, we introduced and simulated a network model of pandemic transmission between regions after confinement in france while considering age 400 classes. regarding the transmission rate β(t) for the network model, we selected the inverse function of the previous β(t) to contribute to the transport effect after lockdown. by using the same model and method, we simulated the pandemic network for all cities in franc to visualize the transport effects of the pandemic between cities. considering age classes, we discussed an integro-differential 405 equation for modelling the network of infectious diseases in the discussion part. because of the large volumes of data and complicated calculations needed for parameters calibration and simulation when considering many geographical areas and many age classes, the requirements in terms of computer hardware and software are rather high. in order to achieve accurate results, appropriate and 410 efficient data processing methods will be applied. moreover appropriate dedicated theoretical work is needed to study the integro-differential model derived in section 6. 31 all rights reserved. no reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted july 29, 2020. . https://doi.org/10.1101/2020.07.27.20161430 doi: medrxiv preprint in future works, we will formulate and study optimal control problems in order to balance the induced sanitary and economic costs. the lockdown strate-415 gies implemented in france should be evaluated and compared to the proposed optimal strategies. all rights reserved. no reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted july 29, 2020. . https://doi.org/10.1101/2020.07.27.20161430 doi: medrxiv preprint early dynamics of transmission and control of covid-19: a mathematical modelling study asymptomatic coronavirus infection: mers-cov and sars-cov-2 (covid-19) covid-19: identifying and isolating asymptomatic people helped 430 eliminate virus in italian village estimated effectiveness of symptom and risk screening to prevent the spread of covid-19 coronavirus : 450 personnes en quatorzaine après des cas de covid-19 déclarés dans uneécole de lyon an seir infectious disease model with testing and conditional quarantine effect of a 445 one-month lockdown on the epidemic dynamics of covid-19 in france population modeling of early covid-19 epidemic dynamics in french regions and estimation of the lockdown impact on infection rate covid-19 pandemic 455 control: balancing detection policy and lockdown intervention under icu sustainability basics and trends in sensitivity analysis predicting the number of reported and unreported 460 cases for the covid-19 epidemic in south korea modeling infectious diseases in humans and animals infectious diseases of humans: dynamics and control stochastic epidemic models with inference on the distribution of points in a cube and the approximate evaluation of integrals the levenberg-marquardt algorithm: implementation and theory a comparison of three methods for selecting values of input variables in the analysis of output from a computer code gouvernement français on fast multi-shot covid-19 interventions for post lock-down mitigation modeling infectious diseases in humans and animals modeling fish population movements: from an individual-based representation to an advection-diffusion equation the mathematics of diffusion the authors are very greatfull to sébastien da veiga, senior expert in statistics and optimization at safrantech (france) for the r codes used for 420 calibration and uncertainty calibration. key: cord-320773-zisujjsx authors: sabat, iryna; neuman-böhme, sebastian; varghese, nirosha elsem; barros, pedro pita; brouwer, werner; van exel, job; schreyögg, jonas; stargardt, tom title: united but divided: policy responses and people's perceptions in the eu during the covid-19 outbreak date: 2020-06-22 journal: health policy doi: 10.1016/j.healthpol.2020.06.009 sha: doc_id: 320773 cord_uid: zisujjsx to understand the public sentiment toward the measures used by policymakers for covid-19 containment, a survey among representative samples of the population in seven european countries was carried out in the first two weeks of april 2020. the study addressed people's support for containment policies, worries about covid-19 consequences, and trust in sources of information. citizens were overall satisfied with their government's response to the pandemic; however, the extent of approval differed across countries and policy measures. a north-south divide in public opinion was noticeable across the european states. it was particularly pronounced for intrusive policy measures, such as mobile data use for movement tracking, economic concerns, and trust in the information from the national government. considerable differences in people's attitudes were noticed within countries, especially across individual regions and age groups. the findings suggest that the epidemic acts as a stressor, causing health and economic anxieties even in households that were not directly affected by the virus. at the same time, the burden of stress was unequally distributed across regions and age groups. based on the data collected, we draw lessons from the containment stage and identify several insights that can facilitate the design of lockdown exit strategies and future containment policies so that a high level of compliance can be expected. the outbreak of covid-19 triggered a wide range of responses from governments in the european union. given that the disease was new and effective medical countermeasures did not exist in early 2020, governments had to adopt non-medical measures aiming at the containment and mitigation of covid-19. with the aim of "flattening the curve," these policies included bans on public gatherings, closures of academic institutions and public places, national and international mobility restrictions, confinement, and several others [1] . italy was the first country in europe to apply intervention measures from the beginning of march 2020 in response to the severity of the covid-19 outbreak. other eu countries followed soon afterward, using similar countermeasures around mid-march 2020 [2] . the adoption of these policies varied in their scale, stringency, and pace across countries. while most european states implemented confinement measures, the extent of limitations of people's freedoms differed across individual countries. lockdowns were usually strictest where the pandemic was deadliest (italy, spain, and france), imposing severe limitations on population movements. some governments chose less stringent versions of confinement or no lockdown at all, for instance, "an intelligent lockdown" in the netherlands or "freedom under responsibility" in sweden [3] . forced to react swiftly to the unfolding epidemic situation, policymakers in every country tried to balance the implementation of containment policies against numerous important factors with the priority mostly given to the protection of the population's health. consequently, there has been a lot of debate in every society about whether measures taken by the government were appropriate or not. some parts of the population have been voicing support for more severe containment policies to minimize the spread of the virus. such attitudes were likely fueled by people's worries about their health and the potential of their national healthcare system to withstand the epidemic. meanwhile, others expressed their concerns about the social and economic consequences of such policies, thereby advocating for less severe containment measures [4] . as the pandemic began to abate, governments started designing the lockdown exit strategies and restarting their economies. however, the risk that the new wave of the epidemic may happen did not disappear, especially given that the vaccine development takes a long time, and herd immunity was not achieved [5] . in this light, the issue of lifting lockdowns has become a new subject of public debate across and within european countries raising discussions about the appropriateness of timing, risks, and potential consequences of ending the confinement [6] . lifting lockdown restrictions creates acute dilemmas to the policymakers since the economic and human costs of any exit strategy seem to be closely linked together. taking a utilitarian approach in this situation could backfire if the society's understanding is not preliminarily secured or expectations are not fulfilled. policymakers and public health experts have to persuade their citizens to make behavior changes and respect future containment interventions while facing the difficulty of enforcing such regulations. therefore, it becomes crucial to understand people's worries about the pandemic and their perceptions of the effects of containment policies, so that the design of further policies and contingency measures is well-informed, and a high level of compliance can be expected from the population. moreover, trust in the government and social institutions may become central to achieving a successful implementation of future measures, whereas lack of it may turn detrimental to the fight against the pandemic. hence it is of paramount importance to understand who people trust most so that public health messages can be amplified using correct means of communication. we provide a timely description of the current situation and draw lessons from the containment stage to inform the design and implementation of the lockdown exit policies. in order to understand the public sentiment towards the covid-19 containment measures and to inform future policy development, we collected information on people's support for these policies, their worries in relation to the unfolding epidemic, and their trust in different sources of information. we surveyed over 7,000 people representative of the adult population in seven european countries: denmark, france, germany, italy, portugal, the netherlands, and the uk. the fieldwork was conducted online during april 2-15, 2020, using multi-sourced online panels provided by the market research company dynata. to ensure that the sampling frame was representative given the online nature of the study, the company applied diverse recruiting procedures to reach the general population (through open recruitment, loyalty programs, affiliate networks, mobile apps). it then used quotas to match the national census shares in each country. the questionnaire was designed by the authors of the study except for the worry items that were adopted from the world health organization (who) covid-19 snapshot monitoring project [7] . the questionnaire was carefully translated into six other languages by native speakers and then implemented using the qualtrics platform first as a pilot (10% of the sample in every country) and next as a large-scale survey. the data from the pilot study were included in the total sample. in each country, we collected data from a sample of 1,000 respondents representative of the national population in terms of region, age, gender, and education. given that the italian region lombardy was the most severely hit by the covid-19 outbreak, we collected 500 additional responses in this region representative in terms of age and gender. learning about perceptions and attitudes of people who reside there could provide essential insights to researchers and policymakers. the extra data collected from lombardy were not included in the representative sample of italy. thus, no weighting was used as the additional lombardy sample was analyzed separately and denoted as lombardy in the results section. we assessed people's approval of policy measures that were taken (or were likely to be taken) by their national government in response to the covid-19 outbreak. in particular, we covered such issues as school closures, bans on public gatherings, border closures, bans imposed on the export of medical equipment, fines for quarantine violations, random temperature checks, curfews, public transport suspensions and utilization of mobile phone data for tracking covid-19 cases and their contacts. on average, 68% of people in the seven european countries approved of the policies taken in their country in response to the pandemic, implying considerable public support. nevertheless, the extent of approval differed by country and by policy measure. the most approved measures were fining 14-day quarantine violations, ban of public gatherings, and border closures (each supported by 83% of respondents). by the time of the survey's fieldwork, restrictions on public gatherings had been adopted in all countries covered by the study, whereas international travel controls had been imposed to a certain extent everywhere, except the uk [8] . prior to complete border closures in mid-march 2020, some countries (for example, italy, france, germany, denmark) had been requiring screening and 14-day quarantine for arrivals from high-risk regions already since february. in contrast, other countries, such as portugal and the netherlands, started later and turned directly to strict measures, such as banning arrivals from high-risk areas and imposing partial border closures. the latter typically implied either limitation on entries of nonresidents or closure of only certain types of borders (land, sea, air), while ensuring "green lanes" for freight vehicles transporting goods. however, complete border closures occurred haphazardly and led to disrupted commerce and stranding citizens. among countries covered in our study, denmark was the first to close all borders in mid-march, whereas the uk did so only in the second half of may 2020. moreover, at the time of fieldwork, the uk did not have routine screenings at its airports or quarantine requirement for travelers [8, 9] . thus, the results for the uk showed the extent of public support that these measures would have received, had they been implemented earlier. meanwhile, the most opposed containment policies were public transport suspension (37% of respondents against it), ban of medical export, use of mobile phone data for tracking, and the imposition of a curfew (each disapproved by approximately 23% of respondents). these trends might reflect within-country regional and age structure of the population. for example, older individuals and those living in remote areas tended to be the most strongly opposed to public transport suspension. in fact, among countries covered by the survey, public transport suspension was implemented only in italy, whereas its volume was reduced in all other states except for germany [8] . the stay-at-home orders were most significantly opposed by the youngest respondents aged below 25. this measure was enforced in all countries covered by the survey except for denmark, where it was introduced as a recommendation [8] . overall, a north-south gradient could often be noticed in the eu regarding policy support: people living in the southern states (portugal, italy, and france) tended to approve of the containment policies more than residents in the northern countries (denmark, germany and the netherlands). noteworthy, the largest share of supporters for every containment measure was noticed among the residents of italy and particularly in lombardy. here, on average, 79% of the population approved of the government's response to the pandemic. interestingly, the most significant share of the population who explicitly opposed each of the containment policies taken by their government was identified in denmark. here, for example, 22% of respondents disapproved of school closures and 48% disapproved of the imposition of a curfew. in comparison, the average disapproval of these measures in other countries was around 8% for schools and 20% for curfews. the most polarizing opinions were observed concerning the use of mobile data for tracking covid-19 cases and their contacts. the most significant share of people explicitly opposing such policy was identified in denmark (34%), the netherlands (31%), and germany (25%). it was particularly disfavored by the youngest age group (33% of respondents aged below 25 against it). this policy received significant media attention as some countries and the european commission started the collaboration with telecom providers to access individual geolocation data for prediction and surveillance of covid-19 spread [10, 11] . as of march 2020, deutsche telekom provided german authorities with the anonymized data on the movement of its users. in italy, vodafone, windtre and telecom italia offered aggregated user data provision to the government for the same purpose. authorities in the lombardy region used mobile phone data to check compliance with the lockdown restrictions [10, 11, 12] . other countries either initiated the development of their own mobile phone tracking apps or cooperated on the creation of common software, such as the pan-european privacy-preserving proximity tracing (pepp-pt) project led by germany. however, the launch of the pepp-pt was delayed at the end of april due to the data protection concerns voiced by experts and even some of the project participants [13] . while proponents of the contact-tracing measures claim that using mobile data is of paramount importance in response to the covid-19 pandemic, many people worry about the government's use of technology due to possible privacy violations, thereby raising debates about the appropriateness of such social control measures [10, 13, 14] . according to our data, people in some european countries expressed considerable reluctance about supporting such policy, which therefore makes future compliance questionable. moreover, such privacy disputes, as in the case of the pepp-pt project launch, might trigger higher reluctance among the potential users to use any contact-tracing app in the future, which could be detrimental for the implementation of a viable tracing technology [13] . to better understand public opinion on certain policies, it is essential to look at the big picture and place obtained results into the national contexts. people's attitudes were likely based on their perceptions of the general state of affairs in their country, particularly in terms of the epidemic situation and restrictions they were subject to at that moment. in view of that, table 1 summarizes the scale of the pandemic and the stringency of government's response in seven european countries at four points of time spaced around april 12 (when the survey's fieldwork was 99% complete in every country). the public health situation in each state is described using total confirmed cases of covid-19 and total deaths attributed to covid-19, both measured per 1 million people and reported by the european centre for disease prevention and control [15] . the stringency of government's response is measured with the covid-19 government response stringency index, a composite measure of containment policies ranging from 1 to 100, where a higher value denotes a stricter response [8] . at the time of the survey's fieldwork, the epidemic situation was worst, and the stringency index was highest in italy and france [8, 15] . clearly, there was a north-south gradient in the stringency of government response: italy, france and portugal imposed more demanding policies than denmark, germany, the netherlands and the uk. nevertheless, although people in southern countries were exposed to more severe containment measures, they approved of them more than people residing in northern states, who experienced less stringent restrictions. turning now to within-country variations, we observed considerable heterogeneity of attitudes towards many policy responses within individual countries with particularly marked differences between regions and age groups in italy, france, and the netherlands. hereinafter, we grouped regions based on the severity of the covid-19 outbreak distinguishing between the most and the least affected areas. noteworthy, lombardy denotes the extra sample collected in italy and was analyzed separately from the representative italian sample. overall, we did not find significant differences in policy support between lombardy and the rest of italy. to illustrate within-country differences, fig.2 . reflects regional and age-related heterogeneity of public opinions in france and italy toward banning the export of medical equipment, such as masks. in fact, this measure was briefly undertaken by germany and france at the onset of the pandemic in early march 2020, leading to political tensions between the eu member states. germany declared that the reason was to avoid shortages of masks, gloves and safety glasses within the country, whereas france argued that the ban was needed for the assessment of inventory and storage capacity [16] . following the call for solidarity, both countries lifted the within-eu export ban on equipment in mid-march [17] . while support for this policy tended to be similar in the most and the least severely affected parts of italy and france, the approval of the export ban conspicuously differed across age groups. older individuals approved more of this policy than younger people, which, besides other factors, may be related to the levels of worry people in these age categories have about the risks that covid-19 poses to their health. we found that 51% of french and 46% of italian respondents aged above 65 perceived risks to their health from covid-19 as high or very high, while the corresponding share among people aged below 25 equaled 30% in france and 17% in italy. to address the mental health implications of the covid-19 outbreak and subsequent containment measures, we assessed levels of worry prevailing in european societies over several domains (health, economic, emotional, work, and future). more specifically, we addressed concerns about losing a close person, becoming unemployed, health system getting overloaded, school closures, small companies running out of business, recession, restricted access to food supplies, blackouts, and society getting more egoistic. these items were adopted from the who covid-19 snapshot monitoring project, which will allow future comparisons with similar data collected for other countries and at different points in time [7] . we found that the mean trend was similar in all countries: people worried most of all about the health system getting overloaded so that the capacities could become insufficient to cope with the surge in covid-19 cases. we observed that even in case of households that had not been directly hit by the novel coronavirus (above 75% of respondents in the total sample), the pandemic might have acted as a stressor causing health and economic anxieties. fig.3 presents people's worry about selected issues across seven eu countries (measured on a likert scale from 1-not worry at all to 5-worry a lot), where the higher intensity of color reflects a larger share of the population who worry "quite a bit" or "a lot". cross-country differences look substantial, and a north-south divide in the worry caused by the covid-19 outbreak is conspicuous. fig.3 . the proportion of respondents who worry "quite a bit" or "a lot" for instance, 84% of respondents in portugal and 81% in italy mentioned that they worried "quite a bit" or "a lot" about the national health system becoming overloaded, while the corresponding shares in denmark and germany were 54% and 62%, respectively. these health concerns might have reflected the development of the pandemic. as showed in table 1 , the progress of the epidemic had a north-south pattern with more covid-19 cases and deaths per million of the population in southern states than in northern. the exception was the uk, where the epidemic was third deadliest after italy and france, but government response was less strict than in countries with a better epidemiological situation [8, 15] . similarly, more people in portugal and italy were concerned with the economic consequences of the pandemic than in other european countries. for example, 68% of portuguese and 56% of italians were worried about losing their jobs, while respective shares in the netherlands and denmark were 27% and 16%, correspondingly. these cross-country differences in economic anxieties may be related to people's perceptions of the economic and financial countermeasures taken by their national government and the eu. during the pandemic, european countries implemented several fiscal and monetary measures to mitigate the economic impact of the covid-19 outbreak. these policies typically included support of wages under the reduced-hour scheme, postponement of tax payments for companies, direct financial supports and grants to small enterprises and self-employed, the extension of unemployment benefits, provision of capital buffers to banks, etc. [1] . nevertheless, there were substantial variations in the timing and specific content of these countermeasures across the states. to briefly overview the scale of economic support provided by the government in each of the seven countries, table 2 summarizes values of the economic support index, a composite measure reflecting income support and debt/contract relief provided by the national government to households [8] . it is measured on a 0 to 100 scale, where a higher value refers to a more substantial economic assistance. at the time of the survey's fieldwork, all countries provided some type of economic relief to their residents. nevertheless, the extent of such support was conspicuously different: france and the uk ranked highest, while denmark, germany, and italy ranked lowest [8] . hence, it may be possible that higher levels of economic concerns in some countries indicated people's beliefs in the insufficiency of the government's response, which will be subject to the analysis in the next waves of the survey. moreover, the composition of employment varies across the eu, especially in terms of informal and temporary employment. temporary contracts provide lower levels of social protection and job security to employees, but their prevalence has increased over the last years, particularly in the netherlands, italy, and france. as of 2019, the share of temporary employees in the total number of employed was highest in southern european countries: portugal (17.4%), france (13.3%), and italy (13.1%). in contrast, it was significantly lower in northern states: the uk (3.8%), denmark (8.3%), and germany (9.3%). the only exception was the netherlands, where temporary workers constituted 13.6% of all employees [18] . thus, such differences in the employment composition may be in part responsible for the cross-country dissimilarities in economic concerns. we also observed differences in the levels of concern within individual countries. fig.4 shows the extent of worry about the health system and a recession in italy. we grouped regions based on the severity of the covid-19 outbreak and distinguished the levels of anxiety across age categories. higher intensity of the color reflects a greater extent of worry. overall, the level of worry in the highly affected regions of the country was not higher than elsewhere in italy, except for the youngest age group. however, economic concerns tended to be unequally distributed across the age groups. for instance, worries about the recession and small companies running out of business were higher among older individuals than younger age cohort. this pattern was similar in all countries covered by the survey. we asked people about the main sources of information from where they received news about covid-19. the data show that overall 94% of respondents closely followed the news on the situation with covid-19, implying a high level of public awareness. regarding the sources of information, 86% of respondents mentioned receiving updates from the tv and 50% additionally searched for information on the internet. presumably, reliable information presented through the television emerged as the best channel to reach the population at large. next, we assessed the extent of people's trust in the information received from various sources in the context of the covid-19 situation. the trust in the following information sources was addressed: national government, the eu, the who, hospitals and gps, national news channels and newspapers, social media, relatives and friends. fig.5 shows mean values of trust in information from six selected sources across seven european states (measured on a likert scale from 1-no trust at all to 5-trust very much). higher intensity of the color reflects a higher level of trust in the information from a specific source. the data show that overall people had the highest levels of trust in information from hospitals, family doctors, and the who, followed by information from the national government and main national news channels. this ranking of sources by trust was similar in all countries covered by the survey, except for france, where citizens had a high level of confidence only in healthcare providers and placed relatively little trust in all other sources. moreover, a north-south divide could be noticed in the level of trust in information from the national government. trust was highest in denmark and the netherlands (more than 70% of respondents trusted "much" or "very much"), whereas it was lowest in france (27% of respondents had a high level of trust). furthermore, a similar north-south gradient was observed concerning the trust in the eu: trust was highest in denmark (45%), germany (40%), the netherlands (39%) and the uk (35%), whereas it was lowest in italy (24%) and france (21%). portugal was an exception to this case since the corresponding value here constituted 46%. finally, we also observed considerable regional heterogeneities in levels of trust within countries with particularly noticeable differences across individual regions in italy, france, and germany. fig.6 shows people's trust in information from the national government in the context of covid-19 in germany and france as an example, where the higher intensity of the color indicates a greater extent of trust. while trust did not differ significantly between regions grouped with respect to the covid-19 severity, it was heterogenous across the age groups. although the survey asked about the level of trust in information from different sources in the context of the covid-19 situation and not about the overall trust in institutions, these two are likely to be related. generally, trust reflects people's perceptions of whether institutions are doing what is right. thus, trust in the information they provide can be considered an indicator of the confidence that citizens have in these institutions [19] . the covid-19 pandemic raised new challenges for policymakers across the eu. the imminent threat to public health at the onset of the pandemic led most governments to impose a lockdown on society. however, as the peak of the pandemic abated, the focus of attention turned to the social and economic consequences of the containment measures. given that without acquired herd immunity the risk of a new wave of the epidemic remains high, and the production and distribution of vaccines may take 12 to 18 months [20] , governments must try to strike the right balance between effects on public health, social life and the economy when considering possible exit-strategies from the current lockdown situation. in the absence of medical intervention, policymakers and public health officials must resort to non-medical behavioral interventions. lifting the lockdown requires that citizens support and adhere to the policy measures that aim to contain the spread of the virus as social and economic activity gradually restarts. given the difficulty of enforcing such regulations, future measures need to be both well-designed and well-communicated to the public. the more people are willing to comply voluntarily with the new measures, the less enforcement and supervision will be needed to achieve high compliance. for this, people's perceptions and attitudes need to be factored in at the policy-design and implementation stages. our survey sought to capture the public sentiment toward measures previously taken by policymakers to contain covid-19 and addressed people's support for policies, worries about the consequences of covid-19, and trust in different sources of information. the first insights obtained from the data showed that containment and mitigating policies undertaken by national governments in response to the initial stages of the covid-19 pandemic were generally wellreceived by the population in all countries covered by the survey. nevertheless, the extent of approval varied across states and specific policy measures. several lessons can be drawn for the design and implementation of policies for the prolongation or gradual removal of lockdown restrictions. first, we observed a north-south divide in people's perceptions, worries and trust across the european countries. this finding suggests that further containment measures and lockdown exit strategies need to be balanced against the factors that worry people in each specific country. one noteworthy example is the level of importance that people in european countries attribute to the concepts of individual freedom and privacy. using mobile data for tracking covid-19 cases and their contacts may be a controversial decision to take even though it is believed by many experts to be a useful tool to manage the covid-19 outbreak. the effectiveness of this policy critically depends on a sufficient level of adoption of the technology by the population [8] . our data suggest that this may not be achieved easily in some european countries. a clear takeaway is that an open dialogue with society on this matter is needed. explaining the need for and the advantages of such intrusive policies through trusted means of communication, while addressing people's concerns explicitly and being open about the risks of using such policy measures may help raise the support and compliance in society to a sufficient degree. another critical issue is the balance between saving lives and saving livelihoods. according to the survey, people in southern european countries are substantially more concerned about the economic aspects of the covid-19 outbreak than people in northern european countries. economic anxieties, if left unaddressed, may have adverse effects on the mental health and wellbeing of the population, as well as cause downward adjustments in consumption behavior, thereby exacerbating the economic situation in a country if the recession indeed happens. second, we found considerable heterogeneities in people's approval of policies within individual countries. this tendency was particularly noticeable in france and italy. one possible determinant of regional differences in public support could be the extent of the devolution of decision-making in the country. on the one hand, devolution could enable regional or local authorities to make better decisions due to their better awareness of region-specific circumstances. on the other hand, it could harm the coordination of policy responses between the central and regional authorities within individual countries. thus, it is crucial to understand the determinants of such differences and address them to secure public support of future policies and ensure high compliance with government measures. furthermore, our results showed that the burden of stress tended to be unequally distributed across and within countries. even in case of households that were not directly hit by covid-19, the pandemic may have acted as a stressor causing health and economic anxieties. such worries may be detrimental to individual mental health and wellbeing, and they may become further exacerbated by the imposition of self-isolation policies. thus, it may be reasonable to consider an asymmetric approach to the design of exit strategies taking region-specific levels of support and worry into account. this includes the identification of vulnerable categories of the population not only in terms of health risks but also with respect to social and economic activities, and addressing their concerns satisfactorily. third, during a pandemic, public trust in the government and the information it provides is of paramount importance. to expect high compliance over extended periods of time, policymakers need to adopt effective strategies and means of communication whereby securing a sufficient level of trust and confidence from the society. as our results suggest, some countries were more successful in this respect than others. society needs to be well-informed about the dilemmas faced by policymakers, and for this, the communication between the government and the citizens must be clear and transparent. the data showed that 94% of respondents closely followed the news on the situation with covid-19 mainly using television to keep themselves updated. thus, television emerged as the best channel to reach the population at large, suggesting that presenting reliable information through this means is an effective strategy to follow. nevertheless, given that the data show regional and age-related heterogeneities in trust and policy support, it may be worth tailoring messages and means of communication to specific groups of the society. for example, cooperation with public figures and well-known experts can be used to deliver government and public health messages in a simple language, or local voices could be used to amplify such messages in individual regions of the country. overall, information provision, public education and effective communication strategies should be among the key guidelines for policymakers when implementing exit strategies and designing future containment measures so that these policies have public support and high compliance. additional waves of the survey are scheduled in june and august 2020. this will allow us to investigate in more detail how the population copes with the health, social and economic consequences of the covid-19 pandemic as the situation evolves. declarations of interest: none estimating the number of infections and the impact of non-pharmaceutical interventions on covid-19 in 11 european do low-trust societies do better in a pandemic? lockdown fatigue hits as europe enforces coronavirus restrictions answering the right questions for policymakers on covid-19. the lancet global health the new york times covid-19 snapshot monitoring (cosmo): monitoring knowledge, risk perceptions, preventive behaviours, and public trust in the current coronavirus outbreak oxford covid-19 government response tracker. blavatnik school of government temporary reintroduction of border control tracking and tracing covid: protecting privacy and data while using apps and biometrics how will governments know when to lift restrictions? european mobile operators share data for coronavirus fight deutsche welle on the responsible use of digital data to tackle the covid-19 pandemic complete our world in data covid-19 dataset seeks solidarity as nations after intense discussions, i welcome that de and fr now allow for export of #covid19 protective equipment. i will continue to follow supply to it closely. no single eu country can win this battle alone. #cooperation #solidarity temporary employees as percentage of the total number of employees trust in government, policy effectiveness and the governance agenda the race against covid-19 this project has received funding from the european union's horizon 2020 research and innovation programme under the marie skłodowska-curie grant agreement no 721402, the work was supported by funding under the excellence strategy by the german federal and state governments, as well as by the university of hamburg, erasmus university rotterdam, and nova school of business & economics lisbon -chair bpi | "fundação la caixa" on health economics. we thank our colleagues for their feedback and work on the adoption of the survey to national contexts: helen banks, joana pestana, maarten husen, laurie rachet jacquet, nicolai fink simonsen. key: cord-282095-cbfyydi3 authors: pierron, denis; pereda-loth, veronica; mantel, marylou; moranges, maëlle; bignon, emmanuelle; alva, omar; kabous, julie; heiske, margit; pacalon, jody; david, renaud; dinnella, caterina; spinelli, sara; monteleone, erminio; farruggia, michael c.; cooper, keiland w.; sell, elizabeth a.; thomas-danguin, thierry; bakke, alyssa j.; parma, valentina; hayes, john e.; letellier, thierry; ferdenzi, camille; golebiowski, jérôme; bensafi, moustafa title: smell and taste changes are early indicators of the covid-19 pandemic and political decision effectiveness date: 2020-10-14 journal: nat commun doi: 10.1038/s41467-020-18963-y sha: doc_id: 282095 cord_uid: cbfyydi3 in response to the covid-19 pandemic, many governments have taken drastic measures to avoid an overflow of intensive care units. accurate metrics of disease spread are critical for the reopening strategies. here, we show that self-reports of smell/taste changes are more closely associated with hospital overload and are earlier markers of the spread of infection of sars-cov-2 than current governmental indicators. we also report a decrease in self-reports of new onset smell/taste changes as early as 5 days after lockdown enforcement. cross-country comparisons demonstrate that countries that adopted the most stringent lockdown measures had faster declines in new reports of smell/taste changes following lockdown than a country that adopted less stringent lockdown measures. we propose that an increase in the incidence of sudden smell and taste change in the general population may be used as an indicator of covid-19 spread in the population. f ollowing similar decisions in china and italy, a strict lockdown was enforced in france beginning on march 17, 2020 to block the progression of covid-19 and alleviate pressure on hospitals. one issue currently faced by governments is how to conduct the progressive relaxation of the lockdown 1 , which needs to be conducted systematically and carefully to prevent subsequent outbreaks while facilitating economic activity and recovery. on may 7, 2020, the french government categorized each geographical area as being red or green, depending on their covid-19 prevalence. compared to green areas, red areas were characterized by: (i) higher active circulation of the virus, (ii) higher level of pressure on hospitals (i.e., ccru occupancy), and (iii) reduced capacity to test new cases (fig. 1a) . in each area, red/ green labels were used to define steps associated with the local relaxation of lockdown. the french ministry of health used the ratio of consultations for suspected cases of covid-19 to general consultations at the emergency room (er) in hospitals as an indicator to assess the active circulation of the virus (detailed in "methods" section). concurrently, changes in smell and taste are prominent symptoms of covid-19 [2] [3] [4] [5] , as has consistently. been demonstrated in many countries (e.g., iran 6 , spain 7 , france 8 , italy 9 , germany 10 , and the uk 2 , among others). more critically, these chemosensory changes generally occur earlier than other symptoms 9 and may constitute more specific symptoms than fever or dry cough 2, 11 . accordingly, monitoring self-reported changes in smell and taste could thus provide early and specific information on the spread of covid-19 in the general population and support health system monitoring to avoid daily ccru admission overflows. using data from a global, crowd-sourced study deployed in 30+ languages (global consortium for chemosensory research survey, gccr, see "methods" section), we tested whether changes in smell/taste at the population level could be used as an early indicator for local covid-19 outbreaks. as pre-registered (see "methods" section), our primary aim was to test the association between self-reported smell and taste changes and indicators of pressure in hospitals (covid-related hospitalizations, ccru admissions, and mortality rates) for each french administrative region over the last 3 months. our secondary aim was to examine temporal relationships between the peak of smell and taste changes in the population and the peak of covid-19 cases and the application of lockdown measures. the potential for self-reported smell and taste loss to serve as an early indicator of the number of covid-19 cases-and hence hospital stress-was tested in a natural experiment by comparing france with italy and the uk, which implemented lockdown with different timing and levels of stringency. here, we show that self-reports of smell/ taste changes are closely associated with hospital overload and are early markers of the spread of infection of sars-cov-2. changes in smell and taste are associated with overwhelmed healthcare systems. the relationship between self-reported changes in smell and taste by french residents (diagnosed as covid-19+ or not, see "methods" section and supplementary table 1 ) and estimators of local healthcare system stress was evaluated geographically. figure 1a depicts the geographical distribution in red and green regions (as defined by the french government) and participants who self-reported changes in their smell and taste. red areas of france account for 40.8% of the population. green areas are clustered into a group with both a low number of self-reported chemosensory changes and a low number of admissions to ccrus (fig. 1b) . red areas show an opposite trend (chi-square <1 × 10 −200 and biserial correlations p < 1.3 × 10 −2 ). a strong relationship exists between self-reported changes in smell and taste and the number of admissions to ccrus (r smell = 0.88, p = 8.9 × 10 −08 ). this correlation remained significant even after removing the two most impacted areas (alsace and ile de france, r smell = 0.72; p < 3 × 10 −04 ), indicating that the significant relationship is not driven solely by these two regions. strikingly, use of self-reported chemosensory changes produced a stronger correlation than the current governmental indicator of virus circulation (fig. 1c) . overall, smell/taste changes are better correlated with the number of covid-19 admissions to hospitals than the current governmental indicator i.e., the ratio of er consultations for suspicion of covid-19 to general er consultations (r smell = 0.81, p = 6.71 × 10 −06 vs. r gov = 0.60, p = 3.35 × 10 −03 ); the same pattern was found for the number of covid-19 related deaths (r smell = 0.75, p = 5.62 × 10 −05 vs. r gov = 0.58, p = 4.97 × 10 −03 see supplementary table 2 ). further, when smaller geographical areas were considered (france is divided into 96 administrative units, called departments), these correlations remained highly significant (e.g., admissions to ccrus: r smell = 0.76, p < 5 × 10 −19 ) (fig. 1c) . moreover, the three relationships (change in smell/taste versus covid-19-related hospitalization, resuscitations, and death) also remained highly significant when considering only individuals who were not clinically diagnosed by a medical professional but considering themself showing some symptoms of covid-19 (e.g., admissions to ccrus: r smell = 0.83, p = 1.65 × 10 −06 ). potential sampling bias due to regional media coverage of our survey (supplementary table 3 ) and self-reported chemosensory changes by region was ruled out by confirming these variables were not correlated (r < 0.01, p > 0.9). notably, relationships between pandemic markers and online searches related to chemosensation were also significant in france. google queries related to smell or taste loss ("perte odorat," "perte goût" in french) were correlated with the three measures of an overwhelmed healthcare system described above (e.g., ccru admissions: r smell = 0.8, p < 4 × 10 −03 , see supplementary table 2 ). changes in smell and taste are early markers of the effectiveness of political decisions. next, we examined the temporal dynamics in france of self-reported changes in smell/taste, the current governmental indicator (ratio of er consults), and the number of ccru admissions due to covid-19 before and after the lockdown period. as shown in fig. 1d , the peak of the onset of changes in smell/taste appeared 4 days after the lockdown and for these individuals, the first reported covid-19 symptoms occur even earlier. conversely, the governmental indicator of er consults only peaked 11 days after the lockdown, while the peak of ccru admissions was shifted later by 14 days. this is consistent with emerging data showing that covid-19-related changes in smell and taste occur in the first few days after infection 6, [12] [13] [14] . the robustness of smell and taste changes over time was assessed in two ways. first, we showed the peak of smell/taste changes remained the same regardless of our survey's completion date ( supplementary fig. 1a) . second, we observed the exact same peak when analyzing a separate french survey performed on 950 individuals and focusing on smell alterations in the french population independently of covid-19 (see "methods" section): the peak of olfactory changes again occurred 4 days after the lockdown decision, and this was independent of survey completion dates ( supplementary fig. 1b) . the robustness of smell and taste changes was also observed over age (supplementary fig. 2a ) and gender ( supplementary fig. 2b ). finally, we also show that the observed peak does not correspond to seasonal occurrence of allergies in france based on the ratio of consultations for allergy to general consultations at the emergency room ( supplementary fig. 3) . further, analyses of google searches confirm this temporal relationship: on the same days where survey participants report experiencing their first symptoms (around march 18, 2020), there was a peak of google queries for terms associated with early covid-19 symptoms (fever, cough, aches, supplementary fig. 4a ). a few days later, the peak of online queries for "taste loss" and "smell loss" is synchronized with the report of smell and taste changes ( supplementary fig. 4b ). one week later, queries for shortness of breath preceded the peak of ccru admissions ( supplementary fig. 4c ). collectively, these results indicate a significant fraction of french covid-19 patients followed the same symptom time course, experiencing initial symptoms at the very start of the lockdown, which might be representative of a peak of infection a few days before the lockdown. this is consistent with the ultimate goal of the lockdown, which was to decrease the number of new infections following implementation. thus, the period immediately prior to lockdown represents the expected peak of new infections. in france, a large population may have been infected two days before lockdown because that weekend was crowded and sunny and occurred over the course of election day. further, there were busier train stations and supermarkets in anticipation of a shortage of supplies during lockdown 12 . these data suggest that the short-term efficacy of a lockdown could be monitored by tracking changes in smell and taste in the population. to assess whether such a prediction might generalize to other countries, we performed parallel analyses with data from fig. 1 changes in smell and taste as indicators of overwhelmed healthcare systems: geographic and time-related approaches. a french regions were assigned a green or red status by the french government to guide local relaxation of lockdown protocols. dots represent people self-reporting smell and taste changes in a web-based survey. base map is from openstreetmap and openstreetmap foundation. b the number of covid-19-related ccru admissions (as of may 11, 2020) correlated with the number of self-reported chemosensory changes (between march 1 and may 11, 2020, total n = 3832). green dots correspond to regions with a post-lockdown level labeled green, and red triangles indicate regions considered red. values are standardized based on the number of inhabitants (inhab.) for each regions. the two red triangles with ccru admissions >5 are alsace and ile de france. the gray band represent the confidence interval of the linear smooth (formula 'y~x') r and p represent value of the test for association between paired samples, using one of pearson's product moment correlation coefficient, without correction for multiple comparisons. c colored bar represent the value of computed correlation coefficients (confidence intervals are depicted as thin black bars) between the number of ccru admissions per area and i) the number of people reporting smell and taste changes (n = 3832, blue), and ii) the governmental indicator (gov. indicator), ratio of er consults for covid-19 (orange). analyses were done both at the level of metropolitan regions (reg) and departments (dep). d temporal relationships in france between smell/taste change symptom onset (blue solid line, n = 1476), the governmental indicator (orange dashed line), and covid-19 cases in ccrus (gray bars) around the lockdown period (vertical dashed line). data are 7-day running averages, normalized to the day with the highest value. with different levels of severity (see fig. 2 ). we monitored the dynamics of confirmed covid-19 cases, self-reported first symptoms, and self-reported taste and smell changes, and compared them as a function of the governmental stringency index. immediately after lockdown, we found that the two countries with the higher stringency index experienced a more rapid decrease in both self-reported smell and taste changes and covid-19 symptoms. further, as expected, the evolution of confirmed covid-19 cases differs according to the stringency index. the governments of italy and france rapidly increased their stringency index, which led to a sharp decrease in covid-19 symptoms and cases. in contrast, in the uk, the number of people in the uk reporting symptoms showed a slower decrease, presumably due to a less severe lockdown policy, and the number of confirmed cases remained high during the observation window. in each country, self-reported smell and taste changes can be regarded as a useful metric to predict the dynamics of confirmed covid-19 cases. that is, when the number of new onsets of chemosensory changes decreases sharply (france and italy), the number of confirmed covid-19 cases also decreases, albeit with a lag of two weeks. on the contrary, a slow decrease in the number of new onset chemosensory changes is associated with a plateau of confirmed cases (uk). the present analyses reveal a strong spatial and temporal relationships between self-reported smell and taste changes and multiple indices of health care system stress, such as admissions to ccrus. this is consistent with cumulative evidence showing a high prevalence of chemosensory alterations in patients affected by covid-19 in europe (france 8,14 , italy 9 , uk 2,15,16 ). participants endorsed smell and taste changes only 3-4 days after their first symptoms. such early chemosensory estimators may represent a cost-effective and easy way to implement alternative surveillance methods to large-scale virology tests, which are difficult to perform, costly, and time-consuming, especially during a pandemic. a prominent question raised by these findings is whether the smell and taste changes observed in our study are solely related to covid-19 or whether they can be explained by other temporal patterns, like seasonal illnesses or allergies. to the best of our knowledge, there are no existing studies that have explored the dynamics of sudden anosmia (as in throughout the year in france. relationship between olfactory disturbances and seasons have been reported in korea, germany or us with a moderate increase of anosmia prevalence in spring [17] [18] [19] overlap, the amplitude of reported changes (either due to allergy or viral affection) were very limited compared to the present report. to further rule out the possibility, we examined whether the annual peak of allergies in france could explain the peak of smell and taste changes observed here. in analyzing existing french governmental data, we found that the annual peak of allergies in france occurred around week 30 (beginning of summer), multiple weeks after the observation window of the present study (from week 5 to week 20, supplementary fig. 3) . further, the french national aerobiological surveillance network (rnsa, https://pollens.fr), which follows pollen concentration in the atmosphere, has also indicated the first week of lockdown was very low risk for seasonal allergies. in addition, when considering google trends data, we did not observe any similar peaks in queries for smell/taste loss in the corresponding time period in previous years. finally, a comparative study in israel 20 showed that in covid-19 suspected patient the frequency of smell change is almost ten time higher in a covid-19 positive patients (68%) than in covid-19 negative (8%). considering that most of the participants of the present study are diagnosed with covid-19 and that their description of a sudden loss of smell/ taste is consistent with the now typical presentation of covid-19 symptoms, it is highly probable that covid-19 infection is the main reason of their smell and taste change. collectively, these data suggest the peak of smell and taste changes studied here are more consistent with sudden covid-19 viral infections rather than an artifact due to seasonal illnesses. the time lag between the onset of covid-19-related symptoms and their declaration by the respondents of our study also deserves comment. although immediate reporting of symptoms would have been ideal, such reporting is not possible within the context of the sudden first wave of a new viral pandemic. a similar time lag has been observed in other large-scale studies focusing on olfaction and covid-19 21 . indeed, this time lag is inevitable given the preparation time required for scientists and clinicians design and launch such a survey, with appropriate ethics approval, once anosmia and ageusia began to emerge as cardinal symptoms of covid-19. the vast majority of participants completed the survey between april 10th and april 19th, 2020, and most of them declared a date of onset of their symptoms roughly a month earlier (although a small fraction of participants did indicate onset prior to 2020). a possible consequence of a time lag between survey completion and the effective date of symptom onset is that subjects' statements may have been influenced by major societal events such as the lockdown decision, potentially creating some recall bias. to examine whether the date of a major event like the lockdown might bias dates of reported smell and taste loss, we explored narrative descriptions provided by our participants. by analyzing responses to the optional open-ended question "please describe the progression or order you noticed your symptoms", we observed that, for france, a mere 11 of 3705 people (who have filled the optional question) used the term "confinement" ("lockdown") in their description of the onset date. separately, another factor that mitigates concerns about a potential recall bias is the stable nature of participant's statements, regardless of their date of completion. that is, logic suggests, the longer the time between the onset date of smell and taste loss and the reporting date, the greater the recall bias should be. however, our data clearly show that regardless of the date of completion, the onset date falls within the same period ( supplementary fig. 1) . finally, other evidence against a potential recall bias comes from google trends data. analyzing real-time google queries in march, we observed a very particular trend in france (supplementary fig. 4) . we first observed a peak of queries for terms associated with early covid-19 symptoms (fever, cough, aches) synchronized with the declared onset of the first symptoms in the survey (around march 18th). a few days later, a peak of online queries for "taste loss" and "smell loss" was seen, and this was synchronized with the date reported of smell and taste changes in our survey. the striking concurrence between google queries and reports in our survey argues against the idea that a recall bias could be driving the effects described here. another important factor to consider in our survey is the way the press and media might have influenced our findings. indeed, when the survey was launched, smell and taste changes were reported as symptoms of covid-19 in the national and local media, which might have influenced respondents to remind themselves of such symptoms and to then report these changes on the survey. such an emphasis on smell and taste loss would have biased attempts to explore the prevalence of chemosensory deficits in covid-19. however, the primary aim of the present investigation was not to focus on the prevalence of anosmia and ageusia with covid-19, but rather to explore use of reported smell and taste loss as indicators of covid-19 pandemic. still, the media coverage of our survey could also have biased the selection of participants geographically, as some french regions received more media coverage than others. however, as reported above, there was no correlation between the number of participants in a given region and the intensity of media and press coverage for the survey in that same region. finally, when participants were asked to describe the chronology of their symptoms, they did not refer to the media coverage as a prominent element influencing their awareness of their smell/taste changes. while this does not exclude an implicit and non-verbalized bias due to media coverage, this pattern suggests a genuine report of symptoms with a high occurrence of covid symptoms just after the lockdown. an interesting question raised by our findings is what impact they might have on government strategies in a pandemic. following lockdown, the rapid decrease of self-reported changes in smell and taste in france may be representative of the effectiveness of this decision in reducing infection rates. similarly, data from italian participants show highly similar patterns, but with a one-week difference compared to the french data. this might reflect highly similar responses by the italian and french governments. conversely, the prevalence of chemosensory changes in the uk shows a more gradual decrease. the uk government began with advice to avoid pubs, clubs and theaters, and to work from home from march 16, with restrictions around march 18. however, a lockdown was not declared until march 23, and this was less stringent than those in france or italy. notably, new covid-19 cases in the uk showed a plateau phase which is not observed in either france or italy. accordingly, we conclude that collecting online information about changes in smell and taste from residents (even retrospectively) may be a valuable metric of the effectiveness of reopening strategies related to the covid-19 pandemic. practically, in areas where smell and taste changes are notable covid-19 symptoms, the proportion of individuals who selfreport changes in their ability to smell or taste might be an early indicator of subsequent demand for healthcare. if confirmed, continuous monitoring of changes in smell and taste perception would then be a highly cost-effective, minimally invasive, and reliable way to track future covid-19 outbreaks. when used this way, we caution that particular attention must be paid to potential selection bias. that is, self-report studies online can be impacted by multiple selection biases, including socioeconomic status, fluency with technology and willingness and interest in participating in scientific research. when considering the present data, at least 3 parameters may contribute to a selection bias in our sample: (1) the age, (2) the gender of the participants, and (3) the format and the advertising of the survey. regarding participant' age, our study cohort (mean 40.7 years, sd = 12.4)) was quite similar to the french population mean (41.1 years, according to insee, https://www.insee.fr/fr/statistiques/ 1893198); however, we did only include individuals over 18 due to issues of consent, and administrative reasons, and seniors were also less represented. for gender, our sample contained a greater proportion of women (67%) compared to men, which might influence the results. however, additional analysis showed no differences in peaks of smell/taste changes across age or gender, minimizing concerns that such selection biases may have influenced present results (see supplementary fig. 2) . we also tested the potential selection bias due to format and the advertising of the survey, by comparing the gccr dataset with an independent second study performed on french residents (see "methods" section). remarkably we observed highly similar results across studies where advertising, inclusion criteria, and survey format were different. based on the present findings, we highlight the paramount importance and robustness of associations between smell/taste changes and covid-19 and we strongly endorse the need for additional large-scale validation studies to assess the causality between the observed association between smell/taste changes and indicators of the covid-19 pandemic. this could be achieved by setting up a simplified interface where selection biases are controlled for (age, gender, motivation, media coverage, socioeconomic level, etc.) through both traditional and online media-and whereby real time information about changes in smell and taste in the general population may be available to decision-makers. subjects' participation in the questionnaire and the reliability of the answers should also be considered. in particular, if a participant knows how their answers may influence enforcement of lockdown, their answers might become less truthful. this motivation can be expressed through different forms of behavior. whereas some individuals may tend to provide statements that minimize their symptoms in order to avoid strict containment measures, others will maximize their declaration to maintain the lockdown, or will provide honest answers in order to participate in the collective effort to better understand the covid-19 pandemic. these motivational factors are a recurrent risk in online studies and different strategies should be held to control for them in future predictive studies. based on the above, a large implementation of the study of smell and taste changes in institutional models should allow for monitoring of covid-19 spread. this might be especially relevant in in areas in which testing proves difficult or delayed and for future outbreaks that may overlap with other seasonal viral diseases which share many of the symptoms (fever, cough etc.) but whose treatment or prevention (vaccination) are less demanding in terms of critical care than covid-19. we advocate that self-report surveys should be used to enhance other strategies such as large-scale pcr tests and covid-19 symptom assessments (including anosmia and ageusia) in primary/secondary care. in summary, we propose that an increase in the incidence of sudden smell and taste change in the general population may be used as a valuable minimally invasive indicator of coronavirus spread in the population. to formally test the temporal relationship between chemosensory changes and spread of the disease, we recommend that a large-scale causal study in different countries be conducted on real-time monitoring of self-reported changes in the ability to smell or taste. such a prospective study will allow for the creation of statistical models that can assist in prediction of future hospital admissions for covid-19. further, it could also help decision-makers take important measures at the local level, either in catching new outbreaks sooner, or in guiding the relaxation of local lockdowns, given the strong impact of lockdown on economic and social activities. online survey. this study is mainly based on data from the global consortium for chemosensory research survey (gccr, https://gcchemosensr.org/)a global, crowd-sourced online study deployed in 30+ languages 22 . the data analyzed here were collected from april 7 to may 14, 2020. the protocol complies with the revised declaration of helsinki and was approved as an exempt study by the office for research protections at the pennsylvania study university (penn state) in the u.s.a. (study00014904; pi hayes). participants in the gccr questionnaire were recruited by word of mouth, as well as through social and traditional media (flyers, social media, television, radio) during the covid-19 pandemic. it was well covered by the french press, as over 70 articles mentioned the project, at both the regional and national level (see supplementary table 3 ). respondents received no monetary incentive for their participation. inclusion criteria were as follows. (i) questionnaire completion was allowed only to participants who indicated they had suffered from a respiratory disease in the past two weeks, whether they noticed a change in their taste/smell or not. (ii) participants aged 18 years old or younger were excluded. for the analyses conducted in this article, only individuals reporting a change in smell and/or taste perception were included, based on the question "have you had any of the following symptoms with your recent respiratory illness or diagnosis?". moreover, to exclude unreliable entries, participants must have reported a quantitative decrease of at least 5 on a 0-to-100 rating scale between their ability to smell and/or taste before and during their recent respiratory illness or diagnosis. therefore, due to this inclusion criteria, "smell/taste change" is equivalent to a quantitative decrease of participant ability to smell and/or taste. we then extracted individuals from the full dataset who reported living in france, italy or the uk. as the country of residence was completed as a text entry, we allowed for typical variations (e.g., "united kingdom" or "uk"), spelling mistakes, use of different languages (e.g., "italie" or "italia"), as well as subdivisions (e.g., "scotland") and major cities ("paris"). metropolitan france was split into 13 so-called "regions" in 2016. however, we considered the former system where france was split into 22 regions here, since the organization of the health system mostly remains based on the structure built before 2016. an alternative, finer granularity, splits metropolitan france into 96 so-called "departments." to retrieve the french department and region of the participants, we used the city of residence they reported in the questionnaire and combined them with the french public website (data.gouv.fr, after a semi-manual correction of spelling). participants came from all metropolitan departments but three (mayenne, creuse, cantal). consequently, the number of responses analyzed in france was between n = 1476 and 4720 depending on the analysis conducted (i.e., on whether the information of interest was present or missing and the date range of analysis, see supplementary table 1 complementary and independent french survey. the data of another online survey were used to evaluate the robustness of the temporal evolution of smell and taste changes. this survey was conducted in the french population between april 8 and may 8, 2020 and aimed at characterizing chemosensory disorders in people with and without covid-19, as well as their consequences on quality of life. the data of 950 respondents were eligible for comparison with data from the gccr survey, i.e., responses where both the date of completion and the date of smell loss onset were provided. only responses that were complete and from people who were responding to the questionnaire for the first time and were over age 18 were included. this survey was approved by the cnrs ethics committee. data collection was strictly anonymous. the protocol complies with the revised declaration of helsinki and the study was approved by the ethics committee of the institute of biological sciences of the cnrs on the 3rd of april 2020 (dpo #trrech-467). all individuals provided informed consent when participating in the survey. online trends. trends of online queries by french region were performed using google trends, a tool returning the popularity of a search term in a specific state or region. google is by far the most used search engine in france (>90% of internet searches, according to statcounter global stats). we looked for the popularity of terms (listed in supplementary fig. 3 , using default selection of "all categories" and "web search"), within the timeframe of february 1, 2020 to may 10, 2020 (from the month of the first official covid-related death in europe to the end of lockdown in france). it should be noted that google trends does not provide the actual numbers of searches but rather a relative score from 0 to 100 (100 corresponding to the day with the greatest number of searches during the specified time period). to compare google trends scores between french regions, we transformed them by computing the relative number of queries per day in the region of interest. for example, despite a value of 100, the peak day might represent only 5% of the total number of queries related to the topic across the timeframe of interest (see above). healthcare system data. the french governmental indicator to estimate the circulation of the virus was calculated from the ratio of consultations for suspected covid-19 to general consultations at the emergency room (er) in hospitals. this ratio corresponds to the medical diagnostic for covid-19 suspicion (codes cim10: u07.1, u07.10, u07.11, u07.12, u07.14, u07.15, u04.9, b34.2, b97.2). the definition of covid-19 has evolved rapidly during the lockdown period but the diagnosis is principally based on symptoms of covid-19 considered as common such as fever, cough, and dyspnea (difficulty breathing). to the best of our knowledge, anosmia and ageusia were officially considered in france as putative symptoms of covid-19 from a letter of the direction générale de la santé (april 1st) and communication of the haut conseil de la santé publique (a letter dated april 8, published online april 15, following a letter from the cnp-orl dated march 20) . areas with values of the french governmental indicator higher than 10% are considered having a high virus circulation. this indicator contributes to the assignment of a red/green label. allergies incidence in previous years were calculated from the ratio of consultations for allergy to general consultations at the emergency room (er) in hospitals. data dealing with the health status across countries (number of covid-19 cases and deaths for each day) were downloaded on may 22, 2020 from the european centre for disease prevention and control databank (ecdc, https:// www.ecdc.europa.eu/en). data regarding healthcare system stress in france (hospitalizations, ccru entries and deaths) were also downloaded on may 22 from the french public health website (géodes, santé publique france, https:// geodes.santepubliquefrance.fr/#c=home). here, we use the term ccru (critical care resuscitation unit) to translate the french hospital service of "réanimation." raw data were normalized across regions with regard to their number of inhabitants as estimated by insee. the temporal evolution of the stringency of government response was retrieved from the oxford covid-19 (https://www.bsg. ox.ac.uk/research/research-projects/coronavirus-government-response-tracker). here, the stringency level of a country is computed according to which measures of a list of items (e.g., school closures, cancellation of public events, international travel controls, etc.) are undertaken. for the post-lockdown situation, the color assigned by the french government to each department was downloaded on may 12 from the government website. only data before may 11 (the initial lift of the lockdown) were included in the analyses. statistical analyses. statistical analyses were pre-registered at the open science framework (osf). data were analyzed using r software (4.0) and its standard packages (maps, ggplot, etc.). data were grouped at the national level (france, italy, uk). in france they were also grouped at the regional level (according to the division into 22 regions in place prior to the 2016 reform). the rationale behind this is that the healthcare system is still structured following this organization, with university hospitals in regional main cities serving patients of the surrounding departments. participants from overseas french territories were not included in the geographical analysis because of too few data (n < 10). the relationship between (1) gccr responses (or online queries), and (2) public health data was determined using parametric (e.g., pearson correlations) statistics as allowed by the normal distribution of the variable of interest. the association between gccr participant and red/green post-lockdown status was tested using chi-square tests and biserial correlations. complementary analyses not planned in the pre-registration included: (i) the analysis using the independent french online survey (see section "complementary and independent french survey" of the methods), (ii) the correlation between regional media coverage and the number of responses to the online survey per region, (iii) the correlation at the level of department, (iv) the correlation excluding extreme points, and (v) the correlation with the government indicator. pre-registered statistical analyses not presented here include: (i) mann-kendall trend test and change-point detection test to detect time series changes, and (ii) part of the google trends analysis. preparing for a responsible lockdown exit strategy real-time tracking of self-reported symptoms to predict potential covid-19 sudden and complete olfactory loss function as a possible symptom of covid-19 a new symptom of covid-19: loss of taste and smell corona viruses and the chemical senses: past, present, and future coincidence of covid-19 epidemic and olfactory dysfunction outbreak in iran acute-onset smell and taste disorders in the context of covid-19: a pilot multicenter pcr-based case-control study utility of hyposmia and hypogeusia for the diagnosis of covid-19 self-reported olfactory and taste disorders in sars-cov-2 patients: a cross-sectional study predictive value of sudden olfactory loss in the diagnosis of covid-19 selfreported olfactory loss associates with outpatient clinical course in covid-19 estimating the burden of sars-cov-2 in france anosmia and dysgeusia in patients with mild sars-cov-2 infection features of anosmia in covid-19 olfactory and gustatory dysfunctions as a clinical presentation of mild-to-moderate forms of the coronavirus disease (covid-19): a multicenter european study smell dysfunction: a biomarker for covid-19 a data science-based analysis of seasonal patterns in outpatient presentations due to olfactory dysfunction olfactory dysfunction from acute upper respiratory infections: relationship to season of onset epidemiology of anosmia in south korea: a nationwide population-based study self-rated smell ability enables highly specific predictors of covid-19 status: a case control study in israel relationship between odor intensity estimates and covid-19 prevalence prediction in a swedish population more than smell-covid-19 is associated with severe impairment of smell, taste, and chemesthesis data acquisition and curation reporting summary. further information on research design is available in the nature research reporting summary linked to this article. the authors declare that the data supporting the findings of this study are available within the paper and its supplementary information files. (source data file). source data are provided with this paper. r scripts are available on the osf server (https://osf.io/gew7p/). the authors declare no competing interests. supplementary information is available for this paper at https://doi.org/10.1038/s41467-020-18963-y.correspondence and requests for materials should be addressed to d.p., j.g. or m.b.peer review information nature communications thanks micael widerstrom and the other, anonymous reviewer(s) for their contribution to the peer review of this work. peer review reports are available.reprints and permission information is available at http://www.nature.com/reprintspublisher's note springer nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.open access this article is licensed under a creative commons attribution 4.0 international license, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the creative commons license, and indicate if changes were made. the images or other third party material in this article are included in the article's creative commons license, unless indicated otherwise in a credit line to the material. if material is not included in the article's creative commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. to view a copy of this license, visit http://creativecommons.org/ licenses/by/4.0/. key: cord-299846-yx18oyv6 authors: amar, patrick title: pandæsim: an epidemic spreading stochastic simulator date: 2020-09-18 journal: biology (basel) doi: 10.3390/biology9090299 sha: doc_id: 299846 cord_uid: yx18oyv6 simple summary: in order to study the efficiency of countermeasures used against the covid-19 pandemic at the scale of a country, we designed a model and developed an efficient simulation program based on a well known discrete stochastic simulation framework along with a standard, coarse grain, spatial localisation extension. our particular approach allows us also to implement deterministic continuous resolutions of the same model. we applied it to the covid-19 epidemic in france where lockdown countermeasures were used. with the stochastic discrete method, we found good correlations between the simulation results and the statistics gathered from hospitals. in contrast, the deterministic continuous approach lead to very different results. we proposed an explanation based on the fact that the effects of discretisation are high for small values, but low for large values. when we add stochasticity, it can explain the differences in behaviour of those two approaches. this system is one more tool to study different countermeasures to epidemics, from lockdowns to social distancing, and also the effects of mass vaccination. it could be improved by including the possibility of individual reinfection. abstract: many methods have been used to model epidemic spreading. they include ordinary differential equation systems for globally homogeneous environments and partial differential equation systems to take into account spatial localisation and inhomogeneity. stochastic differential equations systems have been used to model the inherent stochasticity of epidemic spreading processes. in our case study, we wanted to model the numbers of individuals in different states of the disease, and their locations in the country. among the many existing methods we used our own variant of the well known gillespie stochastic algorithm, along with the sub-volumes method to take into account the spatial localisation. our algorithm allows us to easily switch from stochastic discrete simulation to continuous deterministic resolution using mean values. we applied our approaches on the study of the covid-19 epidemic in france. the stochastic discrete version of pandæsim showed very good correlations between the simulation results and the statistics gathered from hospitals, both on day by day and on global numbers, including the effects of the lockdown. moreover, we have highlighted interesting differences in behaviour between the continuous and discrete methods that may arise in some particular conditions. france was hit by the sars-cov-2 epidemic probably at the beginning of january 2020, the first case being reported on 24 january [1], and went into lockdown on 17 march 2020 [2] . in response to the expected reduction of the number of cases, the french government eased the lockdown restrictions on 11 may 2020 and eased them again on 25 may (except in the ile-de-france region, where the density of population is very high). these measures have been taken to stop the exponential growth of the number of cases, as observed earlier in china [3, 4] . the basic reproduction number r 0 tells us the average number of new infections caused by an infective individual and it describes the exponential growth of the epidemic [5] . if r 0 is greater than 1 the epidemic will spread; otherwise, when r 0 is less than 1, the disease will gradually fade out [6] . compared to the r 0 of h1n1 (1.25) [7] the reproduction number of covid-19 indicates awful potential transmission. the r 0 was estimated as 2.2 [8] , 3.8 [9] and 2.68 [10, 11] by many different research sources around the world. the world health organization (who) published an estimated r 0 of 1.4 to 2.5 [12] . many approaches have already been used to model the covid-19 epidemic using compartment models and deterministic ordinary differential equations (ode) [13, 14] and also to estimate the effects of control measures on the dynamics of the epidemic [15] . these particular approaches give good results, but they do not take into account the stochastic nature or the spatial aspects of the propagation mechanism. however, stochastic differential equations (sde) have been successfully used to tackle the stochastic aspects of epidemic propagation [16] [17] [18] [19] . more recently, multi-region epidemic models using discrete and continuous models, taking into account the effectiveness of movement control have been published [20, 21] , as well as sde multi-region models [22] . stochastic models based on economic epidemiology have been applied to the covid-19 epidemic, for example, in south korea, to determine the optimal vaccine stockpile and the effectiveness of social distancing [23] . approaches using agent-based systems have also been used to model both the stochastic and spatial characteristics of epidemic propagation [24, 25] . in agent-based methods the number of machine instructions needed for each timestep, relative to the size of the data (algorithmic complexity), is at best proportional to the number of agents. those using one agent per individual may need a high computing power when used on large populations. these approaches are often applied to smaller areas (towns mainly) than the entire country, and/or use one agent to model a set of individuals (100 in [24] ). population-centred methods have an algorithmic complexity that does not depend on the size of the population, but on the number of rules considered at each iteration (for example, the number of reactions for biochemistry systems). when used on large populations these methods are much more efficient than entity-centred methods, but they do not take into account the spatial localisation. we adopted here a hybrid model derived from the sub-volumes method that adds coarse-grained spatial localisation capabilities to the standard stochastic simulation algorithm (ssa) used, for example, in the domain of biochemistry. to increase the computing efficiency we also used an original variant [26] of the gillespie algorithm with tau-leaping [27, 28] that automatically adapts the proportion of randomness vs. average-calculation, at each timestep. our implementation allows us to easily switch from this stochastic variant of ssa to a deterministic continuous solver (dcs), and therefore compare the two methods. to test our approach we applied it to the sars-cov-2 epidemic in france where relevant data [29, 30] have been made available throughout the duration of the epidemic. most of the simulation parameters we used have been obtained from statistics gathered in the literature, such as the proportion of cases that needed hospitalisation and the proportion of severe forms among them [31, 32] that needed beds in icu (intensive care unit). the number of infectious individuals and their localisations at the beginning of the epidemic have been inferred from statistical data made available by the french government and from the literature [33] [34] [35] . we used our simulation tool to ascertain the effects of control measures on the dynamics of the epidemic and compared the results to the real statistical data. we focused our study of the impacts of the epidemic only on the part of the population that moves on a daily basis: workers, pupils, students, retired people, etc. people in nursing homes were not taken into account since their environment and way of life are very different. starting from a known initial state, we wanted to compute a stochastic sample of the evolution in time of the number of people at each state of the disease. a transition between such states is often described by a set of probabilistic rules, or by a stochastic automaton. the epidemic spreading can be modeled as a markovian process in the sense that the number of people in each state at time t + ∆t depends only on the numbers at time t (and on other variables that do not depend on t). in most of the cases, it is not possible to find an analytic solution that gives those numbers as a function of time. hopefully, iterative numerical methods exist. one of them is the gillespie algorithm, frequently used to find the evolutions of the quantities of chemical species s(t) = {s 1 (t), ..., s n (t)} that can react according to chemical rules r = {r 1 , ..., r m } and their kinetics k = {k 1 , ..., k m }. starting from the initial value s(0) of the n species, the algorithm computes the values at time t > 0 by iterating the following process: 1. based on the quantities s(t), the rules and their kinetics, compute stochastically at what time each reaction is triggered {t 1 , ..., t m }. 2. let r i being the next reaction: t i = in f {t 1 , ..., t m }. 3. apply r i ; i.e., update the vector s(t i ) by decreasing the quantities of the substrates of r i and increasing the quantities of its products. 4. update the time: t ← t i . this algorithm gives an exact stochastic trajectory of the system, but can be slow when some reactions are quick. these quick reactions will often be triggered, so the time increment at each iteration will be small and the number of iterations per second high. to decrease the computing time, the tau-leaping method uses a fixed timestep, τ. at each iteration, the number of times each reaction is triggered during the time interval τ is stochastically estimated based on the quantities at time t. this method gives an approximation of the stochastic trajectory of the system, which is accurate as τ is small. the value of τ must be chosen to be large enough to minimise the number of iterations per second, but not too large to get good precision. the algorithm used in pandaesim, a variant of the tau-leaping gillespie method, is detailed at the end of this section. the population-centred methods such as those presented here share the same constraint: the entities evolving in the environment are considered homogeneously distributed in the environment. in other words, the spatial localisation is not taken into account. the entity-centred approaches, which compute the behaviour of each individual at each timestep, take into account the spatial localisation of each individual, but need much more computing power. to add coarse grained spatial localisation to our model, we partitioned the territory in sub-regions where one instance of a population-centred ssa is run. these instances use the same timestep and are synchronised. the interactions between sub-regions are modelled by taking stochastic samples of individuals that travel between sub-regions. this is done at a higher time scale since such travelling is less frequent than the travelling inside the original sub-region. most of the individuals that travel go back in their home sub-regions after a variable period of time. thus, the population of each sub-region remains approximately the same, although people enter and leave the sub-region. if this is not taken into account in the model, the population of each sub-region may tend to become the same as time goes on. we describe in the next section how this constraint is implemented in our model. the territory studied is partitioned in two levels of geographical organisation: region and sub-region. a region contains at least two sub-regions, a sub-region belongs to only one region and all the territory is covered (partition). in our case study, france, the first level is the administrative région, each one containing from two to a dozen départements. there are 13 régions and 96 départements in france. of course this can be applied to any partition of a territory. for example in england we could use the nine regions for the first level, and the 46 ceremonial counties and greater london for the second level. the population is divided into four age slices: 0 to 25 years old, 26 to 50 years old, 51 to 75 years old and over 76 years old [36] [37] [38] . each of these four sub-populations has its own values for the population parameters (infection immunity, travelling rate, etc.). we used one instance of a population-centred simulation process for each sub-region, with a one hour timestep. the simulation of the upper level (region) uses a bigger timestep, one day, and mainly processes the people which are travelling to another sub-region. thus, the population distribution is supposed homogeneous inside each sub-region, but can be heterogeneous at the region level and therefore at the level of the entire territory. depending on the age, and except for ill or hospitalised people, each day, people have a probability to travel from their homeplace to some place else either belonging to the same sub-region (local travel) or to another region (remote travel). these probabilities are part of the population parameters mentioned earlier. of course, quarantine type control measures forbid any kind of local or remote travel; people must stay in their respective homes sub-regions. the number of people of each age slice leaving their home sub-regions is a stochastic sample (or averaged value for the deterministic continuous solver) of a percentage of the population of this sub-region. for local travel, they are scattered according to the relative population of each sub-region belonging to their region. the more populated sub-regions attract more of the travellers. for remote travel, people go from their home-regions to the most populated sub-regions of the other regions, where airports and train stations are. the same method is used to dispatch the travellers according to the relative populations of their destination sub-regions. this way of computing how many individuals travel and where they go is a simple way to maintain constant the density of population of each sub-region. the sub-region population-centred model is a variant of the widely used susceptible, exposed, infectious and removed model. we added two states: hospitalised and deceased. the exposed and infectious states have slightly different meanings in our model; they have been renamed to asymptomatic and ill ( figure 1 ). unlike ill people, who show symptoms of the disease, recently infected people are asymptomatic hosts, but both of them are infective. hospitalised patients are also contagious, but to a lesser extent because they are confined inside the hospital. the three red dotted arrows in the figure indicate the potential sources and targets of the infection. we have assumed that people in recovered state are immune to the virus and therefore cannot be reinfected [39] . an incubation period of approximately five to six days before the apparition of the first symptoms has been observed [40, 41] . in consequence, in our model, asymptomatic people are subdivided into six subcategories according to the number of days since contamination. a large majority of cases, around 80%, present a mild form of the disease which is probably even not reported. the other cases need hospitalisation, and among them, from 5% [31] to more than 15% [32] present severe forms wherein patients need to be admitted in icu. the duration of the disease, after the incubation period, depends on the age of the patient an on the severity of the form of the disease. in our model it has been set to a maximum of 15 days, and therefore we have subdivided the ill (resp. hospitalised) people into at most 15 subcategories according to the number of days since the apparition of the first symptoms (resp. the date of the hospitalisation). people with mild infections will recover after a stochastically variable period of time (7 to 15 days) that depends on their age. the severe form of the disease is (stochastically) lethal according to a rate also varying with the age of the patient. the deterministic solver uses fixed average values. all these rates, probabilities and average durations are parameters of the model. their values came or were inferred from observed statistics of real cases. as mentioned before, the simulation algorithm uses a one hour timestep. it mainly computes in a stochastic way the state vector: i.e., the number of people that is in each state and subcategory, at each timestep. there are four state vectors, one for each age slice. of course these four vectors are not independent since whatever their age is, contagious people can infect susceptible people regardless of their own age. basically, from the value of the state vector at time t, the process computes the new value of the state vector at time t + τ (here τ = 1 h). thus, starting from a known initial value of the state vector at time t = 0, we can obtain its value at any time (t = t end ) > 0 by iterating this process until t end is reached, or until a specific value of the state vector is reached. pandaesim automatically stops the simulation when there are no more infective people. our model assumes that people have uniform daily routines. without specific measures, the daily schedule begins at 8 o'clock in the morning for work (or school, university, etc.) with the use of public transportation for one hour. next comes staying at work three hours, followed by a two-hour midday break, four hours in the afternoon at work, another hour in public transportation to go back home and the 13 remaining hours at home. we defined four possible environments, each one having its probability of contagion: home, public transportation, workplace and restaurant. these parameters have default values that reflect the local concentrations of people: very low at home, higher at work and restaurant and much higher in public transportation. to reduce the number of parameters we used the same value for the workplace and the restaurant. many kinds of measures can be used to slow down the propagation of the epidemic; we implemented two examples of such measures: 1. soft quarantine: people do not use public transportation at all and do not go to restaurants during the midday break. 2. full quarantine: this corresponds to what actually happened in france; people were confined at home except for a one hour stroll per day in low populated areas (public parks, forests, etc., were forbidden). again, to reduce the number of parameters, we assumed that the probability of contagion during the stroll was the same as at work. this also allowed us to take into account errands made to get food in more populated places such as groceries or supermarkets. starting from an initial state (number of contagious people in each sub-region), the simulation algorithm iterates the following process at each timestep until either the epidemic ends or the maximum duration of the simulation is reached (defaults to 720 days). 1. first, the infection rate at time t, i rt (t), is computed as the product of the global daily rate of infection, g dri (t), by the infection factor of the current location (home, workplace, public transportation) l in f (t). this infection rate i rt (t) is used the same way the propensity is in the standard ssa. then, for each of the four age slices the deterministic continuous solver computes the average number of individuals of that age that will go from susceptible to asymptomatic state, avnew asympt , as the product of the population in that state and the infection rate at time t: the stochastic discrete solver (sds) computes stochastic integer numbers such that, on the long run, they will average to the same values as the continuous solver. even when the population is an integer number of individuals, this product, avnew asympt , is generally a floating point number because the infection rate is itself a floating point number. this number has an integral part (≥0) and a fractional part (between 0 and 1). the (discrete) number of new asymptomatic hosts is then computed as the integer part of the average number, plus 1 if a uniform random number taken into the interval [0. . . 1] is below the fractional part: as the difference is 0.5 on the average, the higher the value is, the lower the relative impact of this stochastic discretisation becomes and the result is equivalent to a discrete averaged approach. conversely, the lower the value is, the more important the stochastic discretisation becomes. this mechanism allows the simulator to automatically choose the best strategy to adapt to the value range of the population [26] . 3. finally, when the current time indicates the beginning of a new day, t ≡ 0 (mod 24), individuals in each state either remain in the same state but shifted by one day, or change to another state. all the states transitions are computed stochastically by the sds (or deterministically by the dcs) using the method described earlier. • the population in the asymptomatic state that has on average reached the 5/6 day limit is moved to the first day of the ill state. • according to the illness duration by age slice parameter, a proportion of the population in the ill state is moved to the hospitalised or to the recovered state. the others remaining in the ill state one more day. • according to the disease severity by age slice parameter, a proportion of the population in the hospitalised state is moved to the deceased or recovered state. the others remain in the hospitalised state one more day. the global daily rate of infection is then simply computed by multiplying the constant of propagation of the virus, k prop , by the proportion of the total contagious population: by fitting the simulation results after the beginning of the lockdown to the data gathered from hospital statistics, we empirically found a good estimation of k prop for the sars-cov-2 to 0.75. we think that using pandaesim to model another type of epidemic, only this constant, along with the severity parameters, needs to be changed. we applied our simulation tool to the sars-cov-2 epidemic in france. we used the partitions of région and département in the country for the regions and sub-regions of our model. most of the parameters we used were gathered from the literature and statistical data made available by the french government. a few others were obtained empirically, mainly the number of contagious people in each région at the beginning of the simulations, and the constant of propagation of the sars-cov-2. the per-age values of the percentage of lethality [42] , illness duration and percentage of local and remote travellers are shown on table a2, the various rates of contamination on table a3 , and the initial number of contagious people in each département on table a1 in appendix a. in order to test our population-centred algorithm, we first ran simulations without countermeasures and without any travel possibility, either local or remote. these simulations were run using successively the stochastic discrete solver and the deterministic continuous solver. when the initial number of contagious people was relatively high, for example, in the val-de-marne sub-region (180), the results for both solvers were nearly identical: 5207 deaths for the average of 1000 stochastic runs and 5204 deaths for a deterministic run (figures 2 and 3) . the standard deviation for these 1000 runs went from ≈2 at the beginning of the simulations (with a few tens of deaths) to ≈41 at the peak of the infection (a few thousands of deaths), and then ≈5 at the end. the same kinds of results appeared for the ill people with the maximum value of the standard deviation of ≈2300 reached on the 90th day, with 137,381 ill people. on the other hand, when the initial number of contagious people was low, as in loiret (2), the dcs did not find any deaths, whereas 1000 runs of the sds showed two distinct behaviours; 127 of these runs showed the same results as the dcs, no deaths at the end of the epidemic. the 873 other runs took another direction leading to 4499 deaths on average with a standard deviation of ≈264 ( figure 4 ). the reasons for this apparent inconsistency will be explained in the discussion section. using the countermeasure applied in france (lockdown) the simulations showed us retrospectively that the probable date whereat there was a total of 897 contagious people in france (beginning of the simulations) was approximately the end of january 2020. this correlates with the period of time when the first deceased person was reported (24 january). the view of the main window of pandaaesim shown on figure 5 displays the real numbers of deceased people in each département. the map shown on figure 6 displays the mean values of 500 runs of a stochastic simulation. the overall results are very close, 19,877 for the real statistics and 19,764 for the mean value of the simulations. the département by département results are also fairly close, except for a few départements, but the orders of magnitude are more or less identical. to determine whether there is a form of convergence of stochastic trajectories to average values, we ran hundreds simulations and computed the mean value of the number of deaths (and of the other states) at each time step, in each département. the results showed no unique limit values, but the averages obtained with many runs stayed inside a range of values near the real statistics. we also ran pandaaesim using the deterministic continuous solver with the same parameters. the results were completely different: the epidemic ran only for 100 days (2 to 3 weeks less) and reported 7568 deaths (figure 7) , far from the 19,764 obtained with the stochastic simulations. the results département by département are also very different, with more than half the départements showing no deaths at all. again, probable reasons for this inconsistent behaviour are proposed in the next section. we developed a hybrid model and simulation programme derived from standard models and simulation techniques widely used in the fields of epidemic propagation and biochemistry. our approach used an original variant of the gillespie ssa with tau-leaping, where the inner algorithm can be easily switched from stochastic discrete to deterministic continuous. this allowed us to compare these two methods of simulation. to test our approach we applied it to the sars-cov-2 epidemic in france, for which relevant data were available. we also tested the consequences and the efficiency of the lockdown countermeasure applied in france for 55 days. in order to gain spatial localisation but with an efficient population-centred algorithm where the population was supposedly being homogeneous, we partitioned the territory into relatively small units for which an instance of the population-centred simulation was run. the movements of populations between these units were taken into account at a higher scale, with a larger timestep. we first tested one instance of our population-centred algorithm, where no countermeasure was used. using each method (sds and dcs) with the same parameters values, we compared the results in two different situations: (i) with a moderately high number, and (ii) with a very low number of initially contagious people. when the numbers were relatively high, the results of both methods were very similar. this was not surprising because at each timestep the absolute value of the increment computed by each method must be significantly higher than 1, and the stochastic rounding to the inferior or superior integer cannot be relatively very far from the floating point value computed by the continuous method. however, when the numbers are low, the absolute value added at the next timestep is only a bit higher than 0, and therefore the stochastic rounding to 0 or to 1 drastically changes the future trajectory. this is particularly important in this very case where the populations experience an exponential growth. this may look like chaotic behaviour since a small difference in initial conditions can lead to very different futures, but when the numbers grow, the importance of this switch effect is dampened. we used many simulations batches with initially only two contagious individuals in the sub-region. the results of 100, 200, 500 and 1000 simulations showed approximately the same proportions of cases, ≈12%, ending with no death at all, while the rest of the batch converged to approximately 4500 deaths. the same model using the dcs show no death at all. we think this behaviour is a consequence of a bifurcation due to the high non-linearity of the system. when the number of contagious individuals is below a certain threshold, the contagion tends to fade, but if this number goes over the threshold, there is a kind of positive feedback that increases it until a large enough part of the total population is removed. if we assume that the initial number of contagious individuals in our example (2) is below the threshold, the result shown by the dcs is therefore correct. due to both its discrete increments and its stochastic behaviour, the sds can sometimes compute a trajectory that goes above the threshold and switches the other way. in order to deepen the study of this bifurcation phenomenon, we have tried to find the approximate value of the threshold. first we used the dcs with the initial number of contagious individuals varying from 1 to 20. no deaths were found up to 15; then 38 deaths from 16 to 18; and 4508 deaths for 19 and above. then we did the same tests with 200 sds runs, counting the number of runs leading to zero deaths, and in the other case, the average number of deaths. with initially 1 to 5 contagious individuals, the number of runs leading to no deaths decreased from 70 to 2; with six and above initially contagious individuals no more simulations lead to zero deaths. for all the runs not leading to zero deaths, the average number of deaths was ≈4514. the threshold for the sds is somewhere below 5. as expected, this value is very low. then we tested the whole simulator with all the population-centred processes, running independently for 24 timesteps in each sub-region and then synchronised by exchanging a portion of each population either stochastically or deterministically. again, depending on the type of solver chosen and for the reasons mentioned earlier, the results were different but not by too much. with the number of people travelling from a given sub-region being a (small) fraction of the total population of this sub-region, the consequences in terms of infection spreading are very dependent on the value itself: less than 1, it is amplified by the stochastic processing, or else smoothed with the continuous calculation. both global results and sub-regions' local results were found to be very similar using the two methods. this can be explained by noticing that sub-regions with low initial contagious populations "benefit" from the migration of contagious people from more populated sub-regions, and as no countermeasure is applied, the number of contagious people grows rapidly over the threshold. the main difference appears in the shape of the nglobal curves: the deterministic solver showed a bigger dependency on the propagation effect ( figure 8 ). since the dates sub-regions had their peaks of contamination were very different, the propagation effect was slower. although the global number of deaths is approximately the same (379,336 for the dcs, 383,454 for the sds) the slope of the curve obtained with the sds is steeper than the one obtained with the dcs (figure 9 ). this can be explained by the relative sequentiality of the infection peaks showed by the continuous solver, whereas with the stochastic solver all the peaks are almost simultaneous and therefore the resultant is higher. for our last test, we set the simulator with the equivalent of the lockdown countermeasure used in france. the effect of this countermeasure was to decrease the number of contagious people, and while the sds gave results that correlate with the real statistics ( figure 5 ), the dcs did not work well mainly because the initial number of contagious people was too low to be taken into account (figure 7 ). more than half the départements did not show any death and therefore the total number of deaths was largely underestimated. we speculate that if we start from an initial state where there are enough contagious people in most sub-regions, it is very likely that the dcs will yield reliable results. this study gave us the opportunity to compare two different methods to get the trajectory of a complex system. at the beginning we were confident that they would yield very similar results, but facts proved us wrong. the reasons that caused the inconsistency of the behaviour of the stochastic discrete algorithm on the one hand and of the deterministic continuous algorithm on the other hand, lead us to be more confident in the stochastic approach for the simulation of this particular epidemic spreading model. more generally, with this type of model, an exponential growth phase is very sensitive to any variation, even small, in the initial values, and to artefacts, or calculation errors, and can therefore sometimes exhibit chaotic behaviours. nevertheless, this hybrid approach, a mix of an efficient population-centred process that plays the role of an agent in a multi-agent system, seems very promising. the stochastic simulations' results were very similar to the real statistics gathered from hospital data. future works could include improvements to the simulator such as the implementation of other types of countermeasures, the use more accurate methods to model the behaviour of individuals and the use different types of sub-regions to reflect their diversity. in this study we supposed no possible reinfection, so the epidemic effectively stopped after certain amount of time. although simplifying the model, this assumption forbids the possibility of modelling other waves of infection. recent publications discussed the consequences of different transmission scenarios, with and without permanent immunity, that can lead to multiple waves of infection [43] . an interesting perspective would be to include in our model a probability of reinfection in order to test the effectiveness of countermeasures. funding: this research received no external funding. acknowledgments: many thanks to martin davy at sys2diag, for the early version of the parameter dialog box, and the gathering of information about the sars-cov-2. the authors declare no conflict of interest. the following abbreviations are used in this manuscript: in order to fit the simulation results to the real statistics, we estimated the number of asymptomatic hosts in each sub-region (départements) at the beginning of the simulations (table a1) . per-age values of the percentage of lethality (extrapolated from [42] ), illness duration, and percentage of local and remote travellers (table a2 ). rates of contamination according to the location, percentage of hospitalised patients who can infect healing people, and proportion of severe form of the illness (table a3) . first cases of coronavirus disease 2019 (covid-19) in france: surveillance, investigations and control measures portant réglementation des déplacements dans le cadre de la lutte contre la propagation du virus covid-19. legifrance the effect of human mobility and control measures on the covid-19 epidemic in china an investigation of transmission control measures during the first 50 days of the covid-19 epidemic in china on the definition and the computation of the basic reproduction ratio r0 in models for infectious diseases in heterogeneous populations preliminary estimation of the basic reproduction number of novel coronavirus (2019-ncov) in china, from 2019 to 2020: a data-driven analysis in the early phase of the outbreak early estimation of the reproduction number in the presence of imported cases: pandemic influenza h1n1-2009 in new zealand early transmission dynamics in wuhan, china, of novel coronavirus-infected pneumonia novel coronavirus 2019-ncov: early estimation of epidemiological parameters and epidemic predictions nowcasting and forecasting the potential domestic and international spread of the 2019-ncov outbreak originating in wuhan, china: a modelling study transmission interval estimates suggest pre-symptomatic spread of covid-19 coronavirus latest: scientists scramble to study virus samples transmission dynamics of the covid-19 outbreak and effectiveness of government interventions: a data-driven analysis the effectiveness of quarantine and isolation determine the trend of the covid-19 epidemics in the final phase of the current outbreak in china centre for the mathematical modelling of infectious diseases covid-19 working group the effect of control strategies to reduce social mixing on outcomes of the covid-19 epidemic in wuhan, china: a modelling study the behavior of an sir epidemic model with stochastic perturbation the long time behavior of di sir epidemic model with stochastic perturbation a stochastic sirs epidemic model with infectious force under intervention strategies a stochastic differential equation sis epidemic model a multi-regional epidemic model for controlling the spread of ebola: awareness, treatment, and travel-blocking optimal control approaches a multi-regions sirs discrete epidemic model with a travel-blocking vicinity optimal control approach on cells role of media and effects of infodemics and escapes in the spatial spread of epidemics: a stochastic multi-region model with a study on herd immunity of covid-19 in south korea: using a stochastic economic-epidemiological model epidemic spreading in urban areas using agent-based transportation models an open-data-driven agent-based model to simulate infectious disease outbreaks hsim: an hybrid stochastic simulation system for systems biology a general method for numerically simulating the stochastic time evolution of coupled chemical reactions stiffness in stochastic chemically reacting systems: the implicit tau-leaping method données en santé publiques info coronavirus covid 19 clinical characteristics of coronavirus disease 2019 in china critical care utilization for the covid-19 outbreak in lombardy, italy: early experience and forecast during an emergency response cluster of covid-19 in northern france: a retrospective closed cohort study the french connection: the first large population-based contact survey in france relevant for the spread of infectious diseases cmmid covid-working group, estimating the infection and case fatality ratio for coronavirus disease (covid-19) using age-adjusted data from the outbreak on the diamond princess cruise ship estimating the asymptomatic proportion of coronavirus disease 2019 (covid-19) cases on board the diamond princess cruise ship reinfection could not occur in sars-cov-2 infected rhesus macaques the incubation period of coronavirus disease 2019 (covid-19) from publicly reported confirmed cases: estimation and application serial interval of covid-19 among publicly reported confirmed cases projecting the transmission dynamics of sars-cov-2 through the postpandemic period this article is an open access article distributed under the terms and conditions of the creative commons attribution key: cord-340298-0l4pec0z authors: terriau, a.; albertini, j.; poirier, a.; le bastard, q. title: impact of virus testing on covid-19 case fatality rate: estimate using a fixed-effects model date: 2020-05-01 journal: nan doi: 10.1101/2020.04.26.20080531 sha: doc_id: 340298 cord_uid: 0l4pec0z background in response to the sars-cov2 pandemic, governments have adopted a variety of public health measures. there are variations in how much testing has been done across countries. south korea, germany, and iceland take the bet of massive testing of their population. whereas tests were not performed widely in southern european countries. as the former undergo a lower case-fatality rate due to the covid-19 than the latter, the impact of the testing strategy must be investigated. in this study, we aimed to evaluate the impact of testing on the case fatality rate. methods we use data on inpatients across french geographic areas and propose a novel methodology that exploits policy discontinuities at region borders to estimate the effect of covid-19 tests on the case-fatality rate. in france, testing policies are determined locally. we compare all contiguous department pairs located on the opposite sides of a region border. the heterogeneity in testing rate between department pairs together with the similarities in other dimensions allow us to mimic the existence of treatment and control groups and to identify the impact of testing on mortality. results the increase of one percentage point in the test rate is associated with a decrease of 0.001 percentage point in the death rate. in other words, for each additional 1000 tests, one person would have remained alive. conclusion massive population testing could have a significant effect on mortality in different ways. mass testing may help decision-makers to implement healthcare measures to limit the spread of the disease. the increase of one percentage point in the test rate is associated with a decrease of 0.001 percentage point in the death rate. in other words, for each additional 1000 tests, one person would have remained alive. massive population testing could have a significant effect on mortality in different ways. mass testing may help decision-makers to implement healthcare measures to limit the spread of the disease. since it was reported in late december 2019 from hubei province in china, the severe acute respiratory syndrome coronavirus 2 (sars-cov2) has now spread worldwide with more than 2 million confirmed cases by the end of april 2020. 1 the outbreak reached europe via italy at the end of february and quickly affected the entire continent, making europe the epicenter by mid-march. the world health organization (who) declared the sars-cov2 to be a pandemic in mid-march 2020. while research is still underway to find a curative treatment, the increasing number of severe cases admitted to hospital has raised fears of overburdening the health care systems. to prevent such a situation, governments have implemented various public health measures such as mobility restrictions, social distancing, or mass screening strategies. on march 16th, the head of the who pronounced in favor of massive population tests, because "you cannot fight a fire blindfolded". 2 yet, there is a growing debate about the impact of mass testing on mortality rates. 3 we have observed strong differences in testing rates between countries; for example, south-korea, germany, or iceland, have undertaken important screening policies and now report low casefatality rates. on the contrary, countries like spain or france have restricted access to diagnostic tests for inpatients or health care workers and now report higher mortality rates. 4 unfortunately, cross-country comparisons are difficult due to the strong heterogeneity among countries. even in the united states of america, endowments for medical centers and lockdown strategies are very different from one state to another. by contrast, france kept a relatively centralized health system but as the epidemic was expanding, the health regional agencies (ars) were given autonomy in terms of screening strategies implementation; however, at the same time, a strict lockdown approach was instituted for all regions. 5, 6 among french regions, the main difference in their strategies was the intensity of testing policies. screening policies and mortality rate might be related to the fact that testing allows authorities to detect and isolate infected people and to prevent them from transmitting the virus; and also enables early treatment, thus increasing the chances of cure. 7 we propose a novel approach to assess the impact of focused screening strategies on mortality rates, which exploits policy discontinuities at region borders and contiguous department pairs that are located on opposite sides of a region border. this methodology has been used in an economic setting to evaluate the effects of the minimum wage on earnings and employment in the us. 8 we conducted a retrospective study, with a prospective database, including the total of patients who were admitted to hospital and afterwards discharged, the total of casualties and the total of tests performed for screening covid-19 infection (rt-pcr) by out-of-hospital medical laboratories. the sample covers the period from 19/03/2020 to 17/04/2020, which corresponds to a lockdown period in france. all the information was provided daily by the french public health agency (santé publique france; https://www.data.gouv.fr/). the data was gathered from different geographic areas within france and no other countries were included. we merged this dataset with . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. (which was not certified by peer review) the copyright holder for this preprint this version posted may 1, 2020. . information on hospital occupancy rates for intensive care units published by the french ministry of health (https://www.sae-diffusion.sante.gouv.fr/). sociodemographic data were extracted from the national institute of statistics and economic studies (https://statistiques-locales.insee.fr/). our analysis took place at the department level. in our study, we exploit the fact that from 14/03/2020, the french government has activated the third stage of the national plan for the prevention and the control of the epidemic, which translates into non-systematic testing of symptomatic individuals. from this date, testing policies were determined at the region level by the regional public health agencies. we used a fixed-effects model to assess the impact of the number of tests performed over time at a local geographical level (department) on fatality-cases. in fixed-effects models, subjects serve as their own controls, providing a means for controlling omitted-variable bias. otherwise stated, fixed-effects models allow controlling for time-invariant heterogeneity, i.e. all possible characteristics that do not change over time. 9 we used two distinct samples: i) the all-department sample (ad sample) that includes 94 departments distributed across 12 regions; ii) the contiguous border department-pair sample (cbdp sample) that contains all the contiguous department pairs that straddle a region boundary. metropolitan france counts 96 departments. we excluded two departments, haute-corse and corse-du-sud, that are part of a region, corsica, that does not share any direct border with others. among the 94 departments, 69 lie along a region border. as each department may belong to several department-pairs, we have a total of 237 distinct department-pairs. our strategy consisted in comparing all contiguous department pairs sharing a region border (see figure 1 for an example) to identify the effect of testing on the case fatality rate. tests rate and death rates were calculated using the number of rt-pcr tests and the number of deaths related to covid-19 divided by the number of patients admitted to the hospital, respectively. we first estimate the effect of testing on case-fatality rate using the canonical fixed-effects model and the ad sample (specification (1)): where denote the department, the time, is the case fatality rate in department at time , represents the percentage of people hospitalized that are tested in department at time , is a department fixed effect, and an error term. we now turn to our preferred identification strategy, which exploits policy discontinuities at region borders. to achieve identification, we estimate the following model using the cbdp sample (specification (4)): = + 2 + + + . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted may 1, 2020. where denote the department, the department-pair, the time, is the case fatality rate in department in department-pair at date , represents the percentage of people hospitalized that are tested in department in department-pair at date , represents a department fixed effect and a department-pair fixed effect. standard errors are clustered on the region and border segment separately to account for possible correlation in the residuals. 8 although fixed-effects models control for all characteristics which do not change over time, we report some time-invariant variables in table s1 for information. the average number of tested individuals was 613.43 per department with a share of positive cases of approximatively 25 percent. we count a total of 40762 hospitalizations. the observed share of the population above age 65 was roughly 21 percent. as shown by figure 2 , the number of deaths increased quickly, from 154 on march 19 th to 11532 on april 17 th . over the same period, the number of tests increased from 1713 to 133108. the path of mortality and testing was not homogenous over the territory. the autonomy given to regional public heaths agencies generated unprecedented differences in testing rate across regions and strong discontinuities at region borders (see figure 3) . the department fixed effect captures time-invariant heterogeneity across departments. this includes sociodemographic variables (such as the structure of age, race, or gender in the population), but also many variables related to health facilities (number of hospitals, medical density or medical devices). we add time-varying confounding factors in specifications (b) and (c). specification (b) includes the occupancy rate of the resuscitation units, while specification (c) also controls for the rate of positive tests. the first variable controls for the capacity of hospitals to treat patients at different stages of the covid-19 epidemic while the second controls for selection bias. table 1 reports the estimates provided by specifications (a)-(c). our baseline estimates reveal that a 1 percentage point (pp) increase in the tests/hospitalizations ratio leads to a statistically significant decrease in the case mortality rate by slightly less than 0.001pp. finally, table 2 displays the results for specification (d)-(f). our estimates reveal that a 1 pp increase in the tests/hospitalizations ratio leads to a statistically significant drop of case mortality rate by 0.001 pp. putting these numbers into perspective involves that for each additional 1000 tests, one person would have remained alive. sars-cov-2 outbreak is one of the major public health emergencies of international concern for decades. countries have implemented various measures mostly based on mobility restriction, social distancing, and regional or national lockdown. all of these public health measures are aimed at "flattening the curve" of the infected cases to limit avoidable mortality due to overburdened health care systems. we evaluate the effect of mass screening covid-19 on mortality rate in france during the first month of the lockdown. we take advantage of the difference in screening . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted may 1, 2020. . intensity among french regions. we first estimate the effect of testing on case-fatality rate using the canonical fixed-effects model and the ad sample and find that the increase of screening rate of 1 pp allows mortality rate to decrease of nearly 0.001 pp. we confirmed our results by estimating the fixed-effects model using the cbdp sample which compares contiguous french departments sharing region borders. to the best of our knowledge, no large randomized controlled trial (rct) has been implemented to investigate the effect of tests on the case-fatality rate, probably due to time, budget, or ethical constraints. when rct are difficult to implement or unethical, natural experiments (ne) are one of the best alternatives. the principle of ne is to mimic the existence of treatment and control groups using an instrumental variable that induces a change in the explanatory variable but has any direct effect on the outcome. however, in the case of the covid-19 epidemics, finding a suitable instrument remains a hard task. in the absence of rct or ne, many researchers try to approximate using standard methods such as linear regression, logistic regression, or propensity scores. however, such methods are subject to the well-known omitted-variable bias, leading to severe bias in estimating the effects of the variables that are included. consequently, causal inference via statistical adjustment represents a poor alternative to randomized experiments. in such a context, panel data models represent the best way to control for heterogeneity and to improve causal estimation. 8 we use a fixed-effects model because it represents a powerful tool for longitudinal data analysis. 9 however, such a model requires substantial differences between treatment intensities across entities and time to get precise estimates. our data meets these conditions: i) no region has the same test rate path than other regions over the period considered; ii) the test rate varies greatly across regions and time. methods based on regional controls and policy discontinuities have several advantages: i) contiguous border departments are relatively similar, in particular with regard to health trends, which are of major importance in the context of an epidemic; ii) the testing policy is determined at the region level and is largely exogenous from the point of view of a department, which rules out potential reverse causality. 8 until a vaccine is developed, the only way to prevent an unrestrained scenario is to control the spread of sars-cov-2. this is a challenging task because some asymptomatic infected patients could potentially spread the virus. literature reports an alarming proportion of asymptomatic infected cases. epidemiological data from the diamond princess cruise sheep revealed only 18% of positive cases reported no symptoms. 10 two hospitals in new york implemented universal testing for sars-cov-2 with nasopharyngeal swabs in women who were admitted for delivery, and revealed that nearly 90% of patients who were positive for sars-cov-2 at admission reported no symptoms. 11 overall population screening in iceland revealed that only 57% of participants with positive tests reported symptoms of covid-19. 12 this proportion could be even higher, because of false negative results of tests to detect sars-cov-2. 13 testing is part of a strategy to limit the transmission of the virus and who recommends a rapid diagnosis and isolation of cases in combination with a rigorous tracking and precautionary self-isolation of close contacts. several authors support the implementation of mass screening policies. 3, 14 in our opinion, mass screening may positively impact the fatality case rate in different ways. first, unfocused testing, i.e. not limited to symptomatic subjects, could improve the monitoring of the progress of the epidemic and facilitate decision-making by the health authorities. the use of "case definition", given the limited knowledge of the new disease, probably resulted in a low sensitivity to detect infected subjects, resulting in a delayed perception of the progression of the epidemic. 15, 16 screening strategies are subject to the availability of tests which indirectly shapes epidemic curves. 17 while the usa increased their screening capacities between late-february to early-march, the country experienced a rapid increase of total infected cases. 18 second, mass screening may also allow early identification of infected subjects and rapid implementation of isolation measures. early reports from wuhan suggest that public health interventions combining universal symptoms survey, traffic restriction and home quarantine resulted were temporarily associated with an increased control of the outbreak. 19, 20 a modeling from singapore suggests that quarantining of infected individuals and their family members, school closure and workplace distancing could reduce the progression of the epidemic but is associated to a significant economic cost. 21 review from the cochrane database concludes that quarantine is important in reducing the number of covid-19 cases but is dependent on screening strategies. 22 also, a us survey on the impact of school closure on mortality reports that the transmission prevention by school closure needs to be weighted with the potential loss of health-care workers. 23 this supports that public health decisions should be as focused as possible in order to limit the negative impact on the economy and the society. 24 importance of rapid diagnosis and case identification and isolation will become of utmost importance with the end of lockdowns. our study far supports a significant impact of screening strategies on the case-fatality rate in france. notwithstanding, there are some limitations to our results. first, they belong to france and it would be very hazardous to pretend that they apply to other countries because their exposition to covid-19 is different, they adopted different strategies, and have different health structures. second, to provide further evidence on this relation, it would be worth applying this methodology to other countries for which such data are available and in which testing policies are sufficiently heterogeneous across geographical areas. in addition, the data on tests collected by the french public health agency are those made by private laboratories and do not include those made in public hospitals. this represents an important share of tests (between half and two thirds) and we cannot rule out the possibility that this unobservable amount of screening activity may affect our results. lastly, our study cannot quantify the respective contribution of the treatment delivered to screened and infected individuals or the lower dissemination of the virus that results from quarantining policies. covid-19 intensive screening policies were significantly associated with a decrease in the fatality-case rate in france. these results support the implementation of mass screening strategies and could provide important information for decision-makers in the fight against sars-cov2 pandemic. the optimal testing strategy might also concern economic issues. indeed, the bank of france estimated that each fortnight of lockdown costs to france 1.5% of annual gdp (nearly usd48 billions). 25, 26 from a costs/benefits perspective one might naturally wonder what is the optimal policy capable of containing the outbreak and lowering the fatality rate. this is in our research agenda. . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted may 1, 2020. . we thank dr natalia lucia gomez, from the hospital italiano de buenos aires, for it's attentive revision of the manuscript and pr emmanuel montassier, from the university of nantes for his expert advices. . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted may 1, 2020. . standard errors between parentheses. *** p<0.01, ** p<0.05, * p<0.1. source: santé publique france and authors calculations. . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted may 1, 2020. number of regions 12 12 12 standard errors between parentheses. *** p<0.01, ** p<0.05, * p<0.1. source: santé publique france and authors calculations. . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted may 1, 2020. example with "nouvelle aquitaine" and "occitanie" regions. . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted may 1, 2020. . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted may 1, 2020. . test rate: number of rt-pcr tests divided by the number of patients admitted to the hospital for covid-19. death rate: number of deaths in hospital due to covid-19 divided by the number of patients admitted to the hospital. we use shapefiles for regions and departement to construct the maps and compute the contiguity matrix (https://www.data.gouv.fr/fr/datasets/contours-desregions-francaises-sur-openstreetmap/; https://www.data.gouv.fr/fr/datasets/contours-desdepartements-francais-issus-d-openstreetmap/ ). reading: paca belongs to the top 20% of regions that test more and to the bottom 20% of regions that have the lowest fatality ratio. . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted may 1, 2020. note: sample means are reported for all departments in france and for all contiguous border department-pairs with a full balanced panel of observations. source: santé publique france . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. (which was not certified by peer review) the copyright holder for this preprint this version posted may 1, 2020. . https://doi.org/10.1101/2020.04.26.20080531 doi: medrxiv preprint . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. (which was not certified by peer review) the copyright holder for this preprint this version posted may 1, 2020. . https://doi.org/10.1101/2020.04.26.20080531 doi: medrxiv preprint who covid-19 dashboard who announces covid-19 outbreak a pandemic covid-19 mass testing facilities could end the epidemic rapidly covid-19: how doctors and healthcare systems are tackling coronavirus worldwide health system review secretariat-general for national defence and security. national pandemic influenza prevention and control plan likelihood of survival of coronavirus disease 2019 minimum wage effects across state borders: estimates using contiguous counties fixed effects analysis of repeated measures data covid-19 outbreak on the diamond princess cruise ship: estimating the epidemic potential and effectiveness of public health countermeasures universal screening for sars-cov-2 in women admitted for delivery spread of sars-cov-2 in the icelandic population correlation of chest ct and rt-pcr testing in coronavirus disease 2019 (covid-19) in china: a report of 1014 cases covid-19 epidemic in switzerland: on the importance of testing, contact tracing and isolation effect of changing case definitions for covid-19 on the epidemic curve and transmission parameters in mainland china: a modelling study characteristics of and important lessons from the coronavirus disease 2019 (covid-19) outbreak in china: summary of a report of 72 314 cases from the chinese center for disease control and prevention defining the epidemiology of covid-19 -studies needed covid-19) -testing in the u.s control prev the effect of control strategies to reduce social mixing on outcomes of the covid-19 epidemic in wuhan, china: a modelling study association of public health interventions with the epidemiology of the covid-19 outbreak in wuhan, china interventions to mitigate early spread of sars-cov-2 in singapore: a modelling study quarantine alone or in combination with other public health measures to control covid-19: a rapid review impact of school closures for covid-19 on the us health-care workforce and net mortality: a modelling study tackling covid-19: are the costs worth the benefits? update on business conditions in france at the end of gdp and spending -gross domestic product (gdp) -oecd data key: cord-335252-med3c01q authors: fovet, thomas; lancelevee, camille; eck, marion; scouflaire, tatiana; becache, eve; dandelot, dominique; giravalli, pascale; guillard, alexandre; horrach, pierre; lacambre, mathieu; lefebvre, tiphaine; moncany, anne-hélène; touitou, david; david, michel; thomas, pierre title: prisons confinées: quelles conséquences pour les soins psychiatriques et la santé mentale des personnes détenues en france? date: 2020-05-08 journal: encephale doi: 10.1016/j.encep.2020.05.002 sha: doc_id: 335252 cord_uid: med3c01q résumé objectif. en france, les mesures de confinement ont été accompagnées de dispositions spécifiques pour les prisons: suspension des activités, parloirs et interventions extérieures. plus de dix mille personnes détenues ont en outre été libérées pour diminuer le taux d’occupation des établissements et limiter la propagation du virus. l’objectif de cet article est de décrire la réorganisation des soins psychiatriques en milieu pénitentiaire en contexte de pandémie de covid-19 et d’interroger les conséquences du confinement et des libérations anticipées sur la santé mentale des personnes détenues. méthode. ce travail s’appuie sur une enquête menée en avril 2020 auprès des soignants de 42 unités sanitaires en milieu pénitentiaire et des 9 unités hospitalières spécialement aménagées en france. une synthèse de la littérature internationale sur la question des soins psychiatriques en milieu pénitentiaire durant l’épidémie de covid-19 a également été réalisée. résultats. l’épidémie de covid-19 semble avoir été plutôt contenue dans les prisons françaises au cours de la période de confinement mais le poids des mesures mises en place sur la population carcérale est important. les 3 niveaux de soins psychiatriques en milieu pénitentiaire ont instauré des mesures spécifiques pour assurer la continuité des soins, accompagner les personnes incarcérées et contenir l’épidémie. parmi les plus importantes, on note la restriction des consultations, la création de « secteurs covid », la déprogrammation des hospitalisations non urgentes, le renforcement des mesures d’hygiène et le remaniement des effectifs. actuellement, les soignants sont principalement confrontés à des sevrages forcés, des symptomatologies anxieuses et des décompensations de troubles psychiatriques chroniques. certaines libérations anticipées sont aussi très préoccupantes, pouvant entraîner des ruptures de soins, par manque de préparation des relais de prise en charge. discussion. les remaniements en lien avec le confinement donnent une visibilité accrue au fossé qui sépare la psychiatrie en milieu libre de la psychiatrie en milieu pénitentiaire. il nous apparaît important de rappeler la vulnérabilité des personnes incarcérées qui doivent impérativement être considérées dans les politiques de santé publique. abstract objective. the impact of the covid-19 pandemic on the 11 million people currently incarcerated worldwide is the subject of many concerns. prisons and jails are filled with people suffering from many preexisting medical conditions increasing the risk of complications. detainees’ access to medical services is already limited and overcrowding poses a threat of massive contagion. beyond the health impact of the crisis, the tightening of prison conditions worries. on march 16, 2020, in france, the lockdown measures have been accompanied by specific provisions for prisons: all facilities have suspended visitations, group activities and external interventions. over 10 000 prisoners have been released to reduce the prison population and the risk of virus propagation. these adjustments had major consequences on the healthcare system in french prisons. the objectives of this article are to describe the reorganization of the three levels of psychiatric care for inmates in france in the context of covid-19 pandemic and to have a look at the impact of lockdown measures and early releases on mental health of prisoners. methods. this work is based on a survey conducted in april 2020 in france among psychiatric healthcare providers working in 42 ambulatory units for inmates and in the 9 full-time inpatient psychiatric wards exclusively for inmates called “uhsas” (which stands for “unités hospitalières spécialement aménagées”, and can be translated as “specially equipped hospital units”). a review of the international literature on mental healthcare system for inmates during the covid-19 epidemic has also been performed. results. the covid-19 epidemic has been rather contained during the period of confinement in french prisons but the impact of confinement measures on the prison population is significant. the three levels of psychiatric care for inmates have implemented specific measures to ensure continuity of care, to support detainees during coronavirus lockdown and to prevent an infection’s spread. among the most important are: limitation of medical consultations to serious and urgent cases, creation of “covid units”, cancellation of voluntary psychiatric hospitalizations, reinforcement of preventive hygiene measures and reshuffling of medical staff. prolonged confinement has consequences on mental health of detainees. currently, mental health workers are facing multiple clinical situations such as forced drug and substance withdrawal (linked to difficulties in supplying psychoactive substances), symptoms of anxiety (due to concerns for their own and their relatives’ wellbeing) and decompensation among patients with severe psychiatric conditions. early releases from prison may also raise some issues. people recently released from prison are identified as at high risk of death by suicide and drug overdose. the lack of time to provide the necessary link between health services within prisons and health structures outside, could have serious consequences, emphasizing the well-known “revolving prison doors” effect. discussion. the current lockdown measures applied in french jails and prisons point out the disparities between psychiatric care for inmates and psychiatric care for general population. giving the high vulnerability of prison population, public health authorities should pay more attention to health care in prisons. objectif. en france, les mesures de confinement ont été accompagnées de dispositions spécifiques pour les prisons : suspension des activités, parloirs et interventions extérieures. plus de dix mille personnes détenues ont en outre été libérées pour diminuer le taux d'occupation des établissements et limiter la propagation du virus. l'objectif de cet article est de décrire la réorganisation des soins psychiatriques en milieu pénitentiaire en contexte de pandémie de covid-19 et d'interroger les conséquences du confinement et des libérations anticipées sur la santé mentale des personnes détenues. méthode. ce travail s'appuie sur une enquête menée en avril 2020 auprès des soignants de 42 unités sanitaires en milieu pénitentiaire et des 9 unités hospitalières spécialement aménagées en france. une synthèse de la littérature internationale sur la question des soins psychiatriques en milieu pénitentiaire durant l'épidémie de covid-19 a également été réalisée. résultats. l'épidémie de covid-19 semble avoir été plutôt contenue dans les prisons françaises au cours de la période de confinement mais le poids des mesures mises en place sur la population carcérale est important. les 3 niveaux de soins psychiatriques en milieu pénitentiaire ont instauré des mesures spécifiques pour assurer la continuité des soins, accompagner les personnes incarcérées et contenir l'épidémie. parmi les plus importantes, on note la restriction des consultations, la création de « secteurs covid », la déprogrammation des hospitalisations non urgentes, le renforcement des mesures d'hygiène et le remaniement des effectifs. actuellement, les soignants sont principalement confrontés à des sevrages forcés, des symptomatologies anxieuses et des décompensations de troubles psychiatriques chroniques. certaines libérations anticipées sont aussi très préoccupantes, pouvant entraîner des ruptures de soins, par manque de préparation des relais de prise en charge. discussion. les remaniements en lien avec le confinement donnent une visibilité accrue au fossé qui sépare la psychiatrie en milieu libre de la psychiatrie en milieu pénitentiaire. il nous apparaît important de rappeler la vulnérabilité des personnes incarcérées qui doivent impérativement être considérées dans les politiques de santé publique. mots-clés : coronavirus, covid-19, sars-cov-2, épidémie, pandémie, psychiatrie, milieu pénitentiaire, confinement abstract objective. the impact of the covid-19 pandemic on the 11 million people currently incarcerated worldwide is the subject of many concerns. prisons and jails are filled with people suffering from many preexisting medical conditions increasing the risk of complications. detainees' access to medical services is already limited and overcrowding poses a threat of massive contagion. beyond the health impact of the crisis, the tightening of prison conditions worries. on march 16, 2020, in france, the lockdown measures have been accompanied by specific provisions for prisons: all facilities have suspended visitations, group activities and external interventions. over 10 000 prisoners have been released to reduce the prison population and the risk of virus propagation. these adjustments had major consequences on the healthcare system in french prisons. the objectives of this article are to describe the reorganization of the three levels of psychiatric care for inmates in france in the context of covid-19 pandemic and to have a look at the impact of lockdown measures and early releases on mental health of prisoners. methods. this work is based on a survey conducted in april 2020 in france among psychiatric healthcare providers working in 42 ambulatory units for inmates and in the 9 fulltime inpatient psychiatric wards exclusively for inmates called "uhsas" (which stands for "unités hospitalières spécialement aménagées", and can be translated as "specially equipped hospital units"). a review of the international literature on mental healthcare system for inmates during the covid-19 epidemic has also been performed. results. the covid-19 epidemic has been rather contained during the period of confinement in french prisons but the impact of confinement measures on the prison population is significant. the three levels of psychiatric care for inmates have implemented specific measures to ensure continuity of care, to support detainees during coronavirus lockdown and to prevent an infection's spread. among the most important are: limitation of medical consultations to serious and urgent cases, creation of "covid units", cancellation of voluntary psychiatric hospitalizations, reinforcement of preventive hygiene measures and reshuffling of medical staff. prolonged confinement has consequences on mental health of detainees. currently, mental health workers are facing multiple clinical situations such as forced drug and substance withdrawal (linked to difficulties in supplying psychoactive substances), symptoms of anxiety (due to concerns for their own and their relatives' wellbeing) and decompensation among patients with severe psychiatric conditions. early releases from prison may also raise some issues. people recently released from prison are identified as at high risk of death by suicide and drug overdose. the lack of time to provide the necessary link between health services within prisons and health structures outside, could have serious consequences, emphasizing the well-known "revolving prison doors" effect. discussion. the current comme dans tous les lieux d'enfermement, la contagion est une menace constante en prison et les exemples historiques ne manquent pas pour illustrer ce constat. dès le xviii e siècle, john howard dénonce les conditions d'incarcération dans les prisons anglaises ainsi que les conséquences dramatiques des épidémies de typhus (connu alors sous le nom de « fièvre des prisons ») [1] . la pandémie grippale de 1918 dite « grippe espagnole » aurait quant à elle touché environ un quart de la population carcérale, une prévalence bien plus importante qu'en population générale [2] . plus récemment, des épidémies de grippes ont été rapportées dans des établissements pénitentiaires de plusieurs pays [3, 4] . par ailleurs, ces observations épidémiologiques ne s'accompagnent pas toujours des mesures de prévention et de prise en charge adaptées, comme en témoignent les faibles taux de vaccination de la population carcérale au cours de la pandémie de grippe h1n1 [5] . l'impact de la pandémie de covid-19 sur les onze millions de personnes actuellement incarcérées à travers le monde [6] fait donc l'objet de nombreuses inquiétudes et interrogations [7] [8] [9] [10] [11] [12] [13] [14] [15] . la population carcérale apparaît particulièrement fragile et potentiellement plus exposée aux formes sévères de la maladie. en effet, parmi les personnes détenues, la prévalence des maladies chroniques associées à une immunodépression est élevée [16, 17] et un vieillissement de cette population est observé dans de nombreux pays (3 043 personnes sont âgées de plus de 60 ans dans les prisons françaises au 1 er janvier 2018) [18, 19] . de plus, plusieurs facteurs comme la surpopulation ou le cadre de fonctionnement sécuritaire, peuvent constituer des freins à un accès aux soins de qualité en milieu pénitentiaire [9] . de nombreuses recommandations ont émergé ces derniers mois des organismes internationaux et des sociétés savantes pour limiter la propagation de l'infection à sars-cov-2 en population générale. toutefois, les conditions d'incarcération actuelles interrogent quant à la possibilité de mettre en place, en milieu carcéral, l'ensemble des mesures de distanciation sociale actuellement recommandées [9] . au 1 er janvier 2020, 70 650 personnes sont détenues en france pour 61 080 places opérationnelles. plus de la moitié de ces personnes incarcérées se trouvent dans une structure sur-occupée à plus de 120 % et l'administration pénitentiaire dénombre 1 614 matelas au sol [20] . c'est dans ce contexte que l'entrée en vigueur des mesures générales de confinement annoncées le 16 mars 2020 a été accompagnée de dispositions spécifiques pour les prisons comme la suspension de toutes les activités considérées comme non essentielles (travail, formation, culte, etc.), la limitation des mouvements et la suppression des parloirs et interventions extérieures. le rôle décisif que joue le phénomène de surpopulation dans la transmission des infections en milieu pénitentiaire a très rapidement conduit de nombreux auteurs à proposer la libération massive des personnes incarcérées dans le contexte de la pandémie de covid-19 [8, 11, 14, 15] . une revue de la littérature récente a en effet mis en évidence une association entre la transmission des maladies infectieuses et la surface disponible par personne détenue dans les cellules [21] . ainsi, plus de 85 000 prisonniers iraniens ont été libérés et cette politique de « décarcéralisation » a été adoptée par de nombreux autres pays [11] . en france, plus de 10 000 personnes détenues (en majorité des personnes qui présentaient un reliquat de peine inférieur à 6 mois) ont été libérées (assignations à domicile, les données épidémiologiques actuellement disponibles sur le nombre de prisonniers atteints du covid-19 sont limitées [13] . aux états-unis, sur les 141 306 prisonniers fédéraux, 1 926 cas de covid-19 (dont 38 personnes décédées) ont été confirmés. parmi les 36 000 personnels pénitentiaires, 350 cas sont recensés (recherche effectuée le 4 mai 2020 [22] ). au royaume-uni, au moins 15 personnes incarcérées sont décédées du covid-19 [13] . en chine, l'impact de l'épidémie en détention aurait été largement minimisé selon certains auteurs [10] . en france, les données communiquées par la direction [23] . en france, des dispositions ont ainsi été mises en place le 23 mars 2020 pour accompagner les restrictions liées au confinement (crédit téléphonique, gratuité de la télévision, etc.), mais elles ont été jugées insuffisantes par la contrôleure générale des lieux de privation de liberté [24] . l'ensemble des aménagements pris pour limiter l'impact de l'épidémie de covid-19 a eu des conséquences majeures sur le système de soins psychiatriques en milieu pénitentiaire. celui-ci doit pourtant impérativement continuer à effectuer ses missions compte tenu de la prévalence élevée des troubles psychiatriques en détention [25, 26] mais aussi des conséquences potentielles des mesures de confinement sur la population carcérale [27] . cet article se propose de décrire la réorganisation des trois niveaux de soins psychiatriques en milieu pénitentiaire au cours de la pandémie de covid-19 et en france, les établissements autorisés en psychiatrie ont très rapidement créé des unités permettant de prendre en charge les patients souffrant de troubles psychiatriques et du covid-19 [28] . toutefois, une nette reprise de l'activité est décrite depuis mi-avril 2020. en ce qui concerne l'admission des personnes détenues sur les secteurs de psychiatrie générale en soins psychiatriques sur décision d'un représentant de l'état, aucune donnée nationale n'est disponible et l'hétérogénéité des pratiques ne permet pas d'établir un état des lieux. le poids du confinement en détention il est difficile de prédire quelles seront les conséquences du confinement en population carcérale. les rares données actuellement disponibles en population générale font état de phénomènes fréquents de peur de la contamination, d'inquiétude pour les proches, d'irritabilité ou de sentiments de frustration et d'impuissance [41] . pour limiter ces réactions, de nombreuses recommandations insistent sur la nécessité de planifier des activités, de pratiquer des exercices physiques, de maintenir des liens sociaux ou d'entretenir une hygiène de sommeil, par exemple. l'accent est mis sur l'utilisation de réseaux sociaux, de sites internet ou d'applications mobiles pour un accès aux pratiques de relaxation et méditation [27] . là encore, les mesures proposées apparaissent bien éloignées de la réalité du milieu carcéral et quasiment impossible à mettre en oeuvre pour les personnes incarcérées. les mesures de confinement actuelles exacerbent des difficultés bien connues en milieu pénitentiaire comme l'isolement ou l'inactivité contrainte. elles entraînent l'inquiétude des personnes détenues sur leur santé ou celle de leurs proches. des difficultés matérielles (problème d'approvisionnement en linge propre suite à la suspension des parloirs ou difficultés financières dues à l'arrêt du travail en détention par exemple) sont aussi rapportées. enfin, les reports d'audience et la en ce qui concerne le suicide, qui constitue une préoccupation majeure en milieu pénitentiaire [50] , les chiffres semblent stables (la disp de lille a par exemple enregistré 2 suicides entre le 16 [15] . on sait à quel point la période suivant la libération est associée à une mortalité élevée, principalement en raison du suicide et des overdoses [51] . ce risque est d'autant plus important que la personne a connu, au cours de sa détention, une période d'isolement [52] . la sortie de détention est aussi identifiée comme un risque majeur de rupture de prise en charge si elle n'a pu être anticipée [53] . « la santé en prison, c'est de la santé publique » expliquent les épidémiologistes, insistant sur l'importance d'intégrer les mesures visant à limiter la diffusion du covid-19 en détention à la réponse globale de santé publique [9] . mais cette formule s'applique également à la santé mentale et devrait nous interroger sur la place donnée actuellement à la psychiatrie en milieu pénitentiaire en france. car si les prisons françaises restent, pour le moment, peu impactées par la maladie, les remaniements en lien avec le confinement donnent une visibilité accrue au fossé qui sépare encore et toujours la psychiatrie en milieu libre et la psychiatrie en milieu pénitentiaire [57] , bien illustré par les difficultés rencontrées pour le relais des prises en charge à la libération. les multiples recommandations relatives à la santé mentale publiées actuellement témoignent d'une préoccupation forte pour les conséquences psychiques du confinement. il nous apparaît particulièrement important de rappeler la vulnérabilité de certaines populations dont font partie les personnes incarcérées qui doivent être largement prises en compte dans les politiques de santé publique. ainsi, le déconfinement annoncé le 28 avril 2020 et qui reste la source de multiples interrogations, devra s'accompagner d'un suivi rigoureux des indicateurs de santé mentale en détention mais surtout d'une réflexion globale sur l'organisation des soins psychiatriques en france intégrant les soins aux personnes détenues. les auteurs déclarent ne pas avoir de liens d'intérêts. the state of the prisons in england and wales | work by howard an analysis of influenza outbreaks in institutions and enclosed societies influenza outbreak in a correctional facility influenza outbreak in a canadian correctional facility distribution of a(h1n1)pdm09 influenza vaccine: need for greater consideration of smaller jails global prison trends 2020 fighting covid-19 outbreaks in prisons flattening the curve for incarcerated populations -covid-19 in jails and prisons prisons and custodial settings are part of a comprehensive response to covid-19 caring for persons in detention suffering with mental illness during the covid-19 outbreak covid-19, prison crowding, and release policies world health organization. preparedness, prevention and control of covid-19 in prisons and other places of detention prisons are "in no way equipped" to deal with covid-19 preventing major outbreaks of covid-19 in jails covid-19 in prisons and jails in the united states the health of prisoners global burden of hiv, viral hepatitis, and tuberculosis in prisoners and detainees understanding and improving the health of people who experience incarceration: an overview and synthesis direction de l'administration pénitentiaire. statistique de la population détenue et écrouée pour l'année 2020 prison cell spatial density and infectious and communicable diseases: a systematic review bop: covid-19 update n advice of the subcommittee on prevention of torture to states parties and national preventive mechanismsrelatingto the coronavirus pandemic covid-19 en prison : des mesures gouvernementales insuffisantes prevalence of mental disorders in french prisons for men mental disorders on admission to jail: a study of prevalence and a comparison with a community sample in the north of france conséquences psychopathologiques du confinement assurer les soins aux patients souffrant de troubles psychiques en france pendant l'épidémie à sars-cov-2. l'encephale mental health and the covid-19 pandemic patients with mental health disorders in the covid-19 epidemic a nationwide survey of psychological distress among chinese people in the covid-19 epidemic: implications and policy recommendations psychological crisis intervention during the outbreak period of new coronavirus pneumonia from experience in shanghai mental health care for medical staff in china during the covid-19 outbreak the risk and prevention of novel coronavirus pneumonia infections among inpatients in psychiatric hospitals france's forensic psychiatry provision: the long and winding road utilization of hospital-level mental health care services for inmates in france: a transversal study entre détention et psychiatrie, les unités hospitalières spécialement aménagées (uhsa) pour prendre en charge les personnes détenues dont l'état psychique relève d'une hospitalisation general medical care in french psychiatric units for inmates: a national survey immediate psychological responses and associated factors during the initial stage of the 2019 coronavirus disease (covid-19) epidemic among the general population in china public health and solitary confinement in the united states solitary confinement and mental illness in u.s. prisons: a challenge for medical ethics psychological distress in solitary confinement: symptoms, severity, and prevalence in the united states psychiatric effects of solitary confinement reforming punishment: psychological limits to the pains of imprisonment psychopathological effects of solitary confinement effects of sensory deprivation in psychiatric seclusion and solitary confinement the role of boredom proneness in self-reported anger and aggression release from prison -a high risk of death for former inmates solitary confinement placement and post-release mortality risk among formerly incarcerated individuals: a population-based study feuille de route sur trois ans pour la santé des personnes placées sous main de justice substance abuse and dependence in prisoners: a systematic review psychiatric disorders and repeat incarcerations: the revolving prison door psychotic disorders and repeat offending: systematic review and meta-analysis psychiatrie en milieu pénitentiaire : une sémiologie à part ? key: cord-000425-isw6jeir authors: flori, laurence; gao, yu; laloë, denis; lemonnier, gaëtan; leplat, jean-jacques; teillaud, angélique; cossalter, anne-marie; laffitte, joëlle; pinton, philippe; de vaureix, christiane; bouffaud, marcel; mercat, marie-josé; lefèvre, françois; oswald, isabelle p.; bidanel, jean-pierre; rogel-gaillard, claire title: immunity traits in pigs: substantial genetic variation and limited covariation date: 2011-07-29 journal: plos one doi: 10.1371/journal.pone.0022717 sha: doc_id: 425 cord_uid: isw6jeir background: increasing robustness via improvement of resistance to pathogens is a major selection objective in livestock breeding. as resistance traits are difficult or impossible to measure directly, potential indirect criteria are measures of immune traits (its). our underlying hypothesis is that levels of its with no focus on specific pathogens define an individual's immunocompetence and thus predict response to pathogens in general. since variation in its depends on genetic, environmental and probably epigenetic factors, our aim was to estimate the relative importance of genetics. in this report, we present a large genetic survey of innate and adaptive its in pig families bred in the same environment. methodology/principal findings: fifty four its were studied on 443 large white pigs vaccinated against mycoplasma hyopneumoniae and analyzed by combining a principal component analysis (pca) and genetic parameter estimation. its include specific and non specific antibodies, seric inflammatory proteins, cell subsets by hemogram and flow cytometry, ex vivo production of cytokines (ifnα, tnfα, il6, il8, il12, ifnγ, il2, il4, il10), phagocytosis and lymphocyte proliferation. while six its had heritabilities that were weak or not significantly different from zero, 18 and 30 its had moderate (0.1<h2≤0.4) or high (h2>0.4) heritability values, respectively. phenotypic and genetic correlations between its were weak except for a few traits that mostly include cell subsets. pca revealed no cluster of innate or adaptive its. conclusions/significance: our results demonstrate that variation in many innate and adaptive its is genetically controlled in swine, as already reported for a smaller number of traits by other laboratories. a limited redundancy of the traits was also observed confirming the high degree of complementarity between innate and adaptive its. our data provide a genetic framework for choosing its to be included as selection criteria in multitrait selection programmes that aim to improve both production and health traits. increasing robustness by improving resistance/tolerance to pathogens is an important selection objective in most livestock species, particularly in pigs. in the past 30 years, selection for growth, carcass leanness, meat quality and prolificacy, combined with stringent sanitary rules, vaccination and use of antibiotics, has been highly effective in pigs [1] . since the early 2000's, prophylactic use of antibiotics as growth promoters has been forbidden by european legislation. as a result, the health status of numerous farms has deteriorated, leading to an increase in the therapeutic use of antibiotics. indeed, animals highly selected for production traits may be more susceptible to pathogens or less able to maintain performance after infection. deterioration of the global health status may also be due to environmental trends. in this context, including health traits in existing breeding schemes using direct and/or indirect strategies is an emerging trend in pig breeding. direct strategies target animal resistance/tolerance to specific pathogens but may result in increased susceptibility to other diseases [2, 3] . alternatively, an indirect and putatively more global approach focuses on immune traits (its) providing a measure of immune capacity (i.e. immunocompetence) and hopefully predicting the responses to pathogens in general [4] . the choice of relevant its is further based on knowledge of the immune system. this highly interactive and cooperative system is classically separated into two arms referred to as innate and adaptive, which produce a combined response. innate immunity is the first line of defence. its activation is non pathogen-specific and depends on the recognition of evolutionarily conserved pathogenassociated molecular patterns such as lipopolysaccharides constituting bacterial cell walls [5] . innate immunity involves physical barriers, innate immune cells such as dendritic cells (dcs), monocytes, natural killers (nk cells) or cd t lymphocytes, and inflammatory cytokines such as il1b, il6 and tnf. adaptive immunity is antigen-specific and requires the recognition of specific ''non-self'' antigens via a process of antigen presentation and results in an immunological memory. adaptive immunity is divided into cell-and humoral-mediated immunity with different effector functions [6] . in order to include its in a breeding plan to improve pig immunocompetence, the genetic and phenotypic parameters of the different its need first to be estimated. several studies in swine, mice, poultry and cattle demonstrated the possibility of selecting animals with high or low immune response (ir) as characterized by one or a few its [2, 7, 8, 9, 10] . a study on yorkshire pigs selected for eight generations for high and low adaptive ir (hir and lir, respectively) on an index combining four standardized measures of specific antibodies and cellmediated ir, after stimulation with specific antigens (bacillus calmette-guérin and hen egg white lysozyme), has revealed that hir and lir animals differ in response to immunization and infection [2, 11, 12, 13, 14] . other studies have also shown that various innate and adaptive its are genetically controlled. for example, variation in innate its, such as nk cells, monocytes, interferon a (ifna) production or phagocytosis [15, 16, 17] is heritable and several adaptive its have moderate to high heritability values including total white blood cells (wbc), cd4 + t lymphocyte, cd8a + t lymphocyte and b lymphocyte subsets [15, 16, 17] , delayed-type hypersensitivity reaction [15, 18] , lymphocyte proliferative response [15] , interleukin-2 (il2) production by lymphocytes [15] and antibody response [12, 15, 18, 19] . clapperton and colleagues have also reported that variation in acute phase protein levels is heritable [16, 17] . finally, several significant qtls for total leukocyte count ( [20, 21] ; animal-qtldb, http://www.animalgenome.org/cgi-bin/qtldb/index), mitogen-induced proliferation [20] , antibody response [20, 22] , cytokine production (il10 and ifnc) [23] , complement activity [22] , and acute phase protein serum concentration [22] have been detected and mapped to different pig chromosomes. taken together these data demonstrate that variation in some its is under genetic control. however, most of the results reported so far have targeted a limited number of traits and very few studies have combined innate and adaptive its. our global goal is to identify immunocompetence traits for inclusion in selection schemes aiming to improve both zootechnical performances and health traits in pigs. for this purpose, we have launched a genetic and genomic study of numerous its covering innate and adaptive ir [24] . in this report, we present the results of a global genetic study, combining principal component analysis (pca), and genetic parameter estimation applied to a large number of innate and adaptive its in a pig population vaccinated against mycoplasma hyopneumoniae (m. hyopneumoniae). a set of 54 its was measured on a population of 443 pigs three weeks after vaccination against m. hyopneumoniae (tables 1 and 2; table s1 ; figure 1 ). these its comprise either traits related to ir (phagocytosis, lymphocyte proliferation, cytokine production after in vitro stimulations, levels of total and specific antibodies, levels of acute phase proteins) or traits related to total leukocyte and leukocyte subpopulation counts. the various characteristics and descriptive statistics of the traits measured on each animal of the studied population (n = 383 to 442) are summarized in tables 1 and 2 . among the cell-mediated its evaluated after diverse stimulations, higher responses in cytokine levels were observed after phorbol myristate acetate (pma)-ionomycin (pmaiono) stimulation compared to lipopolysaccharide (lps) and concanavalin a (cona) stimulations, except for il2 production. conversely, a higher lymphocyte proliferation was detected after cona and pmaiono stimulations than after lps stimulation. ample phenotypic variation was observed for most traits. the coefficient of variation (cv) was equal to 0.8 on average and ranged from 0.07 to 3.9 (tables 1 and 2 ). limited dispersion (cv#0.9) was observed for traits derived from hemograms, cell subsets characterized by fluorescence-activated cell sorting (facs), phagocytosis capacity and non-specific immunoglobulins. moderate dispersion (0.9,cv#1.5) was observed for most cytokines produced in vitro except for tumour necrosis factor a (tnfa) and mitogen proliferation-related traits. finally, the cv for seric c reactive protein (crp) and haptoglobin (hapt) levels were close to 1.6 and 2, respectively. the highest cv (3.9) was obtained for the specific iggs directed against m. hyopneumoniae. these data clearly indicate that the seric inflammatory protein levels and the specific iggs had the greatest phenotypic variance in our study. in order to analyse the factors causing the variation, we performed a normed pca with 32 traits (tables 1 and 2 ). for cellmediated adaptive ir, we included only those traits related to cytokine production and lymphocyte proliferation after pmaionomycin stimulation. for innate its, we included 10 facscharacterized cell subtypes including the percentages of b lymphocytes (igm + ), cd t lymphocytes (tcrcd + ), three subsets of ab t lymphocytes (cd4 + cd8 + , cd4 -cd8 + and cd4 + cd8 -), nk cells (cd16 + cd2 + ) and three monocyte subsets (cd16 + cd172a + , cd16 + mhcii + , mhcii + cd172a + ). we excluded from the analysis four haematological traits not directly involved in immunity: red blood cell count (rbc), hematocrit (ht), red blood cell distribution width (rdw) and platelet count (plt). the percentage of variance (inertia) explained by the first five components was over 50% (figure 2a ). each of these five components explained more than 5% of the total variance and the first two components accounted for 16.4 and 10.8% of the total variance, respectively. taking into account the 32 components from pca, multivariate normal mixture modelling and modelbased clustering (see materials and methods) were used to identify clusters of its. the highest bayesian information criterion (bic) was obtained using the diagonal model with variable shape and variable variance (vvi in pink on figure 2b ) and k = 3 (first factorial plan on figure 2c ). no parameter was located near the correlation circle indicating that the phenotypic correlations between its are globally weak. a first cluster (k1 in blue on figure 2c ) groups together four hemogram-derived cell counts: white blood cell count (wbc), lymphocyte count (lym), monocyte count (mon) and neutrophil count (neu). this cluster is representative of total cell number traits despite the eosinophil count (eos) not being included. a second cluster (k2 in green on figure 2c ) groups together all the traits related to facscharacterized leukocyte subpopulations (expressed as the percentage of cells with one or two surface antigens) except cd t lymphocytes (tcrcd + ), with a cell response parameter (il10-pmaiono) and the seric level of haptoglobin (hapt). this cluster can be considered as representative of the leukocyte subsets. a third cluster (k3 in red on figure 2c ) includes all other cell response traits, one hemogram-derived cell count (eos) and one facs-characterized leukocyte subpopulation. note that k1 and k2 related traits, which mainly correspond to cell subsets and explain around 25% of the phenotypic variance, show clear clustering ( figure 2c ). these traits are grouped on the first pca axis and separated on the second axis. traits belonging to k3, representative of cell activity (cytokine production, phagocytosis and antibody production), are more spread out on the other axes (data not shown). interestingly, cluster analysis did not highlight any cluster of innate or adaptive its. the estimation of phenotypic correlations (r r p ) with wombat confirmed that the its are weakly correlated, except i) among a few cell count traits (wbc, lym, mon, neu), and ii) between cell count traits and a few leukocyte subsets (wbc, lym, cd16 + cd2 + and cd4 -cd8 + ) for whichr r p greater than 0.4 were estimated (table s2; figure s1 ). weakr r p were mainly positive (330) with 166 negative. no strongly negativer r p (#20.4) were found. taken together, pca and estimations of phenotypic correlations showed that the level of redundancy between the different immune parameters was limited. heritability estimates of the 54 analyzed its was equal to 0.45 on average (se = 0.20; table 1 for the large set of adaptive its, the mean heritability was 0.48 (se = 0.21). no significant difference in average heritability values was detected either between group of its qualifying the innate and adaptive immunity or the humoral and cellular adaptive immunity. in addition, an equivalent proportion of innate and adaptive its had significant heritability values. indeed, 40% (10/25), 40% (2/5) and among the traits involved in cell-mediated immunity, variation in ab t lymphocyte (cd4 -cd8 + , cd4 + cd8 + and cd4 + cd8cells) counts was highly heritable. heritability estimates of the three cytokine (il4, il10, ifng) levels were moderate to high after pmaiono and cona stimulations and weak to moderate after lps stimulation. for those cytokines induced by cona or lps stimulation, confidence interval (95ci) for the heritabilities overlapped zero, except for il4-cona. il2 production after pmaiono, cona and lps stimulations gave high estimates of heritability significantly different from zero for il2-pmaiono and il2-lps. proliferation measurements after various stimulations (prolif-cona, prolif-pmaiono, prolif-lps) provided moderate estimates of heritability not significantly different from zero. among the traits involved in humoralmediated adaptive immunity, heritabilities for total igg and iga antibody levels were higher than for total igm and specific antibodies, and weak h 2 values were obtained for b lymphocyte count (igm + cells). heritability for innate its such as i) total cell number (eos and neu), ii) leukocyte subsets (cd16 + cd2 + cells, cd16 + cd172acells, cd16 + mhcii + cells and tcrcd + lymphocytes), iii) cytokine production (ifna and il12), and iv) phagocytosis were high and significantly different from zero. in addition, several innate its showed weak to moderate heritability, including proinflammatory cytokines (il1b, il8, tnf and il6), mon, cd16 -cd2 + cells, cd16 + cd2cells, mhcii -cd172a + cells, mhcii + cd172acells, cd16 -cd172a + cells and cd16 + cd172a + cells. variation in acute phase proteins was moderately (crp) to highly (hapt) heritable. among the four traits, which measured the total number (mon) or proportions (mhcii + cd172a + , cd16 + cd172a + and cd16 + mhcii + ) of monocytes, moderate to high h 2 , but not significantly different from zero, were estimated, except for cd16 + mhcii + cells. other haematological traits (rbc, ht, rdw and plt) gave high h 2 estimates, of which ht and plt were significant. pairwise genetic correlations are presented in table s2 and illustrated in figure 3 . genetic correlation estimates among most its were generally weak but a few high genetic correlations were observed. the number of positive genetic correlations (310) was higher than that of negative correlations (183), as already observed for phenotypic correlations. positive genetic correlations were higher (in absolute values) than negative ones (table s2, figures 3 and s1). only 28 (2.7% of the total number of correlations) and five r ĝ (0.4% of the total number of correlation estimates) were higher than 0.4 or lower than -0.4, respectively. the unsupervised hierarchical clustering distinguished two main clusters of traits ( figure 3 ). the first cluster of 14 traits (cluster a) could be divided into two groups: i) a group of four traits including one innate immunity cytokine (ifna), two antibody levels (total igg and specific igg-mh), and one facscharacterized leukocyte subpopulation (cd16 + cd172a -) and ii) a group of 10 traits with nine hemogram-based cell counts or facscharacterized leukocyte subpopulations (wbc, lym, mon, neu, eos, cd16 + mhcii + , cd16 + cd2 + , cd4 -cd8 + cells) and two cell activity traits (igm and il10-pmaiono). the second cluster of 18 traits (cluster b) is also subdivided into two groups of traits: i) a group of three different cell response traits (il6, iga and ifng-pmaiono) and one facs-characterized leukocyte subpopulation (igm + cells), and ii) a group of nine cell response traits (il12, il4-pmaiono, il2-pmaiono, tnf, prolif-pma, hapt, crp, il1b, il8 and phag) and four facs-characterized leukocyte subpopulations (cd4 + cd8 -, cd4 + cd8 -, tcrcd + and mhcii + cd172a + cells). in cluster a, the first group of 10 traits showed moderately to highly positive genetic correlations with each other. indeed, r ĝ values greater than 0.4 were estimated between wbc, lym, mon, neu, eos, cd4 -cd8 + and cd16 + cd2 + (nk) cells ( figure 3 , table s2 ). in cluster b, r ĝ values greater than 0.4 were estimated between i) tnf, il8 and phag, and ii) crp and hapt ( figure 3 , table s2 ). strong negative r ĝ values (,20.4) were found between a few traits from both clusters: i) tnf and three cell number traits (wbc, lym, mon), ii) il6 and cd16 + mhcii + , and iii) crp and iga. the measured its globally cover innate and adaptive immunity the large-scale study reported here allowed us to estimate the genetic and phenotypic parameters of numerous its measured on pigs bred in the same environment. innate immunity is represented by 25 traits, humoral-mediated immunity by five traits and cell-mediated adaptive immunity by 18 traits. we have also considered the total number of white blood cells and lymphocytes, and four other haematological traits (rbc, ht, rdw and plt). within cell-mediated immunity, we explored both th1 and th2 responses by measuring cytokine production. figure 1 summarizes the traits that we selected to cover immunity globally. these traits include in vivo measures on blood such as quantification of cell populations by hemogram, dosage of circulating immunoglobulins and acute phase proteins, as well as ex vivo measures obtained after in vitro tests such as lymphocyte proliferation, phagocytosis capacity and cytokine production after blood stimulation. all these its have been widely studied in humans [25, 26] . the trait typology we have used (table 1 ) follows a model based on a clear distinction between innate and adaptive immunity, which may be over-simplistic since both immune systems are closely interconnected [27, 28] . monocytes are involved in innate immunity but are also antigen-presenting cells required for adaptive immunity, and cytokines such as il12 are at the interface between innate and adaptive immunity. similarly, the cell-mediated and humoral adaptive immunity subdivision is artificial. for example, il4 is a cytokine produced by th2 lymphocytes that is usually classified as part of adaptive cellmediated immunity, whereas it is also involved in antibody production and thus adaptive humoral immunity. in addition, the conventional paradigm that cd4 + ab t lymphocytes differentiate into th1 and th2 lineages expressing specific cytokines is collapsing. indeed, recent studies have revealed that cytokine production by the different cd4 + t cell subsets (th1, th2, th, th17 and itreg) is highly flexible, providing new insight into the th cell plasticity [29] . nevertheless, although schematic, the approach used in our report provides a comprehensive overview of genetic variation and co-variation across the entire immune spectrum in pigs (figure 1 ). table 1 illustrates various ranges of phenotypic variation in the measured immune traits, with most having a cv over 0.5. such variability has already been reported in large panels of healthy humans [26] , and in previous studies on pigs [2, 16, 17, 30] . for instance, we substantiated the high level of variation of cytokine production previously reported for il2 production and virusinduced ifna production in a swedish yorkshire population [20] . for innate immunity-related cytokines and il12, stimulation was performed with a mixture of lps, pma and ionomycin. il8 was the cytokine produced with the highest levels followed by il1b, tnf, il12 and lastly il6. these four cytokines are not expected to be expressed at similar levels at all time points after stimulation and a kinetic study would help to improve comparison of cytokine production levels. weaker levels of adaptive ir cytokines are observed after lps stimulation compared to cona or pmaiono. these differences could be related to the distinct modes of action of these molecules. indeed, pma, a plant-derived functional analog of diacylglycerol, in conjunction with ionomycin, a calcium ionophore produced by streptomyces conglobatus, and cona, a lectin originally extracted from the jack-bean canavalia ensiformis, are known to be potent mitogens of blood lymphocytes [31, 32] . lps, a major structural component of the outer membrane of gram-negative bacteria, which binds the cd14/ the five first components, which explain more than 50% of the total variance, are in red. b. plot of the bayesian information criterion (bic) calculated with different models according to number of clusters. six models are compared: eii (spherical with equal volume and equal shape), vii (spherical with variable volume and equal shape), eei (diagonal with equal shape and equal volume), vei (diagonal with variable shape and equal volume), evi (diagonal with equal shape and variable volume), vvi (diagonal with variable shape and variable volume). c. first factorial plan (1: first component, 2: second component) with three clusters identified by multivariate normal mixture modelling and model-based clustering taking into account the 32 components (clusters k1, k2 and k3 are in blue, green and red, respectively). doi:10.1371/journal.pone.0022717.g002 tlr4/md2 receptor complex and promotes the secretion of proinflammatory cytokines, has been extensively used to study innate immune response [33] . we have already shown that transcriptome modifications in peripheral blood mononuclear cells (pbmcs) differ between pmaiono and lps stimulation and that pmaiono and lps target different cells and cellular pathways [34] . the combination of pma and ionomycin induces a stronger stimulation that may be related to a higher production of cytokines as detected in the present study. in addition, the lymphocyte proliferation induced by lps is weaker than that observed with other stimulants, as expected. our study provides the first heritability estimates for innate and adaptive cytokine production and for lymphocyte proliferation after pma-ionomycin and lps stimulations in pig. pro-inflammatory cytokines appear to show less heritable variation than adaptive system-related cytokines. among the adaptive system-related cytokines, estimated heritability was weakest for cytokines produced after lps stimulation, except for il2 production. heritability estimates for lymphocyte proliferation after cona, pma and lps stimulations were moderate and that of lymphocyte proliferation after cona stimulation was comparable to the value obtained by edfors-lilja and colleagues [15] . moderate to high heritability estimates for cell count traits from hemogram or facs also confirmed those previously obtained for wbc, total lymphocytes, neutrophils, eosinophils and some leukocyte subsets (for cd4 + and cd8 + t lymphocytes, cd t lymphocytes, cd11r1 + , cd11r1 + cd8a -, cd11r1 + cd8a + and cd16 + mhcii + ) [15, 16, 17, 19] . likewise, our results confirmed the high heritability estimate for phagocytosis [15] . conversly, a lower heritability estimate for crp than previously reported was observed [15, 16] . overall, the heritability estimates for these traits appear robust regardless of populations, environments and protocols. some discrepancies exist between our heritability estimates and previous results for humoral-mediated adaptive its and some innate its. indeed, in our study, b lymphocyte levels (igm + cells) are not significantly heritable contrary to other results [16, 17] and specific iggs (igg-mh) have lower heritability estimates (0.12, se = 0.19) than previously reported (range from 0.27 to 0.45) for specific antibodies directed against other antigens [12] . our estimated heritability for total igg is higher (0.92, se = 0.20) than in published reports [3, 15, 18, 19] . in addition, our estimated h 2 for ifna production is moderate to high (0.60, se = 0.23), contrary to previous results (range from 0 to 0.08) [15] , and for haptoglobin (0.55, se = 0.21) is higher than previously published (range from 0.14 to 0.23) [17, 19] . these discrepancies could be due to differences in the pig breeds and in environment factors but also to the absence of common standardised protocols between laboratories. in order to better qualify the phenotypes, protocol standardisation is needed. overall, we show that variation in both innate and adaptive its is under substantial genetic control (figure 1 ; tables 1 and 2) . similar heritability estimates for innate and adaptive its and also between cell number and cell response parameters were observed. further, heritability estimates do not differ consistently between in vivo and ex vivo measures with no apparent bias due to phenotyping methods. these data also suggest candidate its for qtl mapping. indeed, mapping studies have already started for total leukocyte count ( [20, 21] ; animalqtldb, http://www.animalgenome.org/cgi-bin/qtldb/index), mitogen-induced proliferation [20] , antibody response [20, 22] , cytokine production (il2, il10 and ifnc) [23, 35] , complement activity [22] , and acute phase protein serum concentration [22] . compared to the previously limited data on genetic correlations between its in pigs, our study provides a large-scale estimation of phenotypic and genetic correlations among 32 its. pca results and correlation estimations highlight the weak phenotypic and genetic correlations between the different its, except mainly for cell subsets. no cluster of innate and adaptive its is revealed. these results illustrate that many of the its included in our study provide more or less independent potential clues for selecting for improved immunocompetence. such complementarity is expected since innate immunity is in place or ready for activation prior to infection or antigenic stimulation and collaborates with adaptive immunity, which is induced by infection or antigenic stimulation. nevertheless, a few highly positive genetic correlations have been detected between total number of white blood cells and some leukocytes subsets, and between some leukocyte subsets such as total number of lymphocytes, cd4 -cd8 + lymphocytes (which contains ab cd4 -cd8 + lymphocytes and nk), and nk cells (cd16 + cd2 + cells). phagocytosis, production of il8 and tnf, two pro-inflammatory cytokines produced by monocytes and macrophages, were positively correlated with acute inflammatory phase proteins produced by hepatocytes i.e. crp and hapt. a high correlation was also found between crp and hapt. clapperton and colleagues [17] have shown that phenotypic correlations between leukocyte subsets and acute phase proteins are weak (,0.2) and not significantly different from zero in agreement with our results. in contrast to our study, clapperton and collaborators have not detected any significant genetic and phenotypic correlations between different leukocyte subsets except when one subset was nested in another [16] . our results provide a framework for including its in multitrait selection for immunocompetence in pigs. criteria for inclusion should take into account heritability, biological relevance, biological sensitivity and feasibility of measurement [25] . the weak genetic correlations between most its suggest that it will be difficult to choose only a few its to select for immunocompetence, and that a combination of many traits may be required [26] . the moderate to high heritabilities estimated for many traits together with the selection study on pigs carried out by wilkie and colleagues a decade ago support the feasibility of selecting for immunocompetence [2, 13] . chickens have also been successfully divergently selected for carbon clearance (phagocytic activity), high antibody response to newcastle disease virus three weeks after vaccination (adaptive humoral ir) and wing web response to pha (high cell-mediated immune response) for more than twelve generations [9, 36] . in order to include immunocompetence in selection for improved health, a major challenge will be to correlate variation in heritable its in healthy animals with inter-individual variability in response to various pathogens. testing this hypothesis will be a key point for further use of its as indirect selection criteria in multitrait selection to improve resistance to disease. some results on genetic and phenotypic relationships between immunocompetence and susceptibility to specific pathogens have already been reported in the literature for pigs. among pigs selected for eight generations for high (hir) or low (lir) response based on an index of four cell and humoral-mediated immunity traits, an increased specific antibody response and lower polyserositis were observed in the hir pigs compared to the lir pigs after challenge with a novel pathogen, mycoplasma hyorhinis [12, 37] . thus, animals with a high ir level to unrelated challenges, as defined by wilkie and colleagues [12, 37] , have a better response to infection with mycoplasma hyorhinis. however, hir pigs develop more severe arthritis than lir pigs. indeed, the levels of humoral and cell-mediated adaptive its included in the wilkie et al index induce the formation of immune complexes and/or the development of inflammatory responses, central to the pathogenesis of mycoplasma hyorhinis-induced arthritis. other correlation tests between its and response to various infections are needed. however, it is important to remain cautious with high responder animals, which could develop autoimmune pathologies or pathological iummne responses. all the studies on immunocompetence and resistance to disease will have to be completed by estimation of genetic correlations with economically important traits already under selection. negative genetic correlations have been reported between some its (monocytes, cd11r1 + cells) and average daily gain [16] . however a larger correlation study considering a higher number of its and pig performances is needed and is ongoing in our population. in the future, a more sustainable production system may require a compromise with a slight decrease in performance traded off for a gain in animal robustness. more studies are required to better understand the correlations between its and production traits and it is not established which levels of its would be good predictors for resistance to pathogens if any. in conclusion, our results show that variation in many its is under significant genetic control in pigs and these findings may provide insights in other species. moreover, based on heritability and correlation estimations, some of the its that we have studied might be incorporated into selection schemes, provided they are associated with improved global health and do not exhibit strong antagonisms with other economically important traits. our experiment was conducted in accordance with the french national regulations for humane care and use of animals in research. no ethics approval was required for the vaccination and the collection of blood samples under the then current regulations. experiments were performed under the individual license numbers 77-01 assigned to marcel bouffaud who was responsible for experiments in the test farm, and 78-16 assigned to a veterinarian, dr silvia vincent-naulleau. the experimentation agreement number for the test farm at le rheu was a35-240-7. a total of 443 large white pigs (castrated males, dam line) tested for performance traits in a pig test station (ue450, inra, le rheu, france) was included in the study. the pigs were distributed in seven contemporary groups and belonged to 307 nuclear families obtained from 106 boars, with an average of 4.1 (+/2 1.7) piglets per boar. animals were born and weaned in 16 different selection herds and arrived in the test station at five weeks of age with no prior vaccination. they were placed into pens of 30 piglets in a post weaning unit and vaccinated against m. hyopneumoniae (stellamune, pfizer, one injection) one day after their arrival in the test station, when 36.3 days old in average. all pigs were apparently healthy with no clinical sign of infection. all animals were sampled three weeks after vaccination. blood samples were collected via the external jugular vein into tubes with or without anti-coagulants, according to further use. blood collected in 10 ml tubes with no anti-coagulant was centrifuged at 3200 g for 15 min at 4uc. the serum was collected and stored at 220uc until use. plasma was collected from blood sampled in heparinised tubes and stored at 220uc before use. hemograms were measured with an ms4-5 counter (eli-techgroup, france) with blood sampled in edta tubes. among a set of 18 traits, nine were included in the genetic analyses: total number of leukocytes, lymphocytes, monocytes, neutrophils, eosinophils, erythrocytes, platelets and hematocrit (tables 1 and 2 ). total concentrations of immunoglobulin subsets were measured by elisa as previously described [38] . plasma samples were diluted 1:6000, 1:4000 and 1:60,000 to detect igm, iga and igg, respectively, in tris-buffered saline and added to plates coated with immunoglobulin class specific pig antibody (bethyl laboratories inc., interchim, france). the different subsets were detected with the appropriate peroxidase anti-pig igm, iga or igg (bethyl laboratories inc.) and were quantified by reference to standard curves constructed with known amounts of pig immunoglobulin subsets. anti-m. hyopneumoniae igg titers were also measured by elisa using a commercial kit (elisa id screenh m. hyopneumoniae indirect, idvet, france). absorbance was read at 450 nm using an elisa plate reader (spectra thermo, tecan, nc, usa) and the biolise 2.0 data management software. haptoglobin and c reactive protein levels were measured in pig serum by colorimetric tests (phase haptoglobin assay, abcys biologie, france) and elisa assays (porcine c reactive protein assay, abcys biologie, france), respectively. absorbance was read at 450 nm using an elisa plate reader (mrx revelation, dynex). pbmcs were purified by density gradient centrifugation. a volume of 13 ml heparinised blood was added to leucosept tubes (greiner bio-one, france) pre-filled with 17 ml ficoll (lymphocytes separation medium, eurobio, france) and centrifuged at 1200 rpm for 35 min. pbmcs were collected at the ficoll interface and washed in 50 ml d-pbs without mgcl 2 and cacl 2 (gibco, invitrogen, france). cells were then incubated in 2 ml bd pharmlyse 1x (bd biosciences, france) at room temperature. purified pbmcs were washed in 50 ml d-pbs without mgcl 2 and cacl 2 , incubated with 2 ml pig serum at 4uc for 20 min, washed again in 50 ml d-pbs without mgcl 2 and cacl 2 and then washed in 50 ml s/w buffer (1 g/l nan 3 , 10 g/l bovine serum albumin in pbs, ph 7.3) at a final concentration of 5.10 6 cells/ml. 10 6 cells were used for each antibody labelling. single, double or triple staining was performed using monoclonal antibodies (mabs) directed against i) cd2 (msa4, isotype igg2a, vmrd) and cd16 (mca1971, isotype igg1, serotec), ii) cd4 (pt90a, isotype igg2a, vmrd) and cd8a (pt81b, isotype igg2b, vmrd) iii) tcrcd (mac320, isotype igg2a, bd biosciences pharmingen), iv) igm (pig45a, isotype igg2b, vmrd) and v) mhcii (msa3, isotype igg2a, vmrd), cd16 (mca1971, isotype igg1, serotec) and cd172a (74-22-15a, isotype igg2b, bd biosciences pharmingen). briefly, pbmcs were stained with primary mabs for 25 min at 4uc, washed in s/w buffer and stained with allophycocyanin-conjugated anti-mouse igg1 (bd biosciences, france), phycoerythrin-conjugated antimouse igg2a (southern biotech, france), fitc-conjugated antimouse igg2b (southern biotech, france), apc-conjugated antimouse igg1 (bd biosciences, france), phycoerythrin-conjugated anti-rat igg2a (bd pharmingen, france), or phycoerythrinconjugated anti-mouse igg2b (southern biotech, france). after washing in s/w buffer, cells were fixed in bd cellfix solution (becton dickinson, germany). data acquisition and analysis were carried out with the facscan and cellquest software (becton dickinson, uk). synthesis of ifna by leukocytes of pigs was tested in vitro by incubating diluted total blood in the presence of pseudorabies virus (prv, suid herpesvirus 1)-infected, glutaraldehyde-fixed, pk15 cell monolayers, according to a protocol previously described for transmissible gastroenteritis virus [39] . confluent pk15 cell monolayers grown in 24 well plates were infected by prv at a multiplicity of infection of 20, fixed with 0.05% glutaraldehyde 8 h post-infection and washed with d-pbs and rpmi1640 before the addition of blood samples. for each animal, monolayers were incubated with diluted heparinized blood (270 ml of blood diluted 1:5 in dmem supplemented with antibiotics) for 18 h at 37uc. plates were then centrifuged at 450 x g for 20 min at 4uc, and supernatants were collected and stored at -20uc. ifna was assayed in the supernatants using a classical sandwich elisa test as previously described [40] . production and dosage of il1b, il6, il8, tnfa and il12 heparinized blood samples (400 ml) were fivefold diluted in 24well plates in 1.6 ml rpmi 1640 medium (biowhittaker, belgium) supplemented with 10% heat-inactivated fetal bovine serum (qb perbio, uk), 2 mmol/l l-glutamine, 100 u/ml penicillin and 100 mg/ml streptomycin. for stimulation, a mixture of 10 ng/ml pma (sigma, france), 1 mg/ml ionomycin (sigma, france) and 1 mg/ml lps from escherichia coli o111:b4 (sigma, france) was added to the diluted blood. for mock stimulation, a volume of pbs equal to the volume of stimulation reagents was added to the diluted blood. after incubation at 37uc for 24 h, culture supernatants were collected by centrifugation at 450 g for 20 min and stored at 220uc before use. the cytokines il1b, il6, il8, tnf and il12 were quantified using commercial elisa tests (duoset elisa development kits, r&d systems, usa). for quantification of basal levels of cytokines in supernatants from mock-stimulated cells, the samples were not diluted for quantification. supernatants collected from stimulated cells were diluted (1:22 for il1b and il8, 1:1 for il6, 1:10 for tnf and 1:2 for il12). all samples were tested in duplicates. absorbance was read at 450 nm using an elisa plate reader (mrx revelation, dynex). results were expressed as pg of cytokine/ml. heparinized blood diluted 1:5 in complete culture medium consisting of dmem (dulbecco's modified eagle medium, eurobio, france) supplemented with 5% fetal calf serum (hyclone, perbio, france), 2 mm l-glutamine, 100 u/ml penicillin and 50 mg/ml streptomycin (eurobio, france) was stimulated with 10 mg/ml cona (sigma, france), or with 50 ng/ml of pma (sigma, france) and 1 mg/ml of ionomycin (sigma, france) or 1 mg/ml lps from escherichia coli (sigma, france). cytokine content was measured in supernatants using elisa tests as already described [41] . briefly, purified fractions of anti-swine il-2, il-4, ifnc (clones a150d 3f1, a155b 16f2 and a151d 5b8 respectively, biosource, france) and il10 (clone 148801, r and d system, france) were used as capture antibodies, in conjunction with the biotinylated anti-swine il-2, il-4, il10 and ifnc monoclonal antibodies (clones a150d 8h10, a155b 15c6 and a151d 13c5, respectively, biosource, clinisciences, france) or anti-swine polyclonal antibody (goat anti-porcine il-10, r and d system, france). streptavidin-horseradish peroxidase (biosource) and tmb (fermentas, md, usa) were used for detection. absorbance was read at 450 nm using an elisa plate reader (spectra thermo, tecan, nc, usa) and the biolise 2.0 data management software. recombinant pig il-2, il-4, il10 and ifnc were used as standards. the detection limits were 700 pg/ ml, 60 pg/ml, 90 pg/ml and 100 pg/ml for il-2, il-4, il10 and ifnc, respectively. results were expressed as pg of cytokine/ ml. lymphocyte proliferation was performed in 96 well plates as already described [42] . briefly, heparinized blood samples were diluted 1:15 in complete culture medium consisting of dmem (dulbecco's modified eagle medium, eurobio, france) supplemented with 5% fetal calf serum (hyclone, perbio, france), 2 mm l-glutamine, 100 u/ml penicillin and 50 mg/ml streptomycin (eurobio, france). for detection of unspecific lymphocyte proliferation, the diluted blood samples were seeded in 96 well plates (200 ml/well) and mock-stimulated for 48 h by incubation in culture medium (control wells), or stimulated for 48 h by incubation with the culture medium supplemented with either 10 mg/ml cona (sigma, france), or 50 ng/ml of pma (sigma, france) and 1 mg/ml of ionomycin (sigma, france), or 1 mg/ml lps (sigma, france). control wells remained unstimulated. after 48 h of incubation at 39uc, 0.5 mci of 3 h-methylthymidine (icn, france) was added to each well. after another 24 h incubation period, the cells were harvested through glassfiber filters (whatman, united kingdom) by means of an automatic harvester (titerteck-skatron, molecular devices, france). incorporation of tritiated thymidine was measured with a liquid scintillation beta counter (kontron instruments, france). results were expressed as a stimulation index of lymphocyte proliferation calculated as mean counts per min (cpm) of the triplicate cultures in stimulated culture/mean cpm in control non-stimulated culture. preliminary statistical analyses were performed using r software [43] . in order to test if trait distributions deviated from gaussian, a d'agostino normality test was used (p = 0.05). since most traits were not sampled from a gaussian distribution, they were all normalized using a box-cox transformation except for phenotypes reaching the value zero, which were normalized using ln(1+x) transformation. significant effects of age at the time of vaccination, of time of vaccination, of breeding unit and of time of experiment were detected for most traits by variance analysis taking into account these effects. normed principal component analysis (pca, [44] , dudi.pca function, ade4 package [45] , r software [43] ) on a subset of its adjusted for age at the time of vaccination, time of vaccination, breeding unit and experiment (table 1) were performed using a linear model. clusters of its were detected using the r package mclust for normal mixture modelling and model-based clustering [46] . it combines model-based agglomerative hierarchical classification, based on the classification likelihood, and the expectation-maximization (e-m) algorithm for maximum likelihood estimation of multivariate mixture models. variance components, genetic parameters and their standard errors were estimated by the restricted maximum likelihood (reml) method [48] , using the wombat software [49] . this is the reference method to estimate genetic parameters with a mixed model. univariate and bivariate mixed linear animal models were employed to estimate heritability and genetic correlations, respectively. for the univariate analyses, the fixed part of the model included experiment time, age at the time of vaccination, vaccination time and herd of origin effects and the random part included a common litter environmental effect and direct genetic effects. in matrix notation y~x b zw a azw c cze where y = the vector of observations; x b , w a and w c are known incidence matrix relating observations to fixed and random effects; ß = a vector of fixed effects and covariates; a = the vector of direct genetic effects; c = the vector of common litter environmental effects; and e = the vector of random residual effects. all random effects were assumed to follow a normal distribution with zero mean. for bivariate analyses, the same effects as for univariate analyses were taken into account in the fixed part of the model and a direct genetic effect was included in the random part of the model. 95% confidence intervals (95ci) were calculated for heritability (h 2 ) estimates (h 2 ). heritability estimates have been classified: high (h 2 .0.4), moderate (0.1,h 2 ,0.4) or weak (h 2 #0.1). a graphical representation of the genetic correlations combined with a hierarchical clustering (euclidian distance, average link) was obtained with the heatmap function from the bioconductor software [50] . comparisons of heritability average between subsets of traits were tested using the wilcoxon mann-whitney test (significance threshold p-value = 0.05). figure s1 heatmap of the phenotypic correlations between 32 its. the correspondence between colour scale and genetic correlation levels are presented on the right-hand side of the heatmap. (tif) estimation of genetic trends in french large white pigs from 1977 to 1998 for growth and carcass traits using frozen semen selection for high immune response: an alternative approach to animal health maintenance? genetic aspects of health and disease resistance in pigs relationships between genetic change and infectious disease in domestic livestock toll-like receptors and innate immunity kuby immunology toward genetic dissection of high and low antibody responsiveness in biozzi mice genetic parameters for antibody response of chickens to sheep red blood cells based on a selection experiment correlated effects of selection for immunity in white leghorn chicken lines on natural antibodies and specific antibody responses to klh and m. butyricum evaluation of immune responses of cattle as a means to identify high or low responders and use of a human microarray to differentiate gene expression use of estimated breeding values in a selection index to breed yorkshire pigs for high and low immune and innate resistance factors immune responsiveness in swine: eight generations of selection for high and low immune response in yorkshire pigs multi-trait selection for immune response; a possible alternative strategy for enhanced livestock health and productivity cytokines in mycoplasma hyorhinis-induced arthritis in pigs bred selectively for high and low immune responses genetic variation in parameters reflecting immune competence of swine pig peripheral blood mononuclear leucocyte subsets are heritable and genetically correlated with performance traits associated with innate and adaptive immunity in pigs: heritability and associations with performance under different health status conditions an evaluation of immune competence in different swine breeds immunological traits have the potential to improve selection of pigs for resistance to clinical and subclinical disease mapping quantitative trait loci for immune capacity in the pig confirmation of qtl on porcine chromosomes 1 and 8 influencing leukocyte numbers, haematological parameters and leukocyte function qtl for traits related to humoral immune response estimated from data of a porcine f2 resource population mapping quantitative trait loci for cytokines in the pig deciphering the genetic control of innate and adaptive immune responses in pig: a combined genetic and genomic study markers to measure immunomodulation in human nutrition intervention studies immunological parameters: what do they mean? innate immune recognition and control of adaptive immune responses innate immunity: impact on the adaptive immune response mechanisms underlying lineage commitment and plasticity of helper cd4+ t cells the association between plasma levels of acute phase proteins, haptoglobin, alpha-1 acid glycoprotein (agp), pig-map, transthyretin and serum amyloid a (saa) in large white and meishan pigs mitogenic activity of 12-o-tetradecanoyl phorbol-13-acetate on peripheral blood lymphocytes from young and aged adults characterization of the response of human thymocytes and blood lymphocytes to the synergistic mitogenicity of 12-otetradecanoylphorbol-13-acetate (tpa)-ionomycin lps/tlr4 signal transduction pathway transcriptome analysis of porcine pbmcs after in vitro stimulation by lps or pma/ionomycin using an expression array targeting the pig immune response mapping quantitative trait loci for innate immune response in the pig immune modulation: the genetic approach mycoplasma hyorhinis infection of pigs selectively bred for high and low immune response ingestion of deoxynivalenol (don) contaminated feed alters the pig vaccinal immune responses induction of alpha interferon by transmissible gastroenteritis coronavirus: role of transmembrane glycoprotein e1 a sensitive immunoassay for porcine interferon-alpha fumonisin b1 alters cell cycle progression and interleukin-2 synthesis in swine peripheral blood mononuclear cells ingestion of low doses of deoxynivalenol does not affect hematological, biochemical, or immune responses of piglets a language and environment for statistical computing. vienna: r foundation for statistical computing on lines and planes of closest fit to systems of points in space the ade4 package. i. one-table methods model-based methods of classification: using the mclust software in chemometrics the significance of the difference between two means when the population variances are unequal recovery of interblock information when block sizes are unequal wombat: a tool for mixed model analyses in quantitative genetics by restricted maximum likelihood (reml) bioconductor: open software development for computational biology and bioinformatics sincere thanks are due to mathieu gautier (inra, umr gabi, jouy-en-josas, france) for his computational help and to thierry tribout (inra, umr gabi, jouy-en-josas, france) for providing the pig genealogy. we thank fabrice andreoletti and stephan bouet (inra, umr gabi, jouyen-josas, france) for their contribution to animal experimentation. we are also grateful to hélène hayes (inra, umr gabi, jouy-en-josas, france) for helping in preparing the manuscript. we warmly thank christopher moran (faculty of veterinary science, university of sydney) for helpful comments and suggestions for the final submission of the manuscript. key: cord-347244-abxv2mkz authors: izopet, jacques; dubois, martine; bertagnoli, stéphane; lhomme, sébastien; marchandeau, stéphane; boucher, samuel; kamar, nassim; abravanel, florence; guérin, jean-luc title: hepatitis e virus strains in rabbits and evidence of a closely related strain in humans, france date: 2012-08-17 journal: emerg infect dis doi: 10.3201/eid1808.120057 sha: doc_id: 347244 cord_uid: abxv2mkz hepatitis e virus (hev) strains from rabbits indicate that these mammals may be a reservoir for hevs that cause infection in humans. to determine hev prevalence in rabbits and the strains’ genetic characteristics, we tested bile, liver, and additional samples from farmed and wild rabbits in france. we detected hev rna in 7% (14/200) of bile samples from farmed rabbits (in 2009) and in 23% (47/205) of liver samples from wild rabbits (in 2007–2010). full-length genomic sequences indicated that all rabbit strains belonged to the same clade (nucleotide sequences 72.2%–78.2% identical to hev genotypes 1–4). comparison with hev sequences of human strains and reference sequences identified a human strain closely related to rabbit strain hev. we found a 93-nt insertion in the x domain of open reading frame 1 of the human strain and all rabbit hev strains. these findings indicate that the host range of hev in europe is expanding and that zoonotic transmission of hev from rabbits is possible. h epatitis e virus (hev) is a major cause of acute hepatitis in many developing countries in asia and africa, where it is transmitted by the fecal-oral route because of poor sanitation practices (1) . acute hepatitis e is also increasingly reported in industrialized countries, where the transmission is mainly zoonotic (2) . the initial discovery of hev transmission from domestic pigs (3) has been followed by evidence that other mammals, such as wild boars and deer, are also potential reservoirs of hev (4) . although the course of hev infection is generally self-limiting and asymptomatic (or symptomatic with acute hepatitis), fulminant hepatitis can occur in pregnant women and in persons with underlying liver disease (5) (6) (7) . hev infections can also become chronic in immunocompromised patients, such as recipients of solid-organ transplants (8) (9) (10) , those with hematologic diseases (11, 12) , and patients infected with hiv (13) (14) (15) . hev, genus hepevirus, family hepeviridae, is a positive-sense, single-stranded, nonenveloped rna virus (16) . the hev genome is ≈7.2 kb long and contains 3 open reading frames (orfs) as well as 5′ and 3′ untranslated regions: orf1 encodes nonstructural proteins, orf2 encodes the capsid protein, and orf3 encodes a small phosphoprotein. phylogenetic analysis of hev sequences has led to the recognition of 4 major genotypes that infect mammals from a variety of species. hev1 and hev2 are restricted to humans and transmitted through contaminated water in developing countries. hev3 and hev4 infect humans, pigs, and other mammals and are responsible for sporadic cases of hepatitis e in developing and industrialized countries (2) . hev3 is distributed worldwide, whereas hev4 largely is found in asia. although hev3 and hev4 infections have been linked to the consumption of raw or undercooked meats, such as pig liver sausages or game meats (17, 18) , the full spectrum of animals that are reservoirs of hev is still unknown. recent studies have characterized new hev genotypes in isolates from rats in germany (19) , wild boars in japan (20) , and farmed rabbits in the people's republic of china (21, 22) . because the potential risk for zoonotic transmission strain in humans, france of hev from rabbits in france is unknown, and cases of autochthonous hepatitis e are commonly reported in this country (23, 24) , we investigated the prevalence of hev in farmed and wild rabbits. we also looked for a genetic link between hev strains circulating in rabbits and hev strains circulating in humans in france. bile specimens (n = 200) were collected in september 2009 from rabbits raised on 20 farms in western france, in the departments of maine et loire (n = 6), vendée (n = 6), deux-sèvres (n = 4), calvados (n = 2), and loire atlantique (n = 2), the main geographic areas of rabbit farming in france. we sampled 10 rabbits from each farm when they were slaughtered at 70-90 days of age (table 1) . all rabbits were healthy and intended for human consumption. all samples were immediately stored at −80°c. liver specimens (n = 205) were collected during september 2007-november 2010 from 18 populations of wild rabbits, established in warrens; each population was considered epidemiologically independent. the populations were located in several departments of mainland france: dordogne (n = 7), finistère (n = 3), deux-sèvres (n = 2), loire-atlantique (n = 2), haute-garonne (n = 1), charentes (n = 1), morbihan (n = 1), and pyrénées-orientales (n = 1) ( table 1 ). the number of rabbits sampled in a given warren ranged from 1 to 44. they were >6 months of age, apparently healthy, and intended for human consumption. each rabbit was eviscerated within a few hours of its death, and a sample of its liver was taken and immediately frozen at −80°c. necropsies were performed on a group of 12 rabbits from the same warren in haute-garonne (w3), and samples of their intestine and cecum were taken, in addition to samples from the liver. serum specimens were collected from immunocompetent and immunocompromised patients who had received a diagnosis of hepatitis e from the department of virology at toulouse university hospital. all samples were stored at −80°c (23, 24) . samples (140 μl of rabbit bile and 50 mg of liver, intestine, and cecum) were disrupted with trizol (invitrogen, saint aubin, france). rna was extracted with qiaamp viral rna mini kits (qiagen, courtaboeuf, france). we used 1-step real-time reverse transcription pcr on the light cycler 480 instrument (roche diagnostics, meylan, france) to amplify a 70-bp fragment. the primers and probes targeted the orf3 region: forward primer hevorf3-s: 5′-ggtggtttctggggtgac-3′, reverse primer hevorf3-as: 5′aggggttggttggatgaa -3′, and probe 5′-fam-tgattctcagcccttcgc-tamra-3′ (25) . each 50-μl reaction mix contained 1 μl of superscript iii platinum one-step quantitative rt-pcr system (invitrogen), 15 μl of rna, primers (200 nmol/l) and probes (150 nmol/l), and 40 u of rnase out (invitrogen). reverse transcription was carried out at 50°c for 15 min, followed by denaturation at 95°c for 1 min. dna was amplifi ed with 50 pcr cycles at 95°c (20 s) and 58°c (40 s). hev rna was quantifi ed by using a transcribed rna standard constructed from a genotype 3f hev strain (genbank accession no. eu495148). the limit of detection was 100 copies/ml. two fragments, one within orf2 (189 bp) and the other within orf1, encompassing the hypervariable region and x domain (≈1,400 bp), were amplifi ed and sequenced in both directions by the dideoxy chain termination method (prism ready reaction ampli taq fs and dye deoxy primers; applied biosystems, paris, france) on an abi 3130xl capillary dna analyzer (applied biosystems, foster city, ca, usa). the primers used for the orf2 fragment were the following: forward primer hevorf2-s: 5′-gacagaattratttcgtcggctgg-3′ and reverse primer hevorf2-as: 5′-tgytggttrtcataatcc tg-3′. the primers used for the orf1 fragment were the following: forward primer hevorf1-s: 5′-tgacggcyacygtkgarcttg-3′ and reverse primer hevorf1-as: 5′-acatcracatccccctgy tgtatrga-3′. the whole genomes of 2 rabbit strains (w1-11 and w7-57) and 1 human strain (tls-18516-human) were amplifi ed by overlapping rt-pcr. the primers are listed in table 2 . the genotype was determined by using reference strains as previously described (26) . phylogenetic analyses were performed with genotype information on reference sequences based on the hev classifi cation proposed by lu et al. (27) . sequences were aligned by using clustalw (mega5, www.megasoftware.net; bioedit version 7.0, www.mbio.ncsu.edu/bioedit/bioedit). phylogenetic trees were created by the neighbor-joining (kimura 2-parameter) method with a bootstrap of 1,000 replicates. the 2 partial sequences of orf1 and the 5 full-length sequences reported in this study have been deposited in genbank. the accession numbers are jq013789 and jq013790 for orf1, and jq013791 to jq013795 for the full-length sequences of w1-11, w7-57, tls-18516-human, tr19 (genotype 3c), and tr02 (genotype 3e), respectively. all bile specimens from the 200 farmed rabbits and the liver specimens from the 205 wild rabbits were tested for hev rna (table 1) . samples from 7 farms (35%) and 9 warrens (50%) tested positive for hev rna. hev rna was found in a 14 bile samples (7%) from farmed rabbits. the median hev rna concentration in the bile samples was 2.3 × 10 7 copies/ml (range 100 copies/ml-10 9 copies/ml). a total of 47 liver samples (23%) from wild rabbits were positive for hev rna; median hev rna concentration was 1.9 × 10 6 copies/g (range 1,400 copies/g-5.8 × 10 7 copies/g). we tested the liver, intestine, and cecum samples from 12 wild rabbits from the same warren (w3) in triplicate to obtain a clear picture of the tissue distribution of hev in infected rabbits. hev rna was detected in all the tissues from 4 rabbits (nos. 4, 7, 9, 12) , in the liver and intestine of 1 rabbit (no. 5), and in the liver only of 1 rabbit (no. 6) ( table 3) . the virus loads in the liver (mean 4.8 log copies/g), intestine (mean 4.0 log copies/g), and cecum (mean 3.6 log copies/g) were not signifi cantly different. phylogenetic analyses, conducted on the basis of a 189-nt fragment within orf2 of the 37 hev strains from rabbits, hev3 strains from humans circulating in france, and hev reference sequences (hev1, hev2, hev3, hev4, rabbit hev, rat hev, wild-boar hev) indicated that the 37 new orf2 sequences from rabbit hevs were clustered. one cluster contained 3 orf2 sequences from previously characterized hev from farmed rabbits from china, 2 orf2 sequences from hevs from farmed rabbits in france, and 13 orf2 sequences from hevs from wild rabbits in france (figure 1 ). this cluster also contained an orf2 sequence from a strain from a person in france (tls-18516-human) (figure 1 ). this strain was found in a serum sample from a 46-year-old man with an elevated alanine aminotransferase level (400 iu/l, reference <35 iu/l). phylogenetic analysis based on a 1,400-nt fragment within orf1, indicated that the orf1 sequences from hev strains from rabbits in france (n = 4) or china (n = 3) and the orf1 sequence from the human strain tls-18516human formed a distinct genetic group among sequences of hev genotypes 1-4 (data not shown). the cluster of rabbit hev sequences was also distinct from the hev sequences from wild-boar and rat hev genotypes that were characterized recently. comparison of the orf1 sequences from rabbit hev strains with reference orf1 sequences from hev genotypes 1-4 showed an insertion of 93 nt in the x domain of the orf1 of all the rabbit hev strains. this insertion was also found in the tls-18516-human strain. the deduced amino acid sequences corresponding to this insertion, located between amino acids 938 and 939 (burmese strain, m73218), were not very similar, except for 2 conserved amino acids at the c-terminal end. we obtained the full-length genomic sequences of hev strains from 2 wild rabbits in france and the tls-18516-human strain. the phylogenetic tree, constructed by the neighbor-joining method using the full-length genomic sequences (including the sequences of genotypes 3f, 3e, and 3c, which were circulating in france), revealed that the hev genomes from the rabbit strains and the tls-18516-human strain belonged to the same clade. this clade was clearly separated from genotypes 1-4, found in other mammals and from the new hev genotypes found in wild boars and rats (figure 2) . the length of the rabbit strain w1-11 genome, excluding the poly(a) tract at the 3′ terminus, was 7,262 nt. the length of the rabbit strain w7-57 was 7,231 nt, and that of the tls-18516-human strain was 7,259 nt. the nucleotide sequences of the rabbit strains and the tls-18516-human strain were 80.3%-85% identical ( table 4 ). the nucleotide sequences of the rabbit or tls-18516human strains were 76.1% to 78.2% identical to those of genotype 3, 72.7% to 73.7% identical to those of genotype 1, 72.2% to 73.5% identical to those of genotype 2, and 72.9% to 74.9% identical to those of genotype 4. these comparisons therefore indicate that the sequences of the rabbit hev strains and the tls-18516-human strain are distinct from all known strains of hev genotypes 1-4 and from the newly described hev genotypes from wild boars and rats. we found that farmed and wild rabbits in france are naturally infected with hev. we also characterized a human hev strain that is closely related to rabbit hev strains; this fi nding thus supports the potential of zoonotic transmission from rabbits to humans. the hevs found in farmed rabbits in several geographic areas of china have been identifi ed (21, 22) . hev was also recently found in farmed rabbits in virginia, usa (28) . our study results show that rabbits in europe are infected with hev and that some farmed rabbits and wild rabbits in france are infected. we found hev rna in 7% of the farmed rabbits and in 23% of the wild rabbits. however, the ages of the rabbits and the tissues tested (bile samples from farmed rabbits and liver samples from wild rabbits) may explain the observed difference in hev prevalence. nevertheless, previous studies have shown that the virus loads in liver and bile samples from swine infected with hev are similar (29, 30) . although the greater prevalence of hev in wild rabbits could be linked to their older age, we could not test for a relationship between the prevalence of hev and rabbit age because we did not know the rabbits' precise ages. our analysis of the distribution of hev in the tissues of infected wild rabbits showed hev rna not only in the liver, but also in the intestine and cecum; our analysis also showed that the virus loads from these organs were not signifi cantly different. this fi nding suggests that 1278 emerging infectious diseases • www.cdc.gov/eid • vol. 18 extrahepatic sites of hev replication exist in rabbits, as has been demonstrated for hev3 in pigs (31) . however, because the intestine and cecum samples may have been contaminated with blood, our results need to be confi rmed in future studies using methods that ensure that tissues other than the liver are not contaminated with blood. to determine whether rabbits could be a reservoir for viruses that cause human infection, we analyzed partial and complete nucleotide sequences of the rabbit hev strains and compared these sequences with those of human hev strains circulating in france. analysis of orf2 showed that the sequences from rabbit hev strains formed clusters, one of which included the sequences of hev genotypes 2 and 4. the bootstrap values were very low because the fragments analyzed were small. in contrast, phylogenetic analyses based on orf1 and the full-length genome indicated that all the rabbit strains from china and france belong to the same clade. one human strain, tls-18516-human, clustered with the rabbit strains and appeared to be somewhat different from the 4 major hev genotypes found in mammals and the newly described hev genotypes from rats and wild boars. although the full-length sequences of the genomes of the rabbit strains and the tls-18516-human strain are more similar to that of hev3 than to those of hev1, hev2, and hev4, they do not seem to belong to the established hev genotype 3 found in humans and swine, as recently suggested (20, 32) . differences in the classifi cation of rabbit hev could be because the full-length genomic sequences were used as the reference for phylogenetic analyses. genotype 3 is highly diverse, with 10 identifi ed subtypes (27) . we included in our analysis the full-length genomes of subtypes 3f, 3c, and 3e, which account for ≈74%, 13%, and 5% of the human and swine hev strains circulating in france (26, 33) . we also included the other full-length genomes representative of hev3 subtypes, but subtypes 3d, 3h, and 3i are not yet available in genbank. our data indicate that the genomes of rabbit hev strains or tls-18516-human were <80% identical with hev3, regardless of which method was used to align the sequences. this fi nding is compatible with the defi nition of a new genotype, as previously proposed (21, 22) . we found a 93-nt insertion in the x domain of the orf1 of the human strain tls-18516-human and of all the rabbit hev strains. this insertion, also found in the rabbit hev strains from china (34), is not present in any known strain of hev genotypes 1-4 or in the new hev genotypes from rats and wild boars. the x domain corresponds to a macro domain found in the nonstructural polyproteins of several positive-stranded viruses such as togaviruses and coronaviruses (35) (36) (37) . this domain can bind polyadenosine diphosphate-ribose regions and could play a role in the replication or transcription of virus rna. whether the insertion in the x domain infl uences the function of the hev macro domain warrants further investigation. several determinants, including this insertion, could be essential for specifying the host range, zoonotic transmission, and pathogenesis of rabbit hev strains (34) . what rabbit hev strains contribute to the epidemiology of hepatitis e in humans is not clear. hev is endemic to southwestern france, and the annual incidence of locally acquired hev infections has been estimated as 3.2% (38, 39) . a case-control study found that the only factor independently associated with hev infection was the consumption of game meat, mostly wild boar, deer, and wild rabbit (23) . however, molecular data from various studies in france indicate that most hev strains identifi ed belong to genotypes 3f, 3c, or 3e, which are prevalent in pigs and wild boars (23, 26, 40) . a recent study showed the same proportions of genotypes 3f, 3c, and 3e in human and pig populations (33) . although this fi nding could indicate that rabbit hev strains are less readily transmitted to humans than hev genotype 3 strains, the primers used for pcr amplifi cation were not specifi cally designed for rabbit hev strains. therefore, the true prevalence of hev rna among rabbits and humans may have been underestimated. in addition, genotyping rabbit hev may have been diffi cult because reference sequences have become available only recently. the immunocompetent or immunocompromised status of the patient that became infected with a rabbit hev strain, as well as the source of his contamination, is unknown because of the lack of medical follow-up. molecular and epidemiologic studies are needed to determine the prevalence of rabbit hev strains among immunocompetent and immunocompromised patients. in conclusion, we have shown that in france, farmed and wild rabbits can be infected with hev. phylogenetic analysis, based on full-length genomes and a molecular signature in the x domain of orf1, indicates that rabbit hev strains could be a new genotype. our identifi cation of *hev, hepatitis e virus; hev1 (m73218-1a-burma, d11093-1b-japan, x98292-1c-india, ay204877-1e-chad, ay230202-1d-morocco); hev2 (m774506-2a-mexico) ; hev3 (af060669-3a-usa, ay115488-3j-canada-sw, ab291963-3b-japan, tls-tr19-3c, tls-tr02-3e, ab481226-3e-japan-sw,eu495148-3f-france,af455784-3g-kyrgyzstan-sw); hev4 (ab099347-4c-japan, ab108537-4g-china, ay594199-4d-china-sw). a human hev strain that is closely related to rabbit hev strains reinforces the potential zoonotic risk for infection with this virus. further studies are needed to demonstrate cross-species transmission directly and to evaluate the contribution of the rabbit reservoir to human hev infection and disease. hepatitis e: an emerging awareness of an old disease hepatitis e: an emerging infection in developed countries a novel virus in swine is closely related to the human hepatitis e virus hepatitis e virus: animal reservoirs and zoonotic risk locally acquired hepatitis e in chronic liver disease clinical course and outcome of sporadic acute viral hepatitis in pregnancy fulminant liver failure from acute autochthonous hepatitis e in france: description of seven patients with acute hepatitis e and encephalopathy chronic hepatitis e with cirrhosis in a kidney-transplant recipient chronic hepatitis e virus infection in liver transplant recipients hepatitis e virus and chronic hepatitis in organ-transplant recipients chronic hepatitis after hepatitis e virus infection in a patient with non-hodgkin lymphoma taking rituximab prolonged hepatitis e in an immunocompromised patient hepatitis e in an hivinfected patient persistent carriage of hepatitis e virus in patients with hiv infection hepatitis e virus seroprevalence and chronic infections in patients with hiv virus taxonomy: classifi cation and nomenclature of viruses: ninth report of the international committee on taxonomy of viruses pig liver sausage as a source of hepatitis e virus transmission to humans zoonotic transmission of hepatitis e virus from deer to human beings novel hepatitis e virus genotype in norway rats, germany analysis of the full-length genome of a hepatitis e virus isolate obtained from a wild boar in japan that is classifi able into a novel genotype the serological prevalence and genetic diversity of hepatitis e virus in farmed rabbits in china a novel genotype of hepatitis e virus prevalent among farmed rabbits in china characteristics of autochthonous hepatitis e virus infection in solid-organ transplant recipients in france acute hepatitis e in south-west france over a 5-year period a broadly reactive one-step real-time rt-pcr assay for rapid and sensitive detection of hepatitis e virus hepatitis e virus genotype 3 diversity, france phylogenetic analysis of global hepatitis e virus sequences: genetic diversity, subtypes and zoonosis hepatitis e virus in rabbits detection of hepatitis e virus in liver, mesenteric lymph node, serum, bile and faeces of naturally infected pigs affected by different pathological conditions hepatitis e virus load in swine organs and tissues at slaughterhouse determined by real-time rt-pcr evidence of extrahepatic sites of replication of the hepatitis e virus in a swine model zoonotic hepatitis e: animal reservoirs and emerging risks close similarity between sequences of hepatitis e virus recovered from humans and swine phylogenetic analysis of the full genome of rabbit hepatitis e virus (rbhev) and molecular biologic study on the possibility of cross species transmission of rbhev structural and functional basis for adp-ribose and poly(adpribose) binding by viral macro domains computer-assisted assignment of functional domains in the nonstructural polyprotein of hepatitis e virus: delineation of an additional group of positive-strand rna plant and animal viruses differential activities of cellular and viral macro domain proteins in binding of adp-ribose metabolites hepatitis e virus infection without reactivation in solid-organ transplant recipients hepatitis e virus antibodies in blood donors a national survey of acute hepatitis e in france we thank pascal bihannic, thierry delhorme, christian bernard, olivier galaup, francis berger, and jacky aubineau for their help with obtaining samples from wild rabbits.dr izopet is chief of the virology laboratory of toulouse university hospital. his research interests include the molecular virology of hev and its pathogenesis. key: cord-292475-jrl1fowa authors: abry, patrice; pustelnik, nelly; roux, stéphane; jensen, pablo; flandrin, patrick; gribonval, rémi; lucas, charles-gérard; guichard, éric; borgnat, pierre; garnier, nicolas title: spatial and temporal regularization to estimate covid-19 reproduction number r(t): promoting piecewise smoothness via convex optimization date: 2020-08-20 journal: plos one doi: 10.1371/journal.pone.0237901 sha: doc_id: 292475 cord_uid: jrl1fowa among the different indicators that quantify the spread of an epidemic such as the on-going covid-19, stands first the reproduction number which measures how many people can be contaminated by an infected person. in order to permit the monitoring of the evolution of this number, a new estimation procedure is proposed here, assuming a well-accepted model for current incidence data, based on past observations. the novelty of the proposed approach is twofold: 1) the estimation of the reproduction number is achieved by convex optimization within a proximal-based inverse problem formulation, with constraints aimed at promoting piecewise smoothness; 2) the approach is developed in a multivariate setting, allowing for the simultaneous handling of multiple time series attached to different geographical regions, together with a spatial (graph-based) regularization of their evolutions in time. the effectiveness of the approach is first supported by simulations, and two main applications to real covid-19 data are then discussed. the first one refers to the comparative evolution of the reproduction number for a number of countries, while the second one focuses on french departments and their joint analysis, leading to dynamic maps revealing the temporal co-evolution of their reproduction numbers. the ongoing covid-19 pandemic has produced an unprecedented health and economic crisis, urging for the development of adapted actions aimed at monitoring the spread of the new coronavirus. no country remained untouched, thus emphasizing the need for models and tools to perform quantitative predictions, enabling effective managements of patients or an optimized allocations of medical ressources. for instance, the outbreak of this unprecedented pandemic was characterized by a critical lack of tools able to perform predictions related to the pressure on hospital ressources (number of patients, masks, gloves, intensive care unit needs,. . .) [1, 2] . as a first step toward such an ambition goal, the present work focuses on the pandemic time evolution assessment. indeed, all countries experienced a propagation mechanism that is basically universal in the onset phase: each infected person happened to infect in average more than one other person, leading to an initial exponential growth. the strength of the spread is quantified by the so-called reproduction number which measures how many people can be contaminated by an infected person. in the early phase where the growth is exponential, this is referred to as r 0 (for covid-19, r 0 * 3 [3, 4] ). as the pandemic develops and because more people get infected, the effective reproduction number evolves, hence becoming a function of time hereafter labeled r(t). this can indeed end up with the extinction of the pandemic, r(t)!0, at the expense though of the contamination of a very large percentage of the total population, and of potentially dramatic consequences. rather than letting the pandemic develop until the reproduction number would eventually decrease below unity (in which case the spread would cease by itself), an active strategy amounts to take actions so as to limit contacts between individuals. this path has been followed by several countries which adopted effective lockdown policies, with the consequence that the reproduction number decreased significantly and rapidly, further remaining below unity as long as social distancing measures were enforced (see for example [4, 5] ). however, when lifting the lockdown is at stake, the situation may change with an expected increase in the number of inter-individual contacts, and monitoring in real time the evolution of the instantaneous reproduction number r(t) becomes of the utmost importance: this is the core of the present work. monitoring and estimating r(t) raises however a series of issues related to pandemic data modeling, to parameter estimation techniques and to data availability. concerning the mathematical modeling of infectious diseases, the most celebrated approaches refer to compartmental models such as sir ("susceptible-infectious-recovered"), with variants such as seir ("susceptible-exposed-infectious-recovered"). because such global models do not account well for spatial heterogeneity, clustering of human contact patterns, variability in typical number of contacts (cf. [6] ), further refinements were proposed [7] . in such frameworks, the effective reproduction number at time t can be inferred from a fit of the model to the data that leads to an estimated knowledge of the average of infecting contacts per unit time, of the mean infectious period, and of the fraction of the population that is still susceptible. these are powerful approaches that are descriptive and potentially predictive, yet at the expense of being fully parametric and thus requiring the use of dedicated and robust estimation procedures. parameter estimation become all the more involved when the number of parameters grows and/or when the amount and quality of available data are low, as is the case for the covid-19 pandemic real-time and in emergency monitoring. rather than resorting to fully parametric models and seeing r(t) as the by-product of their identification, a more phenomenological, semi-parametric approach can be followed [8] [9] [10] . this approach has been reported as robust and potentially leading to relevant estimates of r(t), even for epidemic spreading on realistic contact networks, where it is not possible to define a steady exponential growth phase and a basic reproduction number [6] . the underlying idea is to model incidence data z(t) at time t as resulting from a poisson distribution with a time evolving parameter adjusted to account for the data evolution, which depends on a function f(s) standing for the distribution of the serial interval. this function models the time between the onset of symptoms in a primary case and the onset of symptoms in secondary cases, or equivalently the probability that a person confirmed infected today was actually infected s days earlier by another infected person. the serial interval function is thus an important ingredient of the model, accounting for the biological mechanisms in the epidemic evolution. assuming the distribution f to be known, the whole challenge in the actual use of the semi-parametric poisson-based model thus consists in devising estimatesrðtþ of r(t) with satisfactory statistical performance. this has been classically addressed by approaches aimed at maximizing the likelihood attached to the model. this can be achieved, e.g., within several variants of bayesian frameworks [5, 6, 8, 10] , with even dedicated software packages (cf. e.g., https://shiny.dide.imperial. ac.uk/epiestim/). instead, we promote here an alternative approach based on inverse problem formulations and proximal-operator based nonsmooth convex optimisation [11] [12] [13] [14] [15] . the questions of modeling and estimation, be they fully parametric or semi-parametric, are intimately intertwined with that of data availability. this will be further discussed but one can however remark at this point that many options are open, with a conditioning of the results to the choices that are made. there is first the nature of the incidence data used in the analysis (reported infected cases, hospitalizations, deaths) and the database they are extracted from. next, there is the granularity of the data (whole country, regions, smaller units) and the specificities that can be attached to a specific choice as well as the comparisons that can be envisioned. in this respect, it is worth remarking that most analyses reported in the literature are based on (possibly multiple) univariate time series, whereas genuinely multivariate analyses (e.g., a joint analysis of the same type of data in different countries in order to compare health policies) might prove more informative. for that category of research work motivated by contributing in emergency to the societal stake of monitoring the pandemic evolution in real-time, or at least, on a daily basis, there are two classes of challenges: ensuring a robust and regular access to relevant data; rapidly developing analysis/estimation tools that are theoretically sound, practically usable on data actually available, and that may contribute to improving current monitoring strategies. in that spirit, the overarching goal of the present work is twofold: (1) proposing a new, more versatile framework for the estimation of r(t) within the semi-parametric model of [8, 10] , reformulating its estimation as an inverse problem whose functional is minimized by using non smooth proximal-based convex optimization; (2) inserting this approach in an extended multivariate framework, with applications to various complementary datasets corresponding to different geographical regions. the paper is organized as follows. it first discusses data, as collected from different databases, with heterogeneity and uneven quality calling for some preprocessing that is detailed. in the present work, incidence data (thereafter labelled z(t)) refers to the number of daily new infections, either as reported in databases, or as recomputed from other available data such as hospitalization counts. based on a semi-parametric model for r(t), it is then discussed how its estimation can be phrased within a non smooth proximal-based convex optimization framework, intentionally designed to enforce piecewise linearity in the estimation of r(t) via temporal regularization, as well as piecewise constancy in spatial variations of r(t) by graph-based regularization. the effectiveness of these estimation tools is first illustrated on synthetic data, constructed from different models and simulating several scenarii, before being applied to several real pandemic datasets. first, the number of daily new infections for many different countries across the world are analyzed independently. second, focusing on france only, the number of daily new infections per continental france départements (départements constitute usual entities organizing the administrative life in france) are analyzed both independently and in a multivariate setting, illustrating the benefit of this latter formulation. discussions, perpectives and potential improvements are finally discussed. datasets. in the present study, three sources of data were systematically used: • source1(jhu) johns hopkins university provides access to the cumulated daily reports of the number of infected, deceased and recovered persons, on a per country basis, for a large number of countries worldwide, essentially since inception of the covid-19 crisis (january 1st, 2020 time series. the data available on the different data repositories used here are strongly affected by outliers, which may stem from inaccuracy or misreporting in per country reporting procedures, or from changes in the way counts are collected, aggregated, and reported. in the present work, it has been chosen to preprocess data for outlier removal by applying to the raw time series a nonlinear filtering, consisting of a sliding-median over a 7-day window: outliers defined as ±2.5 standard deviation are replaced by window median to yield the pre-processed time series z(t), from which the reproduction number r(t) is estimated. an example of raw and pre-processed time series is illustrated in fig 3. when countries are studied independently, the estimation procedure is applied separately to each time series z(t) of size t, the number of days available for analysis. when considering continental france départements, we are given d time series z d (t) of size t each, where 1 � d � d = 94 indexes the départements. these time series are collected and stacked in a matrix of size d × t, and they analyzed both independently and jointly. model. although they can be used for envisioning the impact of possible scenarii in the future development of an on-going epidemic [3] , sir models, because they require the full estimation of numerous parameters, are often used a posteriori (e.g., long after the epidemic) with consolidated and accurate datasets. during the spread phase and in order to account for the on-line/on-the-fly need to monitor the pandemic and to offer some robustness to partial/ incomplete/noisy data, less detailed semi-parametric models focusing on the only estimation of the time-dependent reproduction number can be preferred [8, 9, 16] . let r(t) denote the instantaneous reproduction number to be estimated and z(t) be the number of daily new infections. it has been proposed in [8, 10] that {z(t), t = 1, . . ., t} can be modeled as a nonstationary time series consisting of a collection of random variables, each drawn from a poisson distribution p p t whose parameter p t depends on the past observations of z(t), on the current value of r(t), and on the serial interval function f(�): the serial interval function f(�) constitutes a key ingredient of the model, whose importance and role in pandemic evolution has been mentioned in introduction. it is assumed to be independent of calendar time (i.e., constant across the epidemic outbreak), and, importantly, independent of r(t), whose role is to account for the time dependencies in pandemic propagation mechanisms. for the covid-19 pandemic, several studies have empirically estimated the serial interval function f(�) [17, 18] . for convenience, f(�) has been modeled as a gamma distribution, with shape and rate parameters 1.87 and 0.28, respectively (corresponding to mean and standard deviations of 6.6 and 3.5 days, see [5] and references therein). these choices and assumptions have been followed and used here, and the corresponding function is illustrated in fig 1. in essence, the model in eq (1) is univariate (only one time series is modeled at a time), and based on a poisson marginal distribution. it is also nonstationary, as the poisson rate evolves along time. the key ingredient of this model consists of the poisson rate evolving as a weighted moving average of past observations, which is qualitatively based on the following rationale: whenr is above 1, the epidemic is growing and, conversely, when this ratio is below 1, it decreases and eventually vanishes. non-smooth convex optimisation. the whole challenge in the actual use of the semiparametric poisson-based model described above thus consists in devising estimatesrðtþ of r (t) that have better statistical performance (more robust, reliable, and hence usable) than the direct brute-force and naive form defined in eq 2. to estimate r(t), and instead of using bayesian frameworks that are considered state-of-the-art tools for epidemic evolution analysis, we propose and promote here an alternative approach based on an inverse problem formulation. its main principle is to assume some form of temporal regularity in the evolution of r(t) (we use a piecewise linear model in the following). in the case of a joint estimation of r(t) across several continental france départements, we further assume some form of spatial regularity, i.e., that the values of r(t) for neighboring départements are similar. univariate setting. for a single country, or a single département, the observed (possibly preprocessed) data {z(t), 1 � t � t} is represented by a t-dimensional vector z 2 r t . recalling that the poisson law is pðz ¼ njpþ ¼ p n n! e à p for each integer n � 0, the negative log-likelihood of observing z given a vector p 2 r t of poisson parameters p t is where r 2 r t is the (unknown) vector of values of r(t). up to an additive term independent of p, this is equal to the kl-divergence (cf. section 5.4. in [15] ): given the vector of observed values z, the serial interval function f(�), and the number of days t, the vector p given by (1) reads p = r � fz, with � the entrywise product and f 2 r t�t the matrix with entries f ij = f(i − j). maximum likelihood estimation of r (i.e., minimization of the negative log-likelihood) leads to an optimization problem min r d kl (zjr � fz) which does not ensure any regularity of r(t). to ensure temporal regularity, we propose a penalized approach usinĝ r ¼ argmin r d kl ðz j r � fzþ þ oðrþ where o denotes a penalty function. here we wish to promote a piecewise affine and continuous behavior, which may be accomplished [19, 20] using o(r) = λ time kd 2 rk 1 , where d 2 is the matrix associated with a laplacian filter (second order discrete temporal derivatives), k�k 1 denotes the ℓ 1 -norm (i.e., the sum of the absolute values of all entries), and λ time is a penalty factor to be tuned. this leads to the following optimization problem: spatially regularized setting. in the case of multiple départements, we consider multiple vectors (z d 2 r t , 1 � d � d) associated to the d time series, and multiple vectors of unknown (r d 2 r t , 1 � d � d), which can be gathered into matrices: a data matrix z 2 r t�d whose columns are z d and a matrix of unknown r 2 r t�d whose columns are the quantities to be estimated r d . a first possibility is to proceed to independent estimations of the (r d 2 r t , 1 � d � d) by addressing the separate optimization problemŝ which can be equivalently rewritten into a matrix form: is the entrywise ℓ 1 norm of d 2 r, i.e., the sum of the absolute values of all its entries. an alternative is to estimate jointly the (r d 2 r t , 1 � d � d) using a penalty function promoting spatial regularity. to account for spatial regularity, we use a spatial analogue of d 2 promoting spatially piecewise constant solutions. the d continental france départements can be considered as the vertices of a graph, where edges are present between adjacent départements. from the adjacency matrix a 2 r d�d of this graph (a ij = 1 if there is an edge e = (i, j) in the graph, a ij = 0 otherwise), the global variation of the function on the graphs can be computed as ∑ ij a ij (r ti − r tj ) 2 and it is known that this can be accessed through the so-called (combinatorial) laplacian of the graph: [21] . however, in order to promote smoothness over the graph while keeping some sparse discontinuities on some edges, it is preferable to regularize using a total variation on the graph, which amounts to take the ℓ 1 -norm of these gradients (r ti − r tj ) on all existing edges. for that, let us introduce the incidence matrix b 2 r e�d such that l = b > b where e is the number of edges and, on each line representing an existing edge e = (i, j), we set b e,i = 1 and b e,j = −1. then, the ℓ 1 -norm krb > k 1 = kbr > k 1 is equal to p t t¼1 p ði;jþ:a ij ¼1 jr ti à r tj j. alternatively, it can be computed as krb > k 1 ¼ p t t¼1 kbrðtþk 1 where rðtþ 2 r d is the t-th row of r, which gathers the values across all départements at a given time t. from that, we can define the regularized optimization problem: optimization problems (6) and (7) involve convex, lower semi-continuous, proper and non-negative functions, hence their set of minimizers is non-empty and convex [11] . we will discuss right after how to compute these using proximal algorithms. by the known sparsity-promoting properties of ℓ 1 regularizers and their variants, the corresponding solutions are such that d 2 r and/or rb > are sparse matrices, in the sense that these matrices of (second order temporal or first order spatial) derivatives have many zero entries. the higher the penalty factors λ time and λ space , the more zeroes in these matrices. in particular, when λ space = 0, no spatial regularization is performed and (7) is equivalent to (6) . when λ space is large enough, rb > is exactly zero, which implies that r(t) is constant at each time since the graph of départements is connected. optimization using a proximal algorithm. the considered optimization problems are of the form where f and g m are proper lower semi-continuous convex, and k m are bounded linear operators. a classical case for m = 1 is typically addressed with the chambolle-pock algorithm [22] , which has been recently adapted for multiple regularization terms as in eq. 8 of [23] . to handle the lack of smoothness of lipschitz differentiability for the considered functions f and g m , these approaches rely on their proximity operators. we recall that the proximity operator of a convex, lower semi-continuous function φ is defined as [24] prox φ ðyþ ¼ arg min in our case, we consider a separable data fidelity term: as this is a separable function of the entries of its input, its associated proximity operator can be computed component by component [25] : where τ > 0. we further consider g m (�) = k.k 1 , m = 1, 2, and k 1 (r) ≔ λ time d 2 r, k 2 (r) ≔ λ space rb > . the proximity operators associated to g m read: where (.) + = max(0,.). in algorithm 1, we express explicitly algorithm 161 of [23] for our setting, considering the moreau identity that provides the relation between the proximity operator of a function and the proximity operator of its conjugate (cf. eq. (8) of [23] ). the choice of the parameters τ and σ m impacts the convergence guarantees. in this work, we adapt a standard choice provided by [22] to this extended framework. the adjoint of k m , denoted k � m , is given by the sequence ðr ðkþ1þ þ k2n converges to a minimizer of (7) (cf. thm 8.2 of [23] ). input: data z, tolerance � > 0 ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi p m¼1;2 kk m k to assess the relevance and performance of the proposed estimation procedure detailed above, it is first applied to two different synthetic time series z(t). the first one is synthesized using directly the model in eq (1), with the same serial interval function f(t) as that used for the estimation, and using an a priori prescribed function r(t). the second one is produced from solving a compartmental (sir type) model. for such models, r(t) can be theoretically related to the time scale parameters entering their definition, as the ratio between the infection time scale and the quitting infection (be it by death or recovery) time scale [26, 27] . the theoretical serial function f associated to that model and to its parameters is computed analytically (cf., e.g., [28] ) and used in the estimation procedure. for both cases, the same a priori prescribed function r(t), to be estimated, is chosen as constant (r = 2.2) over the first 45 days to model the epidemic outbreak, followed by a linear decrease (till below 1) over the next 45 days to model lockdown benefits, with finally an abrupt linear increase for the last 10 days, modeling a possible outbreak at when lockdown is lifted. additive gaussian noise is superimposed to the data produced by the models to account for outliers and misreporting. for both cases, the proposed estimation procedure (obtained with λ time set to the same values as those used to analyze real data in the next section) outperforms the naive estimates (2), which turn out to be very irregular (cf. fig 2) . the proposed estimates notably capture well the three different phases of r(t) (stable, decreasing and increasing), with notably a rapid and accurate reaction to the increasing change in the 10 last days. the present section aims to apply the model and estimation tools proposed above to actual covid-19 data. first, specific methodological issues are addressed, related to tuning the hyperparameter(s) λ time or (λ time , λ space ) in univariate and multivariate settings, and to comparing the consistency between different estimates of r(t) obtained from the same incidence data, yet downloaded from different repositories. then, the estimation tools are applied to the estimation of r(t), both independently for numerous countries and jointly for the 94 continental france départements. estimation of r(t) is performed daily, with t thus increasing every day, and updated results are uploaded on a regular basis on a dedicated webpage (cf. http://perso.ens-lyon.fr/patrice. abry. regularization hyperparameter tuning. a critical issue associated with the practical use of the estimates based on the optimization problems (5) and (7) lies in the tuning of the hyperparameters balancing data fidelity terms and penalization terms. while automated and data-driven procedures can be devised, following works such as [29] and references therein, let us analyze the forms of the functional to be minimized, so as to compute relevant orders of magnitude for these hyperparameters. let us start with the univariate estimation (5). using λ time = 0 implies no regularization and the achieved estimate turns out to be as noisy as the one obtained with a naive estimator (cf. eq (2)). conversely, for large enough λ time , the proposed estimate becomes exactly a constant, missing any time evolution. tuning λ time is thus critical but can become tedious, especially because differences across countries (or across départements in france) are likely to require different choices for λ time . however, a careful analysis of the functional to minimize shows that the data fidelity term (9), based on a kullback-leibler divergence, scales proportionally to the input incidence data z while the penalization term, based on the regularization of r(t), is independent of the actual values of z. therefore, the same estimate for r(t) is obtained if we replace z with α × z and λ with α × λ. because orders of magnitude of z are different amongst countries (either because of differences in population size, or of pandemic impact), this critical observation leads us to apply the estimate not to the raw data z but to a normalized version z/std(z), alleviating the burden of selecting one λ time per country, instead enabling to select one same λ time for all countries and further permitting to compare the estimated r(t)'s across countries for equivalent levels of regularization. considering now the graph-based spatially-regularized estimates (7) while keeping fixed λ time , the different r(t) are analyzed independently for each département when λ space = 0. conversely, choosing a large enough λ space yields exactly identical estimates across départments that are, satisfactorily, very close to what is obtained from data aggregated over france prior to estimation. further, the connectivity graph amongst the 94 continental france départements leads to an adjacency matrix with 475 non-zero off-diagonal entries (set to the value 1), associated to as many edges as existing in the graph. therefore, a careful examination of (7) shows that the spatial and temporal regularizations have equivalent weights when λ time and λ time are chosen such that the use of z/std(z) and of (10) above gives a relevant first-order guess to the tuning of λ time and of (λ time , λ space ). estimate consistency using different repository sources. when undertaking such work dedicated to on-going events, to daily evolutions, and to a real stake in forecasting future trends, a solid access to reliable data is critical. as previously mentioned, three sources of data are used, each including data for france, which are thus now used to assess the impact of data sources on estimated r(t). source1(jhu) and source2(ecdpc) provide cumulated numbers of confirmed cases counted at national levels and (in principle) including all reported cases from any source (hospital, death at home or in care homes. . .). source3(spf) does not report that same number, but a collection of other figures related to hospital counts only, from which a daily number of new hospitalizations can be reconstructed and used as a proxy for daily new infections. the corresponding raw and (sliding-median) preprocessed data, illustrated in fig 3, show overall comparable shapes and evolutions, yet with clearly visible discrepancies of two kinds. first, source1(jhu) and source2(ecdpc), consisting of crude reports of number of confirmed cases are prone to outliers. those can result from miscounts, from pointwise incorporations of new figures, such as the progressive inclusion of cases from ehpad (care homes) in france, or from corrections of previous erroneous reports. conversely, data from source3 (spf), based on hospital reports, suffer from far less outliers, yet at the cost of providing only partial figures. second, in france, as in numerous other countries worldwide, the procedure on which confirmed case counts are based, changed several times during the pandemic period, yielding possibly some artificial increase in the local average number of daily new confirmed cases. this has notably been the case for france, prior to the end of the lockdown period (mid-may), when the number of tests performed has regularly increased for about two weeks, or more recently early june when the count procedures has been changed again, likely because of the massive use of serology tests. because the estimate of r(t) essentially relies on comparing a daily number against a past moving average, these changes lead to significant biases that cannot be easily accounted for, but vanishes after some duration controlled by the typical width of the serial distribution f (of the order of ten days). confirmed infection cases across the world. to report estimated r(t)'s for different countries, data from source2(ecdpc) are used as they are of better quality than data from source1(jhu), and because hospital-based data (as in source3(spf)) are not easily available for numerous different countries. visual inspection led us to choose, uniformly for all countries, two values of the temporal regularization parameter: λ time = 50 to produce a strongly-regularized, hence slowly varying estimate, and λ time = 3.5 for a milder regularization, and hence a more reactive estimate. these estimates being by construction designed to favor piecewise linear behaviors, local trends can be estimated by computing (robust) estimates of the derivativeŝ bðtþ ofrðtþ. the slow and less slow estimates ofrðtþ thus provide a slow and less slow estimate of the local trends. intuitively, these local trends can be seen as predictors for the forthcoming value of r:rðt þ nþ ¼rðtþ þ nbðtþ. let us start by inspecting again data for france, further comparing estimates stemming from data in source2(ecdpc) or in source3(spf) (cf. fig 4) . as discussed earlier, data from source2(ecdpc) show far more outliers that data from source3(spf), thus impacting estimation of r and β. as expected, the strongly regularized estimates (λ time = 50) are less sensitive than the less regularized ones (λ time = 3.5), yet discrepancies in estimates are significant, as data from source2(ecdpc) yields, for june 9th, estimates of r slightly above 1, while that from source3(spf) remain steadily around 0.7, with no or mild local trends. again, this might be because late may, france has started massive serology testing, mostly performed outside hospitals. this yielded an abrupt increase in the number of new confirmed cases, biasing upward the estimates of r(t). however, the short-term local trend for june 9th goes also downward, suggesting that the model is incorporating these irregularities and that estimates will return to unbiased after an estimation time controlled by the typical width of the serial distribution f (of the order of ten days). this recent increase is not seen in source3(spf)based estimates that remain very stable, potentially suggesting that hospital-based data are much less affected by changes in testing policies. this local analysis at the current date can be complemented by a more global view on what happened since the lifting of the lockdown. considering the whole period starting from may 11th we end up with triplets [5th percentile; median; 95th percentile] that read as given in table 1 : source2(ecdpc) provides data for several tens of countries. figs 5 to 8 reportrðtþ and bðtþ for several selected countries. more figures are available at perso.ens-lyon.fr/patrice.abry. as of june 9th (time of writing), fig 5 shows that, for most european countries, the pandemic seems to remain under control despite lifting of the lockdown, with (slowly varying) estimates of r remaining stable below 1, ranging from 0.7 to 0.8 depending on countries, and (slowly varying) trends around 0. sweden and portugal (not shown here) display less favorable patterns, as well as, to a lesser extent, the netherlands, raising the question of whether this might be a potential consequence of less stringent lockdown rules compared to neighboring european countries. fig 6 shows that whiler for canada is clearly below 1 since early may, with a negative local trend, the usa are still bouncing back and forth around 1. south america is in the above 1 phase but starts to show negative local trends. fig 7 indicates that iran, india or indonesia are in the critical phase withrðtþ > 1. fig 8 shows that data for african countries are uneasy to analyze, and that several countries such as egypt or south africa are in pandemic growing phases. phase-space representation. to complement figs 5 to 8, fig 9 displays a phase-space representation of the time evolution of the pandemic, constructed by plotting one against the other the local average (over a week) of the slowly varying estimated reproduction numberrðtþ and local trend, ð � rðtþ; � bðtþþ, for a period ranging from mid-april to june 9th. country names are written at the end (last day) of the trajectories. interestingly, european countries display a c-shape trajectory, starting with r > 1 with negative trends (lockdown effects), thus reaching the safe zone (r < 1) but eventually performing a u-turn with a slow increase of local trends till positive. this results in a mild but clear reincrease of r, yet with most values below 1 today, except for france (see comments above) and sweden. the usa display a similar c-shape though almost concentrated on the edge point r(t) = 1, β = 0, while canada does return to the safe zone with a specific pattern. south-american countries, obviously at an earlier stage of the pandemic, show an inverted c-shape pattern, with trajectory evolving from the bad top right corner, to the controlling phase (negative local trend, with decreasing r still above 1 though). phase-spaces of asian and african countries essentially confirm these c-shaped trajectories. envisioning these phase-space plots as pertaining to different stages of the pandemic (rather than to different countries), this suggests that covid-19 pandemic trajectory resembles a clockwise circle, starting from the bad top right corner (r above 1 and positive trends), evolving, likely by lockdown impact, towards the bottom right corner (r still above 1 but negative trends) and finally to the safe bottom left corner (r below 1 and negative then null trend). the lifting of the lockdown may explain the continuation of the trajectory in the still safe but. . . corner (r below1 and again positive trend). as of june 9th, it can be only expected that trajectories will not close the loop and reach back the bad top right corner and the r = 1 limit. continental france départements: regularized joint estimates. there is further interest in focusing the analysis on the potential heterogeneity in the epidemic propagation across a given territory, governed by the same sanitary rules and health care system. this can be achieved by estimating a set of localrðtþ's for different provinces and regions [5] . such a study is made possible by the data from source3(spf), that provides hospital-based data for each of the continental france départements . fig 4 (right) already reported the slow and fast varying estimates of r and local trends computed from data aggregated over the whole france. to further study the variability across the continental france territory, the graphbased, joint spatial and temporal regularization described in eq 7 is applied to the number of confirmed cases consisting of a matrix of size k × t, with d = 94 continental france départements, and t the number of available daily data (e.g., t = 78 on june 9th, data being available only after march 18th). the choice λ time = 3.5 leading to fast estimates was used for this joint study. using (10) as a guideline, empirical analyses led to set λ space = 0.025, thus selecting spatial regularization to weight one-fourth of the temporal regularization. first, fig 10 ( top row) maps and compares for june 9th (chosen arbitrarily as the day of writing) per-département estimates, obtained when départements are analyzed either independently (r indep using eq 6, left plot) or jointly (r joint using eq 7, right plot). while the means of r indep andr joint are of the same order ('0.58 and '0.63 respectively) the standard deviations drop down from '0.40 to '0.14, thus indicating a significant decrease in the variability across departments. this is further complemented by the visual inspection of the maps which reveals reduced discrepancies across neighboring departments, as induced by the estimation procedure. in a second step, short and long-term trends are automatically extracted fromr indep and r joint and short-term trends are displayed in the bottom row of fig 10 (left and right, respectively) . this evidences again a reduced variability across neighboring departments, though much less than that observed forr indep andr joint , likely suggesting that trends on r per se are more robust quantities to estimate than single r's. for june 9th, fig 10 also indicates reproduction numbers that are essentially stable everywhere across france, thus confirming the trend estimated on data aggregated over all france (cf. fig 4, right plot) . video animations, available at perso.ens-lyon.fr/patrice.abry/deptregul.mp4, and at barthes.enssib.fr/coronavirus/ixxi-sisyphe/., updated on a daily basis, report further comparisons betweenr indep andr joint and their evolution along time for the whole period of data availability. maps for selected days are displayed in fig 11 ( with identical colormaps and colorbars across time). fig 11 shows that until late march (lockdown took place in france on march 17th),r joint was uniformly above 1.5 (chosen as the upper limit of the colorbar to permit to see variations during the lockdown and post-lockdown periods), indicating a rapid evolution of the epidemic across entire france. a slowdown of the epidemic evolution is visible as early as the first days of april (with overall decreases ofr joint , and a clear north vs. south gradient). during april, this gradient rotates slightly and aligns on a north-east vs. south-west direction and globally decreases in amplitude. interestingly, in may, this gradient has reversed direction from south-west to north-east, though with very mild amplitude. as of today (june 9th), the pandemic, viewed hospital-based data from source3(spf), seems under control under the whole continental france. estimation of the reproduction number constitutes a classical task in assessing the status of a pandemic. classically, this is done a posteriori (after the pandemic) and from consolidated data, often relying on detailed and accurate sir-based models and relying on bayesian frameworks for estimation. however, on-the-fly monitoring of the reproduction number time evolution constitutes a critical societal stake in situations such as that of covid-19, when decisions need to be taken and action need to be made under emergency. this calls for a triplet of constraints: i) robust access to fast-collected data; ii) semi-parametric models for such data that focus on a subset of critical parameters; iii) estimation procedures that are both elaborated enough to yield robust estimates, and versatile enough to be used on a daily basis and applied to (often-limited in quality and quantity) available data. in that spirit, making use of a robust nonstationary poisson-distribution based semiparametric model proven robust in the literature for epidemic analysis, we developed an original estimation procedure to favor piecewise regular estimation of the evolution of the reproduction number, both along time and across space. this was based on an inverse problem formulation balancing fidelity to time and space regularization, and used proximal operators and nonsmooth convex optimization. this tool can be applied to time series of incidence data, reported, e.g., for a given country. whenever made possible from data, estimation can benefit from a graph of spatial proximity between subdivisions of a given territory. the tool also provides local trends that permit to forecast short-term future values of r. the proposed tools were applied to pandemic incidence data consisting of daily counts of new infections, from several databases providing data either worldwide on an aggregated percountry basis or, for france only, based on the sole hospital counts, spread across the french territory. they permitted to reveal interesting patterns on the state of the pandemic across the world as well as to assess variability across one single territory governed by the same (health care and politics) rules. more importantly, these tools can be used everyday easily as an onthe-fly monitoring procedure for assessing the current state of the pandemic and predict its short-term future evolution. updated estimations are published on-line every day at perso.ens-lyon.fr/patrice.abry and at barthes.enssib.fr/coronavirus/ixxi-sisyphe/. data were (and still are) automatically downloaded on a daily basis using routines written by ourselves. all tools have been developed in matlab™ and can be made available from the corresponding author upon motivated request. at the methodological level, the tool can be further improved in several ways. instead of using o(r) ≔ λ time kd 2 rk 1 + λ space krb > k 1 , for the joint time and space regularization, another possible choice is to directly consider the matrix d 2 rb > of joint spatio-temporal derivatives, and to promote sparsity with an ℓ 1 -norm, or structured sparsity with a mixed norm ℓ 1,2 , e.g., kd 2 rb > k 1,2 = ∑ t k(d 2 rb > )(t)k 2 . as previously discussed, data collected in the process of a pandemic are prone to several causes for outliers. here, outlier preprocessing and reproduction number estimation were conducted in two independent steps, which can turn suboptimal. they can be combined into a single step at the cost of increasing the representation space permitting to split observation in true data and outliers, by adding to the functional to minimize an extra regularization term and devising the corresponding optimization procedure, which becomes nonconvex, and hence far more complicated to address. finally, when an epidemic model suggests a way to make use of several time series (such as, e.g., infected and deceased) for one same territory, the tool can straightforwardly be extended into a multivariate setting by a mild adaptation of optimization problems (6) and (7), replacing the kullback-leibler divergence d kl (zjr � fz) by p i i¼1 d kl ðz i j r � fz i þ. finally, automating a data-driven tuning of the regularization hyperparameters constitutes another important research track. factors determining the diffusion of covid-19 and suggested strategy to prevent future accelerated viral infectivity similar to covid pooling data from individual clinical trials in the covid-19 era expected impact of lockdown in ile-de-france and possible exit strategies estimating the burden of sars-cov-2 in france the impact of a nation-wide lockdown on covid-19 transmissibility in italy measurability of the epidemic reproduction number in data-driven contact networks mathematical models in epidemiology a new framework and software to estimate time-varying reproduction numbers during epidemics the r0 package: a toolbox to estimate reproduction numbers for epidemic outbreaks improved inference of time-varying reproduction numbers during infectious disease outbreaks convex analysis and monotone operator theory in hilbert spaces image restoration: total variation, wavelet frames, and beyond proximal splitting methods in signal processing proximal algorithms. foundations and trends ® in optimization wavelet-based image deconvolution and reconstruction different epidemic curves for severe acute respiratory syndrome reveal similar impacts of control measures epidemiological parameters of coronavirus disease 2019: a pooled analysis of publicly reported individual data of 1155 cases from seven countries epidemiological characteristics of covid-19 cases in italy and estimates of the reproductive numbers one month into the epidemic nonlinear denoising for solid friction dynamics characterization sparsest continuous piecewise-linear representation of data the emerging field of signal processing on graphs: extending high-dimensional data analysis to networks and other irregular domains a first-order primal-dual algorithm for convex problems with applications to imaging proximal splitting algorithms: relax them all! fonctions convexes duales et points proximaux dans un espace hilbertien. comptes rendus de l'acadé mie des sciences de paris a douglas-rachford splitting approach to nonsmooth convex variational signal recovery on the definition and the computation of the basic reproduction ratio r0 in models for infectious diseases in heterogeneous populations reproduction numbers and sub-threshold endemic equilibria for compartmental models of disease transmission figs 10 and 11 are produced using open ressources from the openstreetmap foundation, whose contributors are here gratefully acknowledged. mapdata©openstreetmap contributors. conceptualization: patrice abry, pablo jensen, patrick flandrin. key: cord-288079-rr8h5dgy authors: prague, melanie; wittkop, linda; clairon, quentin; dutartre, dan; thiebaut, rodolphe; hejblum, boris pierre title: population modeling of early covid-19 epidemic dynamics in french regions and estimation of the lockdown impact on infection rate date: 2020-04-24 journal: nan doi: 10.1101/2020.04.21.20073536 sha: doc_id: 288079 cord_uid: rr8h5dgy we propose a population approach to model the beginning of the french covid-19 epidemic at the regional level. we rely on an extended susceptible-exposed-infectious-recovered (seir) mechanistic model, a simplified representation of the average epidemic process. combining several french public datasets on the early dynamics of the epidemic, we estimate region-specific key parameters conditionally on this mechanistic model through stochastic approximation expectation maximization (saem) optimization using monolix software. we thus estimate basic reproductive numbers by region before isolation (between 2.4 and 3.1), the percentage of infected people over time (between 2.0 and 5.9% as of may 11th, 2020) and the impact of nationwide household confinement on the infection rate (decreasing the transmission rate by 72% toward a re ranging from 0.7 to 0.9). we conclude that a lifting of the lockdown should be accompanied by further interventions to avoid an epidemic rebound. in december 2019, grouped pneumonia cases have been described in the hubei province, china and sars-cov2 was identified on january, 7 th as the cause of this outbreak (li et al., 2020a; zhu et al., 2020) . sars-cov2 causes the viral disease which has been named covid-19 (world health organization, 2020b) . sars-cov2 rapidly spread all over the world and the pandemic stage was declared on march 11 th by the world health organization (2020c). on april 28 th , over 1,773,084 cases (in accordance with the applied case definitions and testing strategies in the affected countries) including 111,652 deaths have been reported (world health organization, 2020a) . the first case in france was declared on january, 24 th (bernard-stoecklin et al., 2020) and on april 13 th , santé publique france reported 98,076 confirmed cases and 14,967 hospital deaths due to covid-19 includes non-specific symptoms such as fever, cough, headache, and specific symptoms such as loss of smell and taste (gane et al., 2020; greenhalgh et al., 2020) . the virus is transmitted through droplets and close unprotected contact with infected cases. the majority (around 80 %) of infected cases have a mild form (upper respiratory infection symptoms) without specific needs in terms of care. around 20 % of cases need hospitalization and among those are severe forms (severe respiratory distress) which will need to be admitted to intensive care units (icu) with potential need of mechanical ventilation. the percentage of patients in need for icu care varies between 5 % reported from china (guan et al., 2020) and 16 % reported from italy (grasselli et al., 2020) . the number of icu beds in france was 5,058 at the end of 2018 (drees, 2019) (although it is currently being increased, having doubled and aiming to reach 14,000 according to the french minister of health). thus, the availability of icu beds with mechanical ventilation is one of the major issues as facilities are not prepared to deal with the potential increase of the number of patients due to this epidemic. unprecedented public-health interventions have been taken all over the world (kraemer et al., 2020) to tackle this epidemic. in france, interventions such as heightening surveillance with rapid identification of cases, isolation, contact tracing, and follow-up of potential contacts were initially implemented. but as the epidemic continued growing, comprehensive physical distancing measures have been applied since march 15 th , 2020 including closing of restaurants, non-vital business, schools and universities etc, quickly followed by state-wide lockdown on march 17 th 2020. the president has announced on april 13 th 2020, a progressive lifting of the lockdown from may 11 th 2020 onwards. in wuhan (hubei, china), the extremely comprehensive physical distancing measures in place since january 23 rd have started to be relaxed after 2 months of quarantine and lifted completely on april 8 th 2020 (tian et al., 2020; wu and mcgoogan, 2020) . interestingly, these interventions have been informed by mathematical models used to estimate the epidemic key parameters as well as unmeasured compartments such as the number of infected people. another interesting outcome is the forecast of the covid-19 epidemic according to potential interventions. several models have already been proposed to model and forecast the covid-19 epidemic using compartment models (fang et al., 2020; tang et al., 2020; or agent based models (di domenico et al., 2020a; ferguson et al., 2020; wilder et al., 2020) , its potential impact on intensive care systems (fox et al., 2020; massonnaud et al., 2020) , and to estimate the effect of containment measurements on the dynamics of the epidemic (magal and webb, 2020; prem et al., 2020) . most of those rely on simulations with fixed parameters and do not perform direct statistical estimations from incident data (massonnaud et al., 2020) . roques et al. (2020) used french national data but did not use a population approach to model the epidemic at a finer geographical granularity. yet, the dynamics of the epidemic can be very heterogeneous between regions inside a given country resulting in tremendous differences in terms of needs for hospital and icu beds (massonnaud et al., 2020) . moreover, the data collection yields noisy observations, that we deal with a statistical modeling of the observation process, rather than altering the data by e.g. smoothing such as in roques et al. (2020) . in the present study, we use public data from the covid-19 outbreak in france to estimate the dynamics of the covid-19 epidemic in france at the regional level. we model the epidemic with a seirah model, which is an extended susceptible-exposed-infectious-recovered (seir) model accounting for time-varying population movements, non-reported infectious subjects (a for unascertained) and hospitalized subjects (h) as proposed by to model the epidemic in wuhan. parameters from the model are estimated at the regional scale using a population approach which allows for borrowing information across regions, increasing the amount of data and thereby strengthening the inference while allowing for local disparities in the epidemic dynamics. furthermore, we use forward simulations to predict the effect of non-pharmaceutical interventions (npi) (such as lift of lockdown) on icu bed availability and on the evolution of the epidemic. section 2 introduces the data, the model and the necessary statistical tools, section 3 presents our results and section 4 discusses our findings and their limits. because epidemics spread through direct contacts, their dynamics have a strong spatial component. while traditional compartment models do not account for spatiality, we propose to take it into account by: i) modeling the epidemic at a finer, more homogeneous geographical scale (this is particularly important once lockdown is in place); ii) by using a population approach with random effects across french regions which allows each region to have relatively different dynamics while taking all information into account for the estimation of model parameters iii) aligning the initial starting time of the epidemic for all regions. the starting date in each region was defined as the first date with incident confirmed cases of covid-19 directly followed by 3 additional consecutive days with incident confirmed cases as well. this criterion of 4 consecutive days with incident cases is needed in particular for the île-de-france region which had 3 consecutive days with 1 imported confirmed case in late january which did not lead to a spreading outbreak at that time. open-data regarding the french covid-19 epidemic is currently scarce, as the epidemic is still unfolding. santé publique france (spf) in coordination with the french regional health agencies (agences régionales de santé -ars) has been reporting a number of aggregated statistics at various geographical resolutions since the beginning of the epidemic. during the first weeks of the epidemic in france, spf was reporting the cumulative number of confirmed covid-19 cases with a positive pcr test. other french surveillance resources such as the réseau sentinelles (valleron et al., 1986) or the sursaud r database (caserio-schönemann et al., 2014) quickly shifted their focus towards covid-19, leveraging existing tools to monitor the ongoing epidemic in real time, making as much data available as possible (given privacy concerns). in this study, we combined data from three different opendata sources: i) the daily release from spf; ii) the sursaud r database that started recording visits to the emergency room for suspicion of covid-19 on february, 24 th ; iii) the réseau sentinelles which started estimating the weekly incidence of covid-19 in each french region on march 16 th . from the daily release of spf, we computed the daily incident number of confirmed covid-19 cases (i.e. with a positive pcr test) in each region. in 4 . cc-by 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. (which was not certified by peer review) the copyright holder for this preprint this version posted april 24, 2020. . addition we used the incident number of visits to the emergency room for suspicion of covid-19 in each region from the sursaud r database using the oscour r network that encompasses more than 86% of all french emergency services (caserio-schönemann et al., 2014) . although this does not represent the full extend of hospitalized covid-19 cases, it is the only public data available that early in the epidemic, when the majority of covid-19 cases at hospitals were admitted through emergency rooms. finally, we used the réseau sentinelles network's weekly incidence estimates of symptomatic cases (including non confirmed cases) to set the ratio between ascertained and unascertained cases in each region (later denoted as r i ). table 1 presents these observed data. of note, we studied the epidemic in the 12 metropolitan french regions -excluding the corsican region (corse) which exhibits different epidemic dynamics, possibly due to its insular nature. wang et al. (2020) extended the classic seir model to differentiate between different statuses for infected individuals: ascertained cases, unascertained cases and cases quarantined by hospitalization. the model, assuming no population movement, is presented in figure 1 . the population is divided into 6 compartments: susceptible s, latent e, ascertained infectious i, unascertained infectious a, hospitalized infectious h, and removed r (recovered and deceased). this model assumes that infections are well-mixed throughout the population, ignoring any spatial structure or compartmentalization by population descriptors such as age. such assumptions make it particularly relevant to infer the dynamics of the french epidemic at the regional level (a finer geographical scale at which such hypotheses are more likely to hold). figure 1 illustrates the dynamics between those 6 compartments that are characterized by the following system of six ordinary differential equations (ode): cc-by 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. (which was not certified by peer review) model parameters are described in table 2 . of note, given a combination of parameters and initial states of the system ξ = , using a solver of differential equations, it is possible to deterministically compute at any time t the quantities s(t, ξ), e(t, ξ), i(t, ξ), r(t, ξ), a(t, ξ), and h(t, ξ). 6 . cc-by 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted april 24, 2020. observation processes in our case, none of the compartments of the system are directly observed: the only observations considered are i) the number of daily incident infectious ascertained cases denoted y 1 , and ii) the number of daily incident hospitalized infectious cases denoted y 2 . these observations are the only one available both before and after the initiation of lockdown. those two quantities are modeled in equation (1) respectively as observations from the i (in) (t, ξ) = re(t,ξ) de and h (in) (t, ξ) = i(t,ξ) dq random variables, which are the numbers of new incident cases at time t given the parameters ξ in compartment i and h respectively. because these are count processes, we propose to model their observations y 1 and y 2 with poisson likelihoods: where k 1 and k 2 are the respective numbers of cases. 8 . cc-by 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted april 24, 2020. . the initial states of all compartments at the date of epidemic start (t = 0) for region i are also important drivers of the dynamics. some of them can be approximated by quantities directly depending on the observations: i) , and iii) r i (t = 0) = 0. others, namely compartments a and e are not directly observed, and we evaluate these initial quantities. due to variation in data collection protocols and the initial size of regional outbreaks, this estimation is particularly important. indeed, the number of daily incident cases at t = 0 ranges from 1 to 37 cases depending on the region. a i (t = 0) is set as the goal of this study is to model the epidemic of covid-19 in france, but at the regional level using a population approach. this is done using a mixed effect model. in this inference framework, baseline parameters governing the dynamics of the epidemic in each region are assumed to be drawn from a shared distribution which allows for heterogeneity between regions, known as the random effects. we use the log-normal distribution for all parameters to ensure their positivity during estimation. because public health policies changed over the time period of observation of the epidemic, we incorporate explanatory covariates such as physical distancing by lockdown (c 1 and c 2 ) as a time-dependent effect on the transmission of the disease b. covariate c 1 is 0 until 2020-03-17 date of the start of the policy in france and then set to 1. covariate c 2 is 0 until 2020-03-25 assuming that social distancing behaviours build up in a week. in other words, we have ∀i = 1, . . . , 12 (where i is the region identifier): the parameters (b 0 , d q 0 , e(t = 0)) are mean shared values in the population, and can be seen as the country values for these parameters. the inter-region random-effect (u b i , u dq i , u e 0 i ) are normally distributed and assumed independent. so the vector of parameters in the model for each region is cc-by 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted april 24, 2020. . which is the factor by which transmission is modified after the start of lockdown during the first week. the factor by which transmission is modified after that first week of confinement is given by k 2 = exp(−β 1 − β 2 ). the coefficients β = (β 1 , β 2 ) are expected to be negative as lockdown aims at reducing transmission. interestingly, with our approach, we can evaluate whether or not there is a statistically significant effect of lockdown on the transmission by testing β = 0 using a wald test. region-specific model parameters based on the results from the theoretical identifiability analysis of the structural model of the epidemic from equation (1) (see appendix a2), we estimate the parameters (b i , d q i , r i ) as well as the initial state (e 0i ) when the epidemic begins being reported in each region i. we used the monolix software version 2019r2 (lixoft sas, 2019) to estimate those five parameters by maximizing the likelihood of the data given the model and the other fixed parameters. this software relies on a frequentist version of the stochastic approximation expectation maximization (saem) algorithm (delyon et al., 1999) and standard errors are calculated via estimation of the fisher information matrix (kuhn and lavielle, 2005) , which is derived from the second derivative of the log-likelihood evaluated by importance sampling. in addition, we use profile-likelihood to confirm that no further information can be gained from the data at hand on parameters α, d e and d i by running the saem algorithm multiple times while setting these parameters to different values and obtaining similar maximum likelihood values (meaning more data would be needed to be able to estimate those parameters). during inference, practical identifiability of the model is evaluated by the ratio of the minimum and maximum eigenvalues of the fisher information matrix, that will be referred as "convergence ratio" in the reminder of the manuscript. convergence of the saem algorithm was assessed by running multiple saem chains and checking that they all mix around similar probability distributions. we are particularly interested in the trajectories of the model compartments. we use monte carlo methods (parametric bootstrapping) to compute the confidence intervals accounting for the uncertainty in estimating the structural and statistical model parameters. for 10 . cc-by 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted april 24, 2020. . all compartments c(t, ξ i ) (c being s, e, i, r, a, or h) the 95% confidence interval is estimated by sampling from the posterior distributions of the model parameters to simulate 1,000 trajectories, and taking the 2.5% and 97.5% percentiles of these simulated trajectories. we also added to it the error measurement given by the poisson distribution of the outcomes. other outcomes of interest are the number of icu beds needed and the number of death (d) in a given region at a given time. these quantities are not specifically modeled by our mathematical structural model. however, it is possible to roughly approximate them by assuming that they represent a percentage of the hospitalized cases h(t, ξ i ) and removed cases r(t, ξ i ). we assume that icu (t, ξ i ) = 0.25 × h(t, ξ i ) which is consistent with the prevalence of icu cases among hospitalized cases at the french national level. based on the estimation of the infection fatality ratio (ifr) from roques et al. (2020), we get a rough estimation of d(t, ξ i ) as 0.5% of r(t, ξ i ). roques et al. (2020) conclude that covid-19 fatalities are under-reported, and using their ifr estimate we adequately fit the trend of the observed covid-19 deaths but with an offset due to this assumed higher ifr, see appendix a1. model update furthermore β estimations can be easily updated as new data become available using parametric empirical bayes (thus avoiding the need to re-estimate the whole system). it consists in maximizing the likelihood again with respect to β while holding the other parameter distribution fixed to their previously inferred a posteriori distribution. this is how our results are updated with data after march 25 th 2020 in this work. effective reproductive number for each region, we compute the effective reproductive number r e (t, ξ i ) as a function of model parameters: when individuals are homogeneous and mix uniformly, r e (t, ξ i ) is defined as the mean number of infections generated during the infectious period of a single infectious case in the region i at time t. this is the key parameter targeted by npis. we compute analytically its 95% confidence interval by accounting for all 95% confidence interval [x min ; x max ] of parameters and 11 . cc-by 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted april 24, 2020. . trajectories x used in its definition such that: asymptomatic proportion at a given time t the number of incident unascertained cases is equal to the sum of two populations, the number of incident non-tested symptomatic individuals (nt) and the number of incident non-tested asymptomatic individuals (as): where r is the proportion of cases tested positive. collection of data from general practitioners through the re-purposing of the réseau sentinelles network to monitor covid-19 provides a weekly estimation of the number of incident symptomatic cases (tested or not tested) that we previously called r s . this quantity is given over a week but can be evaluated daily by averaging: where r s represents the proportion of infected cases seeing a general practitioner. combining equations (6) and (7) allows to compute the incident number of asymptomatic cases as a function of the compartment e: short-term predictions of attack rates we predict the proportion of infected individuals among the population in each region at a given date by computing: 12 . cc-by 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted april 24, 2020. . given the values of parameters ξ i , we predict the trajectories of the dynamical system compartments using the dlsode differential equation solver in r (soetaert et al., 2020) and we can investigate the impact of npis such as lockdown in various scenarios. this impact is driven by two parameters: • k 2 (k 1 is considered fixed), the decrease ratio of transmission rate of the disease following the first week of npi, defined as this directly translates into a decrease of the effective reproductive number according to equation (4). it reflects the fact that individual by getting confined decrease their number of contacts. of note, the current k 2 is estimated by exp(− β 1 − β 2 ) (see section 2.2.3). • τ , the duration (in days) of the lockdown during which the mixing and transmission are fixed to we evaluate the magnitude of the possible epidemic rebound after confinement according to several values for k 2 . in particular, we predict the rates e(t, ξ i )/n i , a(t, ξ i )/n i and i(t, ξ i )/n i on may 11 th 2020 (currently considered by french authorities as the possible start date for lifting lockdown in france). we also compute the optimal lockdown duration τ opt i needed to achieve the epidemic extinction in region i defined as e i (t, ξ i ) < 1 and a i (t, ξ i ) < 1 and i i (t, ξ i ) < 1 simultaneously. in each scenario we predict the date at which the icu capacities in each region would be overloaded, this date is given by: η i denoting the icu capacities limits in region i. we additionally predict how many more icu beds would be needed at the peak of hospitalization in each scenario, as a proportion of current icu capacity (drees, 2019). finally we provide a rough prediction of the number of deaths for each envisioned scenario assuming a confinement duration of τ opt i days. data fitting because of the lag inherent to diagnostic testing, we also estimated the number of people already infected at this epidemic start by e 0 (notably, the largest numbers of e 0 in table 3 are estimated for île-de-france and grand est the two most affected french regions in this early epidemic). . cc-by 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted april 24, 2020. . on march 25 th , the cumulative number of ascertained cases was 24,623 and the cumulative number of hospitalized cases was 13,388, see table 1 for a regional breakdown. our seirah model fits the data well as can be seen in figure 2 . moreover, the stability of the estimates is good with a convergence ratio of 1.6 (see section 2.3.1), corroborating the good identifiability of the estimated parameters (see appendix a2). table 3 provides the regional estimates of the transmission rates (b i ). of note, regions with higher transmission rate are not necessarily those known to have the highest number of incident ascertained cases. d q i , the number of days from illness onset to hospitalization, can be quite variable between regions and likely accounts for heterogeneity in the observed data. we estimates its population mean at d q 0 = 1.13 days with a standard deviation σ dq = 0.42. to evaluate the validity of our structural ode model (1) and of our inference results, we compare the aggregated predictions of the number of both incident ascertained cases and incident hospitalized cases at the national french level to the daily observed incidences (that are still publicly available from spf at the country level, even after march 25 th -while incident ascertained cases are not openly availilable at the regional level after march 25 th ). figure 3 displays both predictions and observations, illustrating the added value of incorporating data after march 25 th as those encompass new information about the epidemic dynamics and characteristics as we approach the peak in most regions (notably grand est and île de france). of note, worse fit of the observed hospitalizations by our model can be explained by the data discrepancy: while we use the sursaud r data for our inference (which only account for patients arriving through the emergency room), figure 3 displays the data from santé publique france which should contain all covid-19 hospitalizations in france (including hospitalizations not coming from the emergency room, as well as corsica and the french departements d'outre mer that are not taken into account in our model). 14 . cc-by 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted april 24, 2020. . is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted april 24, 2020. evolution of the epidemic without intervention it is also interesting to predict the percentage of infected individuals in each region at future dates, corresponding to attack rates. change of transmission rate during lockdown the parameter β 1 and β 2 measure the effect of the lockdown before and after a week of adjustment. both are significantly different from 0 (p<0.001) such that the lockdown reduced the transmission rate of covid-19 by a divisive factor k 1 estimated at 1.31 [1.27; 1.35] during the first week and k 2 estimated at 3.63 [3.48; 3.80] after this first week. of note, thanks to our update algorithm (see section 2.3.1), it is possible to update those results rapidly as soon as more data are available to inform which scenario of prediction described in section 3.3 is the most likely. effective reproductive number the above quantities directly impact the effective reproductive number as described in figure 4 displays the effective reproductive number trajectories in each region. in tables 5 and 6, we vary k 2 = 3, 5, 10 (the magnitude of the reduction of transmission during lockdown after the first week). this gradient of simulation is important because the actual french value of k 2 remains currently unknown. in section 3.2 we showed that we can estimate a lower bound for k 2 ≥ k 2 = 3.48. from table 5 , we show that the higher k 2 is, the lowest 17 . cc-by 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted april 24, 2020. . . cc-by 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted april 24, 2020. table 4 : estimation of the effective reproductive ratios r e during each of the 3 considered periods (before lockdown, during the first week of lockdown, and beyond 1 week of lockdown) for each region with 95% confidence intervals. the numbers of ascertained (i), unascertained (a) and latent (e) infected individuals are on may 11 th 2020. however, it is not equal to 0, which means the epidemic is not extinct (and ready to bounce back as soon as lockdown is lifted). in table 6 , we predict the optimal (i.e. shortest) duration of the lockdown to achieve extinction of the epidemic in each region, which is, table 7 presents the proportion of infected individuals at various dates, i.e. instantaneous attack rates. we also predict them for three horizon dates assuming confinement would be maintained until these dates: 2020-05-15, 2020-06-08 and 2020-06-22. we predict the national french attack rate on may 15 th 2020 to be 3.8% [3.1%; 4.8%]. . cc-by 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted april 24, 2020. . cc-by 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted april 24, 2020. . . cc-by 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted april 24, 2020. is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted april 24, 2020. table 7 : model predictions for the proportion of infected and immunized in the population (deaths not taken into account), assuming continued lockdown until then. . it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted april 24, 2020. . lockdown lift on may 11 th 2020 we simulated the effect of lifting the lockdown on may 11 th 2020 assuming that after this date the transmission goes back to its value before lockdown. figure 5 shows the predicted dynamics for each region. in every region, we observed a large rebound occurring either in june or july. the timing and magnitude of this rebound is largely influenced by the importance of the first wave, that is successfully contained thanks to the lockdown. these results strongly argue for enforcing other npis when lockdown is lifted in order to contain r e below 1 and prevent this predictable rebound of the epidemic. in this work, we provide estimations of the key parameters of the dynamics of the covid-19 epidemic in french regions as well as forecasts according to npis especially regarding the proportion of infected when lifting the lockdown policy. the point estimates of the basic reproductive ratios for french regions fluctuated between 2.4 and 3.4 before lockdown took effect, but according to the uncertainty around these estimates they are not substantially different from one region to another. therefore, observed differences in the number of cases were due to the epicemic starting first in grand est and île-de-france regions. these estimates were close to those reported before isolation using other models (alizon et al., 2020; flaxman et al., 2020) . the model provided estimates of the impact of the lockdown on the effective reproductive ratio and although recent data led to a substantial reduction of r e after the lockdown, it remains close to 1 thus without a clear extinction of the epidemic. these estimates should be updated with more recent data that may lead to an estimated r below 1. on the other hand, it is an argument to add other measures such as intensive testing and strict isolation of cases. in addition, the model provides estimates of the size of the population of people who have been or are currently infected. as already reported (di domenico et al., 2020a) , this proportion of subjects is around 2 to 4 percent, so excluding any herd immunity and control of the epidemic by having a large proportion of people already infected and therefore not susceptible. with our estimates of basic reproduction ratio, the epidemic would become extinct by herd immunity with a proportion of 89.5% (95% ci [88.0%; 90.7%]) of infected people. interpretation of our results is conditional on the mechanistic model illustrated in figure 1 , and careful attention must be given to the parameters set 24 . cc-by 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted april 24, 2020. . table 2 , as updated estimates are published every day. first and foremost, our model takes only two kinds of infectious cases into account: confirmed cases i, and unascertained cases a. our observation model takes i as the number of infectious cases confirmed by a positive pcr sars-cov-2 test. thus, a can be interpreted as 25 . cc-by 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted april 24, 2020. . unconfirmed symptomatic cases that can be diagnosed by a gp visit (possibly through remote teleconsultation). this is a very simple representation of the covid-19 infection, which can have various degrees of severity (e.g. asymptomatic, mild, severe) that could be themselves modeled into different compartments. however, very little data is currently available to gather sufficient information to be able to distinguish between those infectious states. second, our model does not have a compartment for covid-19 patients in icu, and the number of occupied icu beds is simply taken as a fixed percentage (25% based on an estimate from the bordeaux chu university hospital). meanwhile icu bed capacity does not account for the recent surge of available icu beds in response to the covid-19 epidemic. compared to , our model does not feature an inflow of susceptibles n (and matching outflow) but population movement across regions are limited during the isolation period (see appendix a3 for a thorough discussion). deaths were also not distinguished from recoveries in the r compartment, but over the observation period this did not impact the main estimates. third, our model does not take into account the age-structure of the population on the contrary to the recently posted report salje et al. (2020) using french data. interestingly, although the models were different and the data not fully identical, our results were comparable. actually, our approach captures a part of the unexplained variability between regions through the random effects. this variability might be explained at least partly through the difference in age-structure and probability of hospitalization according to the age. we would like to underline the interest of making the data publicly accessible as done by santé publique france on the data.gouv.fr web portal, hence allowing our group to work immediately on the topics. furthermore, we have made our code fully available on github www.github.com/sistm/ seircovid19 and we are currently working on packaging our software for facilitating its dissemination and re-use. in conclusion, the lockdown has clearly helped controlling the epidemics in france in every region. the number of infected people varies from one region to the other because of the variations in the epidemic start in these regions (both in terms of timing and size). hence, the predicted proportion of infected people as of may 11 varies, but stays below 10 % everywhere. it is clear from this model, as in other published models (di domenico et al., 2020b; flaxman et al., 2020) , that a full and instantaneous lockdown lift would lead to a rebound. additional measures may help in controlling the number of new infections such as strict case isolation, contact tracing (di domenico et al., 2020b) and certainly a protective vaccine for which the 26 . cc-by 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted april 24, 2020. . strategy of administration to the population remains to be defined (amanat and krammer, 2020; lurie et al., 2020; thanh et al., 2020) . the data from the sursaud r database regarding covid-19 is available from the data.gouv french government platform at https://www.data. gouv.fr/fr/datasets/donnees-des-urgences-hospitalieres-et-de-sosmedecins-relatives-a-lepidemie-de-covid-19. the source code used for this work is available on github at www.github.com/sistm/seircovid19. . cc-by 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted april 24, 2020. . strategies to reduce social mixing on outcomes of the covid-19 epidemic in wuhan, china: a modelling study. lancet public health. raue, a., kreutz, c., maiwald, t., bachmann, j., schilling, m., klingmüller, u., and timmer, j. (2009) . structural and practical identifiability analysis of partially observed dynamical models by exploiting the profile likelihood. bioinformatics, 25(15):1923 -1929 . roques, l., klein, e., papaix, j., sar, a., and soubeyrand, s. (2020 . using early data to estimate the actual infection fatality ratio from covid-19 in france. medrxiv. tian, h., liu, y., li, y., wu, c.-h., chen, b., kraemer, m. u. g., li, b., cai, j., xu, b., yang, q., wang, b., yang, p., cui, y., song, y., zheng, p., wang, q., bjornstad, o. n., yang, r., grenfell, b. t., pybus, o. g., and dye, c. (2020) . an investigation of transmission control measures during the first 50 days of the covid-19 epidemic in china. science. valleron, a.-j., bouvet, e., garnerin, p., ménares, j., heard, i., letrait, s., and lefaucheux, j. (1986) . a computer network for the surveillance of communicable diseases: the french experiment. american journal of public health, 76(11):1289-1292. wang, c., liu, l., hao, x., guo, h., wang, q., huang, j., he, n., yu, h., lin, x., pan, a., wei, s., and wu, t. (2020) . evolving epidemiology and 31 . cc-by 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted april 24, 2020. . figure s1 : cumulative incidence hospitalization for covid-19 at france national level according to either santé publique france or the sursaud r database 33 . cc-by 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted april 24, 2020. . of note, the earlier slowing down of emergency admissions for covid-19 suspicions observed in the sursaud r data compared to the si-vic data could lead to an optimistic biais regarding the epidemic stage and evolution in our estimations and predictions . however, since the si-vic data are not publicly accessible before march 18 th , we are currently unable to use this data source to estimate the impact of the lockdown. under-reporting or over-estimation of covid-19 deaths according to roques et al. (2020) , the infection fatality ratio (ifr) for covid-19 is 0.5% (95%-ci: 0.3; 0.8). figure s2 shows that using this ifr over-estimate compared to the death currently reported by santé publique france. figure s2 : observed incident number of ascertained cases and hospitalization at france national level compared to predicted ones, based on estimations with data collected up to either march 16 (before lockdown), march 25 th or april 6 th (delimited by the vertical line). . cc-by 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted april 24, 2020. . based on ode model (1), the statistical analysis of the population evolution can be turned into a parameter estimation problem. theoretical identifiability (i.e. the possibility of learning the true values of the model parameters) of epidemiological compartment models is rarely checked. although it relies on unrealistic hypothesis (namely that incident number of ascertained cases and hospitalized cases are observed continuously, without noise and for an infinite length of time in our case), this framework provides important guarantees for the results. to determine which parameters from table 2 can be accurately identified from the available observations we first evaluate the identifiability of this seirah structure with the daisy software from bellu et al. (2007) (based on differential algebra results). we conclude that there is global identifiability of ξ = (b, d q , r, d e , d i ) with known α and d h , even if initial conditions are unknown -which is our case for (e 0 , a 0 ). practical identifiability (for which most existing evaluation methods rely on estimations and are based on fisher information matrix or likelihood profiling (raue et al., 2009)) of these parameters will be discussed in section 2.3. figure s3 present the occupancy numbers of the six compartments of the model when one parameter is varied while the others are fixed. because the epidemics start date and state is different in each region, it is necessary to estimate in priority e 0 and a 0 . b, d e and d i have a similar impact on the simulations, and we chose to prioritize the estimation of the transmission b as it is a important actionable driver of the epidemics, that can potentially be reduced by interventions. r and d q have little impact on i (in) but are informative for h (in) . since r can be estimated from other data sources, we prioritized the estimation of d q . in addition, there is a strong rational to estimate b and d q as they are directly involved in the computation of the reproductive number. so in the end, we assume d e , d i , and r fixed and we will estimate e 0 , a 0 , b, and d q . both d e and d i are set to a common value across regions, estimated from previous studies referenced in table 2 . to set r, we used data collected from general practitioners through the re-purposing of the réseau sentinelles network to monitor covid-19 and provide a weekly estimation of the number of incident symptomatic cases (regardless of their confirmation status through a pcr test) at the region level. thus, for each region, r i is set to r s the ratio of observed incident number of ascertained cases over the incident number of symptomatic cases (as defined by the incident symptomatic cases reported by the réseau sentinelles network). values are provided in table 1. 35 . cc-by 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted april 24, 2020. . is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. (which was not certified by peer review) the copyright holder for this preprint this version posted april 24, 2020. . the model ( (1)) is a simplified version of the ode system presented in wang et al. (2020): in the ode system (9), consider the possibility of an exogenous flow of susceptible modeled by a sustained susceptible inflow n. while this make sense for modeling an outbreak in a region surrounded by other regions free of the epidemic (such as the first outbreak of covid-19 in wuhan (hubei, china) in late 2019 -early 2020), it is not suitable for a pandemic where the pathogen is circulating in all regions, which is the current situation in all of france covid-19. nonetheless we study the possible asymptotic steady states of this model, i.e. the constant values s (0) , e (0) , i (0) , r (0) , a (0) , h (0) possibly reached by the system when t → ∞. without loss of generality, we assume the initial number of removed subjects is set to r(t = 0) = 0. by definition, the steady 37 . cc-by 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. (which was not certified by peer review) the copyright holder for this preprint this version posted april 24, 2020 . . https://doi.org/10.1101 states verify: with β 1 = d e /r (1/d q + 1/d i ) and β 2 = d h /d q . in the particular case of n = 0, the fifth equation in (10) can be simplified and imposes a (0) = d i (1 − r)β 1 i (0) /d e which in combination with the third equation leads to i (0) = 0 and thus a (0) = h (0) = e (0) = 0. assuming that a(t) = i(t) = 0 for all times t and applying the population conservation principle we identify (n, 0, 0, 0, 0, 0) as a steady state. since the ode imposes that s is strictly decreasing as long as s(t) > 0 and a(t) > 0 (or i(t) > 0) and that s and a (or i) cannot reach 0 in finite time if a 0 = 0 (or i 0 = 0), this steady state (n, 0, 0, 0, 0, 0) is unstable. having a 0 = 0 (or i 0 = 0) leaves us with s (0) = 0 as the only possible asymptotic steady state value and since conservation principle imposes s 0 + e 0 + i 0 + r 0 + a 0 + h 0 = n , we end up with r (0) = n and we identify s (0) , e (0) , i (0) , r (0) , a (0) , h (0) = (0, 0, 0, n, 0, 0), as the only possible asymptotic steady when a 0 = 0 (or i 0 = 0). but in presence of an inflow of susceptible n = 0, can we expect an asymptotic behavior with extinction of the epidemic. if we set a (0) = i (0) = 0 in (10), the first and fourth equations give us: hence the potential steady state corresponding to the extinction imposes s (0) = n , r (0) = 0 and (0, 0, 0, n, 0, 0) is no longer a possible steady state point. the only steady state (n, 0, 0, 0, 0, 0) corresponds to the case where no one gets infected nor removed i.e when the epidemic does not start at all. . cc-by 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. (which was not certified by peer review) the copyright holder for this preprint this version posted april 24, 2020. . in this new setting, a (0) , s (0) and r (0) are now functions of i (0) and given by: with β 4 = (1 − r)β 1 d i /d e . from algebraic manipulation of (10), we derive that a necessary and sufficient condition for i (0) to constitute a steady set is to be a positive real solution of the fourth-order polynomial equation: bd e s (0) (nd i + (αβ 4 + 1) (n − (1 + β 2 )i (0) ))(n − (1 + β 2 )i (0) ) −β 1 n (nd e + n − (1 + β 2 )i (0) ) nd i + n − (1 + β 2 )i (0) = 0 39 . cc-by 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. (which was not certified by peer review) the copyright holder for this preprint this version posted april 24, 2020. . sars-cov-2 vaccines: status report. immunity daisy: a new software tool to test global identifiability of biological and physiological systems the french syndromic surveillance system sursaud r convergence of a stochastic approximation version of the em algorithm report #8: expected impact of school closure and telework to mitigate covid-19 epidemic in france report #9: expected impact of lockdown in île-de-france and possible exit strategies statistique annuelle des établissements de santé transmission dynamics of the covid-19 outbreak and effectiveness of government interventions: a data-driven analysis modelling the impact of covid-19 upon intensive care services in new south wales isolated sudden onset anosmia in covid-19 infection critical care utilization for the covid-19 outbreak in lombardy, italy: early experience and forecast during an emergency response covid-19: a remote assessment in primary care the effect of human mobility maximum likelihood estimation in nonlinear mixed effects models the incubation period of coronavirus disease 2019 (covid-19) from publicly reported confirmed cases: estimation and application early transmission dynamics in wuhan, china, of novel coronavirus-infected pneumonia substantial undocumented infection facilitates the rapid dissemination of novel coronavirus (sars-cov2) monolix v2019r2 developing covid-19 vaccines at pandemic speed predicting the number of reported and unreported cases for the covid-19 epidemic in south korea, italy, france and germany. medrxiv. impact of non-pharmaceutical interventions on the outbreak of coronavirus disease the role of age distribution and family structure on covid-19 dynamics: a preliminary modeling assessment for hubei and lombardy coronavirus disease 2019 (covid-19) situation report -84 naming the coronavirus disease (covid-19) and the virus that causes it who director-general's opening remarks at the media briefing on covid-19 -11 characteristics of and important lessons from the coronavirus disease 2019 (covid-19) outbreak in china: summary of a report of 72,314 cases from the chinese center for disease control and prevention a novel coronavirus from patients with pneumonia in china the authors thank romain greffier for his time in discussing the aggregated features of covid-19 patients care at the bordeaux university hospital. the authors thank the opencovid-19 initiative for their contribution in opening the data used in this article. bph thanks vincent pey for discussions about the clinical characteristics of the covid-19 infection. this work is supported in part by the gestepid inria mission covid19. the authors have no competing interests to declare. mp, lw, rt and bph designed the study. mp and bph analyzed the data. mp, dd and bph implemented the software code. qc performed identifiability and asymptotic analysis of the model. mp, lw, rt and bph interpreted the results and wrote the manuscript. one of the difficulties in modeling the covid-19 epidemic in real-time is the reliability and comparability of data sources an reporting artifacts.differences bewteen the sursaud r database and santé publique france reports the santé publique france epidemiological reports cite si-vic as their source. si-vic is a special system that can be activated for the identification and monitoring of victims from terror attacks and exceptional health situations, and that was activated on march 13 th 2020 for the covid-19 epidemic and is maintained by the french agence du numérique en santé. because si-vic is a declarative web-based plateform (with data populated by regional health agencies, emergency services and hospitals) it was likely missing some case declaration when first launched. on the contrary, the sursaud r database. the figure s1 illustrate the discrepencies and differences between the two database. key: cord-267867-q52nvn0n authors: chevalier, christophe; saulnier, aure; benureau, yann; fléchet, dorian; delgrange, david; colbère-garapin, florence; wychowski, czeslaw; martin, annette title: inhibition of hepatitis c virus infection in cell culture by small interfering rnas date: 2016-12-14 journal: mol ther doi: 10.1038/sj.mt.6300186 sha: doc_id: 267867 cord_uid: q52nvn0n hepatitis c virus (hcv) infection is a major cause of chronic liver disease and hepatocellular carcinoma, yet fully efficacious treatments are missing. in this study, we investigated rna interference (rnai), a specific gene silencing process mediated by small interfering rna (sirna) duplexes, as an antiviral strategy against hcv. synthetic sirnas were designed to target conserved sequences of the hcv 5′ nontranslated region (ntr) located in a functional, stem–loop structured domain of the hcv internal ribosome entry site (ires), which is crucial for initiation of polyprotein translation. several sirnas dramatically reduced or even abrogated the replication of selectable subgenomic hcv replicons upon cotransfection of human hepatoma cells with viral target and sirnas, or upon transfection of cells supporting autonomous replication of hcv replicon with sirnas. importantly, three sirnas also proved capable of strongly inhibiting virus production in cell culture. one sirna, targeting a sequence that is highly conserved across all genotypes and forms a critical pseudoknot structure involved in translation, was identified as the most promising therapeutic candidate. these results indicate that the hcv life cycle can be efficiently blocked by using properly-designed sirnas that target functionally important, highly conserved sequences of the hcv ires. this finding offers a novel approach towards developing ires-based antiviral treatment for chronic hcv infections. hepatitis c virus (hcv) infection is a major cause of chronic liver disease and hepatocellular carcinoma, yet fully efficacious treatments are missing. in this study, we investigated rna interference (rnai), a specific gene silencing process mediated by small interfering rna (sirna) duplexes, as an antiviral strategy against hcv. synthetic sirnas were designed to target conserved sequences of the hcv 5 nontranslated region (ntr) located in a functional, stem-loop structured domain of the hcv internal ribosome entry site (ires), which is crucial for initiation of polyprotein translation. several sirnas dramatically reduced or even abrogated the replication of selectable subgenomic hcv replicons upon cotransfection of human hepatoma cells with viral target and sirnas, or upon transfection of cells supporting autonomous replication of hcv replicon with sirnas. importantly, three sirnas also proved capable of strongly inhibiting virus production in cell culture. one sirna, targeting a sequence that is highly conserved across all genotypes and forms a critical pseudoknot structure involved in translation, was identified as the most promising therapeutic candidate. these results indicate that the hcv life cycle can be efficiently blocked by using properly-designed sirnas that target functionally important, highly conserved sequences of the hcv ires. this finding offers a novel approach towards developing iresbased antiviral treatment for chronic hcv infections. hepatitis c virus (hcv) frequently establishes persistent infections in the liver, leading to the development of chronic hepatitis, and, potentially, liver cirrhosis and hepatocellular carcinoma at later stages. chronic hcv infection affects 2.2% of the world's population and is known to be the leading factor necessitating liver transplantation in patients in developed countries. no vaccine is available for hcv and the current treatment, which consists of administering pegylated interferon α and ribavirin, has limited efficacy against certain hcv genotypes, and also produces significant adverse effects (reviewed in ref. 1) . the development of alternative, specific therapies for chronic hcv infection is therefore a major public health objective. hcv, a member of the flaviviridae family, contains a singlestranded positive-sense rna genome that encodes a unique precursor polyprotein; this polyprotein is further processed into structural proteins and nonstructural proteins, thereby ensuring genome replication. the long open reading frame is flanked by 5 and 3 nontranslated regions (ntrs) that are highly conserved among the majority of hcv genotypes and contain elements that are essential for genome replication (reviewed in ref. 1 ). in addition, hcv 5 ntr contains a highly structured element, namely, an internal ribosome entry site (ires), which is essential for the initiation of hcv polyprotein translation. 2 anti-hcv drug development has been hampered both by the lack of efficient cell culture systems that support virus replication and by the unavailability of accessible animal models. hcv subgenomic rna replicon systems 3 have permitted the assaying of the inhibitory effect of antiviral candidates on hcv genome replication in vitro in human hepatoma cells. the recent development of stable cell culture systems permitting robust production of infectious hcv particles in vitro, based on the jfh-1 strain of hcv genotype 2a, 4 will facilitate the investigation and testing of new antiviral strategies. in addition, alternative animal models that have recently been developed seem to show promise in evaluating candidate antiviral therapeutics. these animal models include transgenic mice engrafted with human hepatocytes, 5 as well as nonendangered primate species (tamarins, marmosets) infected by gb virus b, a hepatotropic virus that is closely related to hcv, 6, 7 or by gb virus b derivatives containing functional hcv sequences. 8 among possible therapeutic strategies, rna interference (rnai) is an attractive path to explore. first described as a natural defense mechanism against plant viruses, rnai was subsequently shown to be a universal phenomenon of post-transcriptional gene silencing, which is initiated by double-stranded rna and leads to specific degradation of homologous rnas. this process involves the generation of 21-nucleotide small interfering rnas (sirnas) which, in association with a multiprotein complex named "rna induced silencing complex", are used as guides to target specific rna substrates by watson-crick base-pairing. such sirnas, when introduced directly into mammalian cells or expressed from viral vectors, can lead to the degradation of targeted sequences (reviewed in ref. 9 ). during the past few years, rnai has been extensively investigated as an alternative specific therapy to treat cancers, genetic diseases, and infections by various human pathogens of medical importance, using both in vitro and in vivo model systems (reviewed in refs. 10, 11) . rna viruses, in particular viruses with liver tropism such as hcv, are ideal candidates for nucleic acid-based therapies that have been shown to efficiently target the liver (reviewed in ref. 12) . recently, various studies have reported variable inhibitory effects of hcv-specific synthetic sir-nas or small hairpin rnas (shrnas) expressed from viral vectors, that target sequences encoding various hcv nonstructural proteins as well as sequences within the 5 ntr. [13] [14] [15] [16] [17] [18] [19] these studies all relied on the use of hcv subgenomic or genomic replicon models in cell culture. in this study we designed sirnas targeting highly conserved sequences within the hcv 5 ntr. the efficiency of these hcvspecific sirnas was first evaluated using previously described self-replicating hcv rnas. 20, 21 this allowed us to select four sir-nas that abrogated or substantially reduced genome replication. by using these sirnas in recently described cell culture systems that support the production of genotype 2a jfh-1 4 in order to investigate the potency of virus-specific sirnas in inhibiting hcv infection, we selected seven sirnas that target conserved sequences of the pivotal domain (domain iii) of the hcv ires within the 5 ntr 2 (figure 1a) . this domain is particularly important for initiation of polyprotein translation, as it directly contacts 40s ribosomal subunits 22 and binds translation initiation factor 3, eif3. 23 domain iii has also been reported to be involved to some extent in genome replication. 24 sirnas were designed essentially according to previously described criteria; in particular, whenever possible, the asymmetric thermostability of sirna duplexes was respected. 25, 26 as a messenger, hcv positive strand rna genome is an ideal target for sir-nas, but it harbors strong secondary and tertiary structures, in particular within the ires. this makes it difficult to design optimal sirnas according to the criteria referred to earlier. on the other hand, the high nucleotide conservation of this region the footprints of the seven hcv-specific sirnas selected to target the internal ribosome entry site (ires) are represented on the schematic structure of domains iii-iv of the 5 nontranslated region (ntr) from the h77 strain of hcv genotype 1a. 2 sirna si240 targets a viral sequence that contains a nucleotide change in hcv genotype 1a and 1b sequences, as indicated. (b) the two replicons used in this study, ntat2aneo 21 and nneo 20 with 5 ntrs derived from hcv genotype 1a and 1b strains, respectively, are schematically represented, with cell culture adaptive mutations (s2005i in ns5a and r2889g in ns5b, respectively) indicated by arrowheads. both replicons encode neomycine phosphotransferase (neo) c-terminally fused to either human immunodeficiency virus tat protein followed by foot-and-mouth disease virus autocatalytically cleaved 2a protein (ntat2aneo), or to a few amino acid residues from core protein ( c in nneo). permitted the identification of sirnas that were homologous to several hcv genotypes ( table 1) . selected sirnas were named according to the nucleotide position within the genome of the h77 strain of hcv genotype 1a 27 that corresponds to the first 5 nucleotide targeted by the sirna antisense strand: si205, si214, si240-1a, si240-1b, si244, si284, si313 ( table 1) . sirna si313 targets the 3 end of domain iii and 5 end of domain iv involved in a pseudo-knot located upstream from the initiator codon (figure 1a) . the sequences of all sirnas, with the exception of si240, matched both hcv genotype sequences 1a and 1b that were present in the subgenomic replicons used in the first part of the study (figure 1b) . two sirnas si240 (si240-1a and si240-1b) were synthesized with sequences homologous to genotypes 1a and 1b, respectively. when targeted to genotype 1b 5 ntr, si240-1a formed a g:u wobble base-pairing at position 18 of the sirna antisense strand, whereas si240-1b formed an a:c mismatch with the genotype 1a replicon (figure 1c ). the g:u non-canonical base-pairing is known to be non-disruptive in double-stranded rna structures and both g:u wobble basepairing and a:c mismatch, if present at certain positions of the sirna, were recently reported to be well tolerated for sirnamediated gene silencing. [28] [29] [30] [31] therefore, we proceeded to analyze the effects of si240-1a and si240-1b on both homologous and heterologous replicon targets. finally, as a non-specific, negative control in these experiments, we used an irrelevant sirna, referred to as siirr ( table 1) , that had previously been used for targeting a sequence within the 5 ntr of a poliovirus strain, 30 and that did not share homology with the hcv sequence. first we evaluated whether selected sirnas were able to inhibit the replication of an hcv subgenomic replicon (ntat2aneo) in cell culture. we did this by using a previously described system that allowed simple monitoring of rna replication by measurement of an enzymatic activity, namely, secreted alkaline phosphatase (seap) activity 21 (see materials and methods and figure 1b) . all hcv sequences of ntat2aneo replicon are derived from the n strain of genotype 1b, with the exception of the 5 ntr, which is derived from the h77 strain of genotype 1a. following coelectroporation of en5-3 cells with 5 μg of in vitro-transcribed ntat2aneo rna and 2 μg of each synthetic sirna, the culture supernatant was replaced daily with fresh medium during the first 4 days after transfection and then on day 7, and seap activity was measured in all culture supernatants collected. during the first 2 days, seap signals essentially reflected translation of the input rnas, since seap levels were roughly similar in cells transfected with ntat2aneo replicon and those transfected with a replication-defective rna encoding an inactive ns5b rna polymerase deleted in its active site (δgdd, see materials and methods). rna replication was then clearly detected between days 3 and 7 after transfection (figure 2a , compare seap levels in cells transfected with ntat2aneo and δgdd). cotransfection of the irrelevant sirna (siirr) did not significantly affect ntat2aneo replication (figure 2a) . hcvspecific sirnas si214 and si284 showed only transient and nil effect on hcv replication, respectively, and seap expression levels on day 7 after transfection were similar to those induced by transfection of ntat2aneo alone (figure 2a) . these two sirnas were therefore not further utilized in the course of our study. in sharp contrast, sirnas si205, si240-1a, si240-1b, si244, and si313 proved capable of abolishing hcv replication in this system, yielding seap levels as low as those obtained with replication-defective rna δgdd. in the presence of these sirnas, seap levels were even slightly lower at days 1, 2, and 3 after transfection than those obtained with δgdd, as a result of sirna-mediated degradation of input rnas, rendering them unavailable for translation (figure 2a) . the inhibitory effect of these five sirnas on hcv replication was shown to be prolonged to 14 days after transfection (data not shown). we next monitored dose-response effects of each of the five selected sirnas using decreasing sirna doses in the range of 2,000-4 ng (67-0.13 nmol/l per μg replicon) in coelectroporation experiments with 5 μg of ntat2aneo replicon (figure 2b ). from these experiments, inhibition doses 50 were calculated for each sirna (figure 2b) . the sirnas si240-1a and si244 proved to be the most efficient inhibitors of hcv rna replication with inhibition doses 50 of 19 ng (0.64 nmol/l) and 33 ng (1.11 nmol/l), respectively, and all sirnas exhibited inhibition doses 50 within the 19-141 ng range. the inhibition dose 50 of si240-1b (127 ng) was shown to be substantially higher than that of si240-1a (19 ng), in agreement with the fact that si240-1b is not fully homologous to ntat2aneo 5 ntr (genotype 1a). this demonstrates the sequence-specificity of these sirnas. we also found that the co-administration of two sirnas targeting nonoverlapping hcv sequences, in suboptimal doses, resulted in additive inhibitory effects (data not shown). this might prove useful in future studies in which a strategy based on combined sirnas is sought to be developed in order to prevent the occurrence of escape mutations. the five sirnas (si205, si240-1a, si240-1b, si244, and si313) that specifically target domain iii of the hcv ires and strongly inhibit subgenomic rna replicon replication when used in the nanomolar range were retained in the remaining part of the study. next, in order to determine whether sirnas are capable of curing replicon-containing cells, we investigated whether hcv replicon rna was eliminated in sirna-treated cells in which seap activities were at basal level. total rna was extracted from cells cotransfected with replicon and sirna on day 10 after transfection and analyzed for hcv rna. this was done by semi-quantitative reverse transcription-polymerase chain reaction (rt-pcr) using an hcv primer pair that allows amplification of a 1630 base-pair fragment spanning the ns3 coding region, as well as a primer pair that allows the detection of a housekeeping messenger rna (glyceraldehyde-3-phosphate dehydrogenase) for the purpose of rna quantity normalization (figure 2c ). hcv rna was readily detected in similar abundance in cells transfected with ntat2aneo or cotransfected with ntat2aneo and siirr (figure 2c , lanes 4, 5). very low levels of residual transfected δgdd rna molecules could be occasionally detected under these experimental conditions (figure 2c, lane 3) . in duplicate experiments, using cells cotransfected with ntat2aneo and si240-1a, si244, or si313, viral rna was either not detected or detected at very low levels, comparable to the results obtained with δgdd rna (figure 2c , lanes 8-13). hcv rna was consistently detected in cells cotransfected with ntat2aneo and si205 (figure 2c, lanes 6, 7) , but at levels lower than in cells cotransfected with ntat2aneo and siirr. these results correlated well with inhibition levels of reporter seap activity, and confirmed that hcv rna replication is strongly inhibited, if not abolished, in cells treated with 2 μg of the three most efficient sirnas (si240-1a, si244, and si313). in order to work with a cell culture system mimicking an established persistent infection, we analyzed the inhibitory capacity of sirnas in en5-3 cells that stably contain and continuously replicate ntat2aneo replicon. the ongoing replication of ntat2aneo replicon in these cells was reflected by seap levels that were more than tenfold higher than in transient experiments after hcv replicon electroporation. two micrograms of each sirna were electroporated into these cells, and their effect on hcv rna replication was determined at different timepoints. data obtained at day 7 after transfection, corresponding to cumulative seap levels secreted between days 4 and 7, are represented in figure 3 . the seap activity generated by the hcv replicon in the presence of either of the homologous sirnas was reduced by 50% or more relative to that of ntat2aneo in mock-transfected cells. when a higher dose (5 μg) of sirnas was used, seap activity was reduced by 70-93% (data not shown). the sirna-mediated inhibitory effect was not as dramatic as in cotransfection experiments, probably due, in part, to limited transfection efficiency (~70%, data not shown), thereby permitting ongoing rna replication in cells that did not receive sirna in the transfected culture. in addition, viral rnas undergoing replication within replication complexes may be less accessible than transfected viral rnas to sirnas. nonetheless, we demonstrated strong, dose-dependent inhibitory potency of hcv domain iii-specific sirnas in cells supporting continuous hcv rna replication. we next addressed the question of the efficacy of sirnas in inhibiting hcv genome replication in cells placed under selective pressure, and examined whether there was any hcv escape from specific sirna treatment. in these experiments, we used the hcv replicon-free subclone 2-3c of huh7 cells, 32 and either ntat2aneo or nneo replicon. in contrast to ntat2aneo, nneo replicon (see figure 1b and ref. 20) contains a 5 ntr derived from the n strain of genotype 1b and encodes neomycine phosphotransferase gene in the first cistron. 2-3c cells were coelectroporated with either ntat2aneo or nneo replicon and each sirna, then placed under g418 selective pressure. g418resistant cell clones were counted on day 21 after transfection. data from two or three experiments carried out with each replicon were pooled for sirnas homologous to both replicons (si205, si244, si313), and the mean of these data is shown in figure 4 . a mean of 34,000 ± 23,800 resistant cell clones was obtained after transfection of hcv replicon alone. the data are represented with respect to 10,000 g418-resistant clones in each experiment. treatment with each of the hcv-specific sirnas resulted in a substantial reduction in the number of g418-resistant cell clones (figure 4) . sirna si205 reduced the formation of g418-resistant cell clones by ~2 logs, whereas si244 and si313 reduced cell clone formation by 3 logs or more (figure 4a) . this is in good correlation with results obtained in the seap reporter system. both si240-1a and si240-1b were more efficient on their homologous target than on heterologous targets (figure 4b) . the nature and the position of the mismatch between si240 and its target (g:u and a:c for si240-1a on hcv genotype 1b and si240-1b on genotype 1a, respectively, at nucleotide 18 of the sirna antisense strand) are reasonably well tolerated by the rnai machinery, resulting in ~2 log reduction in g418-resistant cell clone formation, an inhibition that was, however, 1-1.5 log less efficient than on homologous targets. these experiments confirmed, as already described, 28, 29, 31 that perfect base-pairing between sirnas and targeted messenger rna is required for most effective silencing. to look for potential emergence of sirna escape mutants, several g418-resistant cell clones, obtained after a single treatment with each sirna, were isolated and independently amplified. the 5 ntrs of the hcv replicon rnas recovered from 3 to 13 cell clones for each sirna were reverse-transcribed and pcr-amplified. the resulting pcr products were subjected to sequencing. no nucleotide substitution was found within or in the vicinity of the sirna-targeted region of the hcv replicon rnas. these data suggest that either a single sirna treatment may not be sufficient to select for hcv replicon escape mutants in this system, or that replication competence does not tolerate nucleotide variation in this genomic region. we sought to determine whether the sirnas si240-1a, si244, and si313, shown to be the most efficient in inhibiting the replication of hcv subgenomic replicons both transiently and under selective pressure, are also capable of efficiently blocking virus infection in cell culture. we utilized the jfh-1 strain of hcv genotype 2a that was shown to infect huh7 and huh7-derived cell lines and produce infectious particles. 4 this in vitro infectious system therefore recapitulates the entire hcv life cycle. sequences of si240-1a and si313 were perfectly homologous to the 5 ntr sequence of this genotype 2a strain. in contrast, si244 was shown to hybridize to the jfh-1 5 ntr with a c:a mismatch at the third position of the sirna antisense strand (see figure 1c) . such a mismatch at the 5 end of the antisense strand is expected to hamper rnai. 29, 33 for this reason, we also used an engineered chimeric virus that contains 5 ntr and core-coding sequences derived from genotype 1a within the backbone of the jfh-1 genome (jfh-1/c(+) 6 (figure 5a) . in contrast, hcv-specific sirna si313 considerably reduced the number of cells infected with either jfh-1 or the chimeric virus (figure 5a) . similarly, si244 efficiently inhibited infection with the 1a/2a chimeric virus (figure 5a, left) , but had only a weak effect, if any, on jfh-1 infection (figure 5a , right), consistent with the existing mismatch between si244 and genotype 2a 5 ntr (see figure 1c) . interestingly, si240-1a sirna also appeared to have a higher inhibitory effect on the infection with the chimeric virus (figure 5a , left) than with jfh-1 (figure 5a, right) , in spite of perfect homology with the 5 ntr sequences of both viruses. we hypothesized that these differential effects of si240-1a might be linked to delayed replication kinetics of the chimeric 1a/2a virus, as compared to jfh-1 (d. delgrange, a. pillez, l. cocquerel, g. paranhos-baccala, y. rouillé, j. dubuisson et al., unpublished results), a phenotype that may impact on target/sirna ratios at early time-points following infection. the effects of the three sirnas on both viruses were also monitored by western blot analysis of e2 expression in infected cells (figure 5b) . the data obtained confirmed that si313 is the most efficient sirna, causing complete inhibition of e2 expression in cells infected with either jfh-1 or jfh-1/ c(+)6-1a. in order to verify whether virus production from sirnatreated cells is reduced accordingly, viral particles were titrated in supernatants from cells transfected with each sirna and infected with either jfh-1 or the chimeric virus, both by realtime quantitative rt-pcr (genome equivalents, figure 5c ) and by determination of infectious focus-forming units ( table 2) . for each of the sirnas, quantification of virus production in sirnatreated cultures (figure 5c, table 2 ) was in good agreement with core and e2 intracellular expression levels (figure 5a and b) . in particular, supernatants from si313-treated, hcv-infected cells exhibited a >92% reduction in genome equivalents/ml titer and a >96% reduction in focus-forming units/ml infectious titer, as compared to mock-treated hcv-infected cells. taken together, these data demonstrate that hcv 5 ntr-specific sirnas are capable of strongly inhibiting virus production in cell culture. in exploring rnai as a potential new therapeutic approach against hcv infection, one of the reasons for our choice to target domain iii of the hcv 5 ntr was that domain iii contains well-conserved nucleotide sequences across all hcv genotypes ( table 1) . 34 the high degree of conservation seen in these sequences may be required for preserving virus function, since this 5 ntr domain controls the initiation of polyprotein translation and modulates rna replication. 23, 24 recent nuclear magnetic resonance and electron cryomicroscopy studies of domain iii 22, 35, 36 have helped identify the structure-function relationships of the various subdomains of domain iii. the results of these studies suggest that subdomains iiia/c and iiid, as well as subdomain iiif that forms a pseudoknot structure (figure 1a) , directly contact the 40s subunit body and act synergistically for the proper positioning of the aug codon. in the present study, si240 and si244 (that hybridize to subdomain iiid) and si313 (that essentially hybridizes to the iiif pseudoknot), therefore target regions that are essential for ires structural integrity and functioning. these three sirnas proved able, at low doses in the nanomolar range, to substantially reduce or even abolish hcv subgenomic rna replication, as measured by reporter seap activities and g418-resistant cell clone formation (figures 2 and 4) , and resulted in the elimination of hcv replicons from treated cells (figure 2c) . importantly, we also demonstrated that these sirnas, particularly si313, considerably limit hcv infection in cell culture. we did this using two hcv strains that can be propagated in huh7-derived cells: (i) the jfh-1 strain of hcv genotype 2a, 4 and (ii) a chimeric derivative of jfh-1 carrying 5 ntr and core sequences of hcv genotype 1a (d. delgrange, a. pillez, l. cocquerel, g. paranhos-baccala, y. rouillé, j. dubuisson et al., unpublished results). our data contrast with those obtained by others who concluded that domains ii 17 and iii 17, 19 of hcv 5 ntr are relatively resistant to sirna, because there is reduced accessibility to these domains, given their association with different proteins and factors involved in translation. our data indicate that regions involved in complex secondary and tertiary structures should not be disregarded as targets for rnai. the data also underscore the importance of sirna design. 25, 26 in contrast, two of the sirnas tested in the present study (si214 and si284, targeting domains iiic and iiie, respectively) exhibited poor or only transient inhibitory effect on hcv rna replication (figure 2a) . interestingly, two other studies reported that sirnas having sequences identical to that of si284 showed relative discrepancies in their effects in cells supporting autonomous replication of hcv subgenomic replicon. 13, 19 these conflicting results may be explained by the different strategies used for sirna synthesis and/or delivery. in addition, the sensitivity of the replicon systems used may have an impact on the sirna efficiencies reported. we observed some differences in sirna efficiencies between: (i) transient replication systems in which viral target rna and sirna were co-introduced into cells (figures 2 and 4) , and (ii) stable replication systems in which sirna was introduced in cells supporting ongoing replication of the viral target, whether dealing with a subgenomic replicon (figure 3 ) or genomelength infectious rna (figure 5, table 2 ). two micrograms of sirna did not inhibit viral rna replication in the stable subgenomic replicon system as efficiently as in cotransfection experiments (compare figures 2 and 4) . in addition, si240-1a and si244 caused elimination of viral rna upon cotransfection with subgenomic replicon (figure 2a) , whereas they substantially reduced, but did not abolish the production of virus particles ( figure 5, table 2 ). this was in spite of the fact that the molar ratio of sirna to targeted hcv rna was higher in cells stably containing hcv replicons or infected with virus than in cells cotransfected with replicon and sirna. in another study, high doses (4,000 pmol) of sirnas proved necessary in order to strongly inhibit hcv rna replication in subgenomic replicon-containing cells. 18 we speculate that viral rna templates engaged in the replicase complex and nascent rnas are probably not as accessible to the rnai machinery as transfected viral rnas are. our data can also be explained in part by the fact that a proportion of cells in the culture were infected or supported ongoing viral rna replication but did not receive sirna, given that the efficiency of electroporation is ~70% in en5-3 and 2-3c cells (data not shown). the specificity of hcv sirnas was demontrated by studying their dose-effect responses (figure 2b) , as well as by using si240 to target the replicon 5 ntr from two hcv genotypes (1a and 1b; figures 2-4 ) and si244 to target the 5 ntr of genotype 1a or 2a virus strains ( figure 5, table 2 ). when used with heterologous targets, the sirnas si240-1a and si240-1b exhibited a noncanonical wobble base-pairing and a mismatch at the 3 end (position 18) of the sirna antisense strand, respectively (figure 1c) . for both 1a and 1b replicons, perfect base-pairing was preferred for maximum rnai efficacy, but a g:u wobble and an a:c mismatch were both tolerated (figure 4) . this is consistent with the fact that wobble base-pairing is known to be well tolerated in double-stranded rna helices, providing stability similar to a watson-crick base-pairing. in agreement with our data, wobble base-pairs, especially when located at the 3 end of the sirna antisense strand, were recently suggested to have no disruptive effect on rnai. 28, 29, 33 more surprisingly, but also consistent with our data, it was reported that an a:c mismatch was not deleterious in double-stranded rna structures. 29 contrasting with these tolerated mismatches between sirna and viral target rna, we found that a c:a mismatch at the 5 end of the sirna antisense strand strongly reduces the inhibitory effect of si244 on jfh-1 infection in cell culture ( figure 5, table 2) , and this is consistent with the rnai mechanism. 29, 37 taken together, these data provide strong evidence against the involvement of a spurious interferonmediated mechanism of inhibition in the suppression of viral replication we observed. the sequence of rna viruses, such as hcv, is known to evolve continuously, resulting in the production of many quasispecies, because of the high error rate of rna-dependent rna polymerases with no proof-reading activity. this property allows rna viruses to escape rapidly to a selective pressure such as antiviral treatments, when sequence changes are compatible with virus functions. this holds true for anti-protease and antipolymerase compounds that are currently under development for hcv treatment. 38, 39 however, we did not observe, after a single sirna treatment, the emergence of escape mutations within sirna-targeted genomic regions in rna extracted from several g418-resistant cell clones. in other studies, sirna-resistant viral mutants harboring single nucleotide substitutions or deletions within or at the vicinity of the sirna-targeted sequence were reported to have emerged in cells expressing constitutively an shrna (in the case of human immunodeficiency virus infections), 40, 41 and after repeated treatments with an sirna targeting the polymerase coding sequence (in the case of hcv subgenomic replicons). 16 the lack of escape mutations observed in our study could either be due to the fact that a single sirna was used in the treatment, or to the fact that sirnas target highly conserved, functionally important genomic sequences, in which substitutions would not be compatible with translation/replication competence. additional studies will be needed in order to determine which of these hypotheses holds true. from the point of view of evaluating the potency of sirnabased antiviral strategies to eradicate persistent infections, it was recently shown that sirnas may be used to cure in vitro persistent infections caused by rna viruses such as poliovirus or lymphocytic choriomeningitis virus. 30, 42 for assaying such a therapeutic strategy in animal models, synthetic sirnas need to be further chemically stabilized and formulated to be efficiently delivered. successful utilization of stabilized, lipid-encapsulated sirnas was reported in a mouse model of hepatitis b virus replication. 43 in addition, organ or cell-type specific delivery of sirnas has been achieved through sirna binding to cholesterol 44 or to antibodies directed against cell surface antigens. 45 synthetic shrnas also showed a more prolonged inhibitory effect in murine liver than sirnas did. this was demonstrated upon co-delivery of shrna or sirna targeting domain iv of the hcv ires and plasmid dna encoding a reporter protein placed under the translational control of the hcv ires. 46 alternatively, optimization of sirna delivery may rely on the use of viral vectors that allow continuous synthesis of shrnas in cells, thereby leading to sustained rnai. 31, 47 it is crucial, however, to control intracellular shrna production levels, since it has recently been reported that constant, high expression of shrnas in the liver could be lethal in mouse models. 48 nevertherless, controlled doses of sirna or expression of shrna have already been shown to be efficient and safe in several animal models of viral infection, including in a mouse model of hepatitis b virus infection, 47 a rhesus macaque model of severe acute respiratory syndrome-associated coronavirus infection, 49 and mouse models of west nile virus and japanese encephalitis virus infections. 31 in the present study, we have identified promising anti-hcv sirnas (si313 and, to a lesser extent, si244 and si240) that target highly conserved sequences in domain iii of the 5 ntr and efficiently silence hcv infection in cell culture. it has now become possible to monitor the effect of these sirnas in vivo using a tamarin/marmoset primate model of infection with a chimeric gb virus b-containing hcv domain iii of the hcv ires, that we have previously described. 8 during revision of this manuscript, kanda et al. 50 reported that delivery of an hcv 5 ntr-specific shrna to hepatoma cell lines infected by hcv resulted in the reduction of viral production. hepatoma cell lines, 2-3c, 32 en5-3, 21 and huap are derived from the hepatocarcinoma cell line huh7, and were cultured in dulbecco's modified eagle medium (invitrogen, cergy-pontoise, france) supplemented with 10% fetal calf serum, 100 u/ml penicillin and 100 μg/ml streptomycin. en5-3 cells stably express seap under the control of the human immunodeficiency virus long terminal repeat promoter, 21 and were cultured in the presence of 2 μg/ml blasticidin (invivogen, toulouse, france). en5-3 cells supporting autonomous replication of ntat2aneo hcv replicon were cultured in the presence of 2 μg/ml blasticidin and 0.5 mg/ml geneticin (g418, invitrogen, cergy-pontoise, france). genotype 2a jfh-1 and chimeric jfh-1/c(+)6-1a virus stocks were generated by transfection of huap cells with corresponding in vitro transcribed genomic rnas. plasmid pjfh-1 containing the genome-length complementary dna (cdna) of the jfh-1 isolate of hcv genotype 2a (genbank accession no. ab047639, ref. 4) was generously provided by t. wakita. plasmid pjfh-1/c(+)6-1a was derived from pjfh-1 by substituting nucleotides 154-341 of the 5 ntr, as well as the capsid coding sequence by corresponding sequences of the h77 strain of genotype 1a (d. delgrange, a. pillez, l. cocquerel, g. paranhos-baccala, y. rouillé, j. dubuisson et al., will be described elsewhere). plasmids pjfh-1 and pjfh-1/c(+)6-1a were linearized with xbai and treated with mung bean nuclease prior to in vitro transcription using megascript t7 kit (ambion, courtaboeuf, france). huap cells were electroporated with in vitro transcribed rna, as previously described, 51 and supernatants collected 8-10 days after transfection were used to re-infect naïve huap cells. supernatants collected at 10 days after infection were used as virus stocks and stored at 80 °c. design and synthesis of sirnas. sirnas targeting hcv ires were designed using previously described criteria. 25, 26 the sequence of the sense-strand of each sirna selected is shown in table 1 . sirnas are referred to by the nucleotide position within the genome of the h77 strain of hcv genotype 1a 27 that corresponds to the first nucleotide targeted by the sirna antisense strand. si240 was synthesized with a sequence homologous to hcv subtype 1a or 1b, i.e., containing a u (si240-1a) or a c (si240-1b) residue at position 18 of the sirna antisense strand. an irrelevant sirna, named siirr, targets the 5 ntr of the sabin strain of poliovirus type 3 30 and was used as negative control in this study. the sequences of all sirnas were compared with known genes by using basic local alignment search tool within the genbank database, and no significant homology to other genes was found. the two strands of each sirna were chemically synthesized at the institut pasteur ("plate-forme 7") and annealed at a final concentration of 500 pmol/μl, as previously described. 25 the quality and quantity of hybridized sirnas were analyzed by electrophoresis on nondenaturing 3% agarose gels. hcv subgenomic replicons and in vitro transcriptions. two dicistronic, subgenomic hcv replicons that encode a selectable reporter gene, neomycine phosphotransferase (neo) were used in this study (see figure 1 ). these replicon cdnas, kindly provided by s.m. lemon, are inserted downstream of the t7 rna polymerase promoter and have been previously described as pnneo/3-5b 20 and pntat2aneo. 21 we will refer to these replicons as nneo and ntat2aneo, respectively, in this study. nneo and ntat2aneo each carries a cell culture adaptive mutation, r2889g and s2005i, respectively. both replicon cdnas are derived from the cdna of the n strain of genotype 1b hcv with the exception of the 5 ntr sequence of the ntat2aneo, which is derived from the h77 strain of hcv genotype 1a. ntat2aneo rna encodes human immunodeficiency virus tat protein fused to the 2a protease of foot-and-mouth disease virus, followed by neo. 21 upon transfection of en5-3 cells with ntat2aneo replicon, the expression of tat-2a drives the synthesis of secreted seap. 21 cdnas bearing a 30-nucleotide deletion, including the three codons (gdd) of the active site of the rna polymerase ns5b within the backbone of nneo and ntat2aneo, 20, 21 and referred to as δgdd, were used as replicationdeficient rnas. plasmid dnas were linearized with xbai prior to in vitro transcription using megascript t7 kit. the dna template was removed by treatment with turbo dnase (ambion, courtaboeuf, france) and rnas were purified by phenol-chloroform extractions and precipitated with ethanol. the quality and quantity of replicon rnas were analyzed by electrophoresis on a non-denaturing 0.8% agarose gel and by optical density measurements. cell transfections. en5-3 or 2-3c cells (2 × 10 6 ) were electroporated with 5 μg replicon rna, or coelectroporated with 5 μg replicon rna and various doses of each sirna, in a 4 mm-gap width cuvette by applying one pulse at 240 v, 900 μf (easyject plus, equibio, kent, uk). cells were resuspended in complete medium immediately after the pulse and seeded at various concentrations. en5-3 cells supporting autonomous replication of ntat2aneo replicon were similarly transfected by electroporation with each sirna. en5-3 transfected cells (4 × 10 5 ) were seeded in 6-well plates and supernatants were collected at 1, 2, 3, 4, and 7 days after electroporation and replaced with fresh medium. seap activity was measured in supernatant aliquots with the phospha-light system (applied biosystems/tropix, courtaboeuf, france) using a luminescent substrate according to the manufacturer's recommendations. luminescent signals were read for 1 second using a lumat lb 9507 luminometer (berthold technologies, thoiry, france). to select for g418-resistant cell clones, variable fractions of 2-3c transfected cells were seeded in 100 mmdiameter dishes and supplemented with nneo-δgdd or ntat2aneo-δgdd transfected cells to adjust the final number of cells to 5 × 10 5 cells per dish. twenty-four to forty-eight hours following transfection, cells were placed under the selective pressure of 0.5 mg/ml geneticin (g418, invitrogen, cergy-pontoise, france) for 3 weeks, with the medium being replaced twice a week. g418-resistant cell clones supporting viral rna replication were fixed and stained with a 0.1% crystal violet solution, or selected and expanded under selective pressure for viral rna sequencing. viral rna sequencing. total rna was extracted from expanded sirnaresistant clones with trizol reagent (invitrogen, cergy-pontoise, france). viral rna was reverse transcribed and amplified with hcv specific primers designed to generate pcr products spanning nucleotides 100-451, using the superscript one-step rt-pcr kit with platinum taq (invitrogen, cergy-pontoise, france). sequencing reactions were carried out on uncloned rt-pcr products using big dye terminator version 1.1 kit (applied biosystems, courtaboeuf, france) and analyzed on a abi 3700 capillary dna sequencer (applied biosystems, courtaboeuf, france). en 5-3 transfected cells (2 × 10 5 ) were seeded in 6-well plates (duplicate wells per transfection) and cultured for 4 days, then trypsinized and passaged at a 1:5 dilution in new 6-well plates. on day 10 after transfection, total rna was extracted novel insights into hepatitis c virus replication and persistence internal ribosome entry site-mediated translation in hepatitis c virus replication replication of subgenomic hepatitis c virus rnas in a hepatoma cell line production of infectious hepatitis c virus in tissue culture from a cloned viral genome hepatitis c virus replication in mice with chimeric human livers comparison of tamarins and marmosets as hosts for gbv-b infections and the effect of immunosuppression on duration of viremia chronic hepatitis associated with gb virus b persistence in a tamarin after intrahepatic inoculation of synthetic viral rna a chimeric gb virus b with 5 nontranslated rna sequence from hepatitis c virus causes hepatitis in tamarins the silent revolution: rna interference as basic biology, research tool, and therapeutic silencing viruses by rna interference antiviral rnai therapy: emerging approaches for hitting a moving target therapeutic short hairpin rna expression in the liver: viral targets and vectors small interfering rna-mediated inhibition of hepatitis c virus replication in the human hepatoma cell line huh-7 interference of hepatitis c virus rna replication by short interfering rnas inhibition of intracellular hepatitis c virus replication by synthetic and vector-derived small interfering rnas hepatitis c virus replicons escape rna interference induced by a short interfering rna directed against the ns5b coding region alternative approaches for efficient inhibition of hepatitis c virus rna replication by small interfering rnas clearance of replicating hepatitis c virus replicon rnas in cell culture by small interfering rnas suppression of hepatitis c virus replicon by rna interference directed against the ns3 and ns5b regions of the viral genome selectable subgenomic and genome-length dicistronic rnas derived from an infectious molecular clone of the hcv-n strain of hepatitis c virus replicate efficiently in cultured huh7 cells subgenomic hepatitis c virus replicons inducing expression of a secreted enzymatic reporter protein hepatitis c virus ires rna-induced changes in the conformation of the 40s ribosomal subunit mechanism of ribosome recruitment by hepatitis c ires rna sequences in the 5 nontranslated region of hepatitis c virus required for rna replication analysis of gene function in somatic mammalian cells using small interfering rnas rational sirna design for rna interference transcripts from a single full-length cdna clone of hepatitis c virus are infectious when directly transfected into the liver of a chimpanzee sirnas can function as mirnas a systematic analysis of the silencing effects of an active sirna at all single-nucleotide mismatched target sites complete cure of persistent virus infections by antiviral sirnas a single sirna suppresses fatal encephalitis induced by two different flaviviruses virus-host cell interactions during hepatitis c virus rna replication: impact of polyprotein expression on the cellular transcriptome and cell cycle association with viral rna synthesis risc is a 5 phosphomonoester-producing rna endonuclease genetic diversity and evolution of hepatitis c virus-15 years on structures of two rna domains essential for hepatitis c virus internal ribosome entry site function structure of the hepatitis c virus ires bound to the human 80s ribosome: remodeling of the hcv ires specificity of microrna target selection in translational repression anti-hcv therapies in chimeric scid-alb/upa mice parallel outcomes in human clinical application mutations conferring resistance to a hepatitis c virus (hcv) rna-dependent rna polymerase inhibitor alone or in combination with an hcv serine protease inhibitor in vitro human immunodeficiency virus type 1 escapes from rna interference-mediated inhibition hiv-1 can escape from rna interference by evolving an alternative structure in its rna genome rna interference-mediated virus clearance from cells both acutely and chronically infected with the prototypic arenavirus lymphocytic choriomeningitis virus potent and persistent in vivo anti-hbv activity of chemically modified sirnas therapeutic silencing of an endogenous gene by systemic administration of modified sirnas antibody mediated in vivo delivery of small interfering rnas via cell-surface receptors small hairpin rnas efficiently inhibit hepatitis c ires-mediated gene expression in human tissue culture cells and a mouse model clearance of hepatitis b virus from the liver of transgenic mice by short hairpin rnas fatality in mice due to oversaturation of cellular microrna/short hairpin rna pathways using sirna in prophylactic and therapeutic regimens against sars coronavirus in rhesus macaque small interfering rna targeted to hepatitis c virus 5 nontranslated region exerts potent antiviral effect subcellular localization of hepatitis c virus structural proteins in a cell culture system that efficiently replicates the virus functional analysis of cell surface-expressed hepatitis c virus e2 glycoprotein courtaboeuf, france). five micrograms of total cellular rna were heatdenatured at 65 °c for 5 minutes and used as template for reverse transcription with 50 u of superscript ii reverse transcriptase (invitrogen, cergy-pontoise, france) and 250 ng of random hexanucleotide primers (roche, meylan, france) for 50 minutes at 42 °c. the resulting cdnas were treated with 2 u of rnaseh (invitrogen, cergy-pontoise, france) for 30 minutes at 37 °c, and purified using qiaquick pcr purification kit (qiagen, courtaboeuf, france). one fifth of the cdna was used for programming pcrs with either hcv-specific primers that allowed amplification of a fragment spanning nucleotides 3457-5085 of the n strain of hcv genotype 1b, or primers specific to cellular housekeeping gene glyceraldehyde-3-phosphate dehydrogenase, so as to normalize for total rna content. pcrs were performed using platinum taq dna polymerase (invitrogen, cergy-pontoise, france) under the following cycling conditions: 1 cycle at 94 °c for 3 minutes, followed by 35 cycles of: (i) 30 seconds at 94 °c; (ii) 30 seconds at 55 °c; and (iii) 1.5 minutes at 72 °c. pcr products were analyzed by electrophoresis on 1% agarose gels. huap cells (4 × 10 6 ) were electroporated with 2 μg of sirna, and 5 × 10 4 electroporated cells were seeded on coverslips in 24-well plates and subsequently infected at 16 hours after transfection with jfh-1 or jfh-1/c(+)6-1a virus stocks at ~1 focus forming unit per cell. after 2 hours of incubation at 37 °c, virus inoculum was washed off and cells were fed with culture medium. at 42 hours after infection, cells were fixed and processed for core detection by indirect immunofluorescence, as previously described, 51 using anti-core monoclonal antibody acap27 (kindly provided by j.f. delagneau, bio-rad, marnes-la-coquette, france). cells were counterstained with hoechst dye to enable nuclei to be detected. for titration of infectious viral particles, 5 × 10 4 huap cells were infected with 1:10 dilutions of supernatants from cells transfected with sirnas and infected with hcv. the foci of infected cells were detected by immunofluorescence with anti-core monoclonal antibody at 72 hours after infection and counted to determine titers in focus forming units/ml. huap cells (4 × 10 6 ) seeded in 24-well plates were lyzed in 50 mmol/l tris-hcl (ph 7.5), 150 mmol/l nacl, 5 mmol/l edta, and 0.5% (vol/vol) igepal buffer containing a mixture of protease inhibitors (complete, roche, meylan, france) and processed for e2 or β-actin detection by immunoblot analysis as previously described, 51 using anti-e2 monoclonal antibody 3/11 52 or anti-β-actin monoclonal antibody (sigma-aldrich, saint quentin fallavier, france), respectively. rnas were isolated from cell culture supernatants using qiaamp viral rna kit (qiagen, courtaboeuf, france) and quantified by real-time quantitative rt-pcr using primer pair and probe targeting a sequence spanning nucleotides 130-290 within the hcv 5 ntr : fp 5 -cgggagagc catagtgg-3 ; rp 5 -agtaccacaaggcctttcg-3 ; probe 5 -fam-ctgcggaaccggtgagtacac-tamra-3 . assays were performed using taqman one-step rt-pcr master mix reagents kit and an abi prism 7700 sequence detector instrument (applied biosystems, courtaboeuf, france), according to the manufacturer's instructions. key: cord-355549-6xnjj5h5 authors: cécile, couchoud; florian, bayer; carole, ayav; clémence, béchade; philippe, brunet; françois, chantrel; luc, frimat; roula, galland; maryvonne, hourmant; emmanuelle, laurain; thierry, lobbedez; lucile, mercadal; olivier, moranne title: low incidence of sars-cov-2, risk factors of mortality and the course of illness in the french national cohort of dialysis patients. date: 2020-08-25 journal: kidney int doi: 10.1016/j.kint.2020.07.042 sha: doc_id: 355549 cord_uid: 6xnjj5h5 the aim of this study was to estimate the incidence of covid-19 disease in the french national population of dialysis patients, their course of illness and to identify the risk factors associated with mortality. our study included all patients on dialysis recorded in the french rein registry in april 2020. clinical characteristics at last follow-up and the evolution of covid-19 illness severity over time were recorded for diagnosed cases (either suspicious clinical symptoms, characteristic signs on the chest scan or a positive reverse transcription polymerase chain reaction) for sars-cov-2. a total of 1,621 infected patients were reported on the rein registry from march 16th, 2020 to may 4th, 2020. of these, 344 died. the prevalence of covid-19 patients varied from less than 1% to 10% between regions. the probability of being a case was higher in males, patients with diabetes, those in need of assistance for transfer or treated at a self-care unit. dialysis at home was associated with a lower probability of being infected as was being a smoker, a former smoker, having an active malignancy, or peripheral vascular disease. mortality in diagnosed cases (21%) was associated with the same causes as in the general population. higher age, hypoalbuminemia and the presence of an ischemic heart disease were statistically independently associated with a higher risk of death. being treated at a selfcare unit was associated with a lower risk. thus, our study showed a relatively low frequency of covid-19 among dialysis patients contrary to what might have been assumed. 1-rein registry, agence de la biomédecine, saint-denis la plaine, france 2-chru-nancy, inserm, cic, epidémiologie clinique, nancy, france 3-nephrology department, caen university hospital, france 4-nephrology department, aphm university hospital, marseille, france 5-nephrology department, ghr mulhouse sud-alsace, france 6-university of lorraine, chru-nancy, vandoeuvre, france 7-calydial, vienne, france 8-nephrology department, nantes university hospital, france 9-nephrology department, ap-hp pitié-salpêtrière hospital, paris, france 10-nephrology-dialysis-apheresis department, nîmes university hospital, france introduction due to their frequent contact with hospitals and their comorbid condition, dialysis patients are identified as high-risk patients for severe forms of infection from sars-cov-2. guidelines to mitigating risks have been published (1) (2) (3) (4) (5) (6) (7) . however, few studies including case reports or the experience of centres have included sufficient numbers of patients to have a complete overview of their real risk and course of illness (8) (9) (10) (11) (12) (13) (14) (15) (16) . in those studies, case fatality varied from 14% to 31%. on march 16 th , 2020, the french national end-stage kidney disease rein registry began to record all patients on dialysis in france who were diagnosed with covid-19. the aim of this first report from the french rein registry is to describe the population of infected dialysis patients and their course of illness, estimate the incidence and lethality of covid-19 disease and identify the risk factors associated with the probability of death. from march 16 th , 2020 to may 4 th , 2020, 1 621 patients were declared as being infected with sars-cov-2 on the rein registry. this represents 3.3% of all 48 669 dialysis patients treated in 1245 dialysis units in metropolitan france and overseas territories. the clinical and care situation at the first report in the registry was "hospitalized -moderate disease" for 48%, "mild disease treated at home" for 39%, "severe disease in an intensive care unit" for 5%, "death" for 2% and asymptomatic for 2% of cases. the first diagnosis was made in 73% of cases with a pcr on a nasopharyngeal swab, 17% on characteristic signs on the ct scan and 8% on suspicious clinical symptoms. finally, a positive pcr was available for 1269 patients (79%). in all, 38% were treated at home. outpatients were younger (median age 68.7, iqr 56.7-80.4, vs 73.7, iqr 63.7-81.6), more often non-smokers and had less dysrhythmia and incapacity for transfer (suppl table1). their mortality was lower (8.5%) compared to patient who were hospitalised (22.4%). in all, 9% of patients were admitted to an icu unit. those patients were younger than the others (median age 67.2, iqr 58.3-74.5, vs 72.4, iqr 61.3-81.6), less often had cerebrovascular disease, had a higher bmi and were less often treated by hospital-based hd (suppl table2). among the 87 patients for whom information was available, 51% received invasive mechanical ventilation (suppl figure 1). the mortality of icu patients was higher (34%) compared to patients who were not admitted to icus (15.5%). the clinical situation at the last report in the registry for patients who were still alive, was "hospitalized -moderate disease" for 11%, "mild disease treated at home" for 16%, "in intensive j o u r n a l p r e -p r o o f care" for 2% and "recovered" for 67% and asymptomatic 4%, with a median follow-up of 19 days (iqr 6-28). not all parts of france were affected in the same way. the prevalence of covid-19 patients varied from less than 1% in the 5 overseas territories and 8 metropolitan regions to over 5% in 3 northeastern regions (especially in alsace, 10%, one of the first french clusters) and in the île-de-france, 9%, the most densely-populated region ( figure 1 ). these variations were not explained by age and were parallel to those of the general population (figure 1 ). at that time, the percentage of infected persons in the french population was 0.2% and the mortality among confirmed cases was 19% (no systematic screening). the cumulative incidence of new cases after an exponential increase has now stabilized itself ( figure 2 the clinical characteristics of infected dialysis and control populations are represented in table 1 . compared to the 25 455 selected controls (treated in centres where at least one patient was infected), the probability of being a case was higher in males (or 1. among the infected patients, 344 died due to a cause related to sars-cov-2 after a median time of 6 days (iqr 3-13). the lethality in diagnosed cases was 21%. in the univariate analysis, higher age, being a former smoker, having a chronic respiratory disease, cardiovascular comorbidities ( e.g. peripheral vascular disease, ischemic heart disease, congestive heart failure or dysrhythmia) and frailty (hypoalbuminemia or inability to walk) were associated with a higher risk of death in sars-cov-2 infected dialysis patients. dialysis in self-care units or out-centres or being a current smoker were associated with a lower risk of death. in fact, most of these clinical characteristics and care j o u r n a l p r e -p r o o f were associated with older age. in the multivariate model, only older age, hypoalbuminemia and the presence of an ischemic heart disease were statistically independently associated with a higher risk of death (table 2 ). being treated in a self-care unit was associated with a lower risk of death. neither chronic respiratory disease, obesity, diabetes nor smoking status were associated with a higher risk of death. the sensitivity analysis including the region of treatment gave similar results. the trajectory of care is represented in figure 3 for the 287 deceased patients for whom at least 2 different clinical situations were reported in the registry. for severe cases hospitalized in intensive care units, the median time until death was 7 days (iqr 4-14), whereas the median time for hospitalized patients until death was 5 days (iqr 3-9) and, for patients at home, 6 days (iqr 3-11). the trajectory of care is represented in figure 4 for the 799 patients who recovered (clinical situation coded as recovery or asymptomatic). the median time in hospital until recovery was 15 days (iqr 10-21), similar to that for patients who were at home (16 days, iqr 11-21). so far, more than 1600 dialysis patients have been diagnosed with sars-cov-2 infection in france. our study shows that the prevalence of sars-cov-2 infection in dialysis patients varied throughout the country from 0 to 10%. mortality in this population of diagnosed cases is high at 21% and is mainly associated with a higher age (13% mortality in patients aged under 75 compared with 30% of patients aged over 75). the trend of the sars-cov-2 epidemic in patients on dialysis shows a parallel development as in the general french population, with north eastern regions and the ile-de-france being more affected. our global prevalence is 3% of dialysis patients but this reaches 10% in the most affected regions. in the absence of other population-based data, it can only be compared with the 14% of the haemodialysis centre in wuhan at the epicentre of the chinese epidemic (12) . however, the nonsystematic detection of asymptomatic patients in france may lead to an underestimation of the true dissemination of sars-cov-2 in the french dialysis population. although the lockdown seemed to have significantly reduced the amount of contact among the general population, dialysis patients have to leave confinement to go to their dialysis units and, consequently, are still in contact with a large number of people. the risk of contamination may occur during transport, at the dialysis unit or during hospitalisation, but also at home with the family or caregivers. home dialysis was associated with a lower probability of being infected suggesting a protective effect of staying at home. dialysis centres affected later learned from units contaminated early on in the epidemic's progression and reorganized their patients' circulation and care (14,17). indeed, since the beginning of the epidemic, protective strategies have been broadcast by the sfndt (société francophone de nephrology dialyse transplantation) with weekly covid-19 webinars inviting all french nephrologists to discuss the overall covid-19 themes and topics available on the sfndt website (https//www.sfndt.org˃actualites) . thanks to this collaboration, the worst may have been avoided. however, we must now remain vigilant and protect our healthcare workers. the initial incidence of the disease in some dialysis units seemed very high, especially in the initial regions. the incidence in dialysis units is now decreasing, mirroring the decrease in the general population. this can also be associated with the implementation of all the necessary preventive actions prone by the sfndt, including 1/wearing a mask during transport and for the entire period of care, 2/systematic tracking of patients and screening at the entrance to dialysis units based on fever and symptoms or contact with an infected person and 3/restricting areas for covid-19 cases (18),. as in the general population, male gender, diabetes and frailty, but not age, were associated with a higher risk of being infected. a selection bias, due to the fact that these patients may have a more severe form of the disease and are therefore more easily diagnosed, cannot be ruled out. as in the cohort of 627 haemodialysis patients at the haemodialysis centre in wuhan, diabetes was associated with a higher risk of infection. this result was still significant when introducing regions in the model to take into account the fact that the epidemic was mainly located in the north east of france where the prevalence of diabetes is higher. smoking, even after taking comorbidities into account, was associated with a lower risk of infection, as discussed in the general population (19) . surprisingly, being treated in a self-care unit was associated with a higher risk of being infected. at self-care units, care is provided without supervision by an on-site nephrologist (16) . the presence of a nurse is mandatory and patients are helped with the hd process. all these units collaborate with a hospitalbased dialysis unit. moreover these units treat younger patients who may have had more contact at risk than elderly. despite the lockdown. these small units, with fewer caregivers on site, could have tarried in implementing protection strategies as proposed by others (1) . further analyses are required to evaluate the impact of other risk factors, such as living in an institution or in a deprived neighbourhood area associated with overcrowded housing. international comparison of case fatalities should be made with caution given the case-mix, the various healthcare arrangements and the different dynamics of the epidemic. our mortality among diagnosed cases, 21% so far, is higher than the 13% reported for the dialysis center in wuhan (12). the older age and more frequent comorbidities of french dialysis patients may explain a higher mortality than in china (20, 21) . furthermore, the non-systematic detection of asymptomatic patients favors more seriously ill patients. in france, case fatality was lower than the 29% reported in j o u r n a l p r e -p r o o f 4 outpatient dialysis facilities in italy (14), the 30% in a single center in madrid (15) , or 31% in a single center in new york (16) . higher mortality in these studies may be explained by a selection bias for more severely ill hospitalized patients. compared to the general population, the dialysis lethality observed in our cohort was similar to the 20% case fatality rate observed with patients aged over 80 in italy (20) . it is also similar to the mortality rate for confirmed cases in the french general population, where at least 84% of the people who died had a comorbid condition and 92% were aged 65 or older. apart from age, which seems to be the major factor in the general population (22), nutritional status, indirectly assessed by albumin levels and the presence of ischemic heart disease, seem to be the main risk factors. further in-depth analyses are planned in order to better estimate the excess mortality in dialysis patients at this period, taking into account the underlying mortality risk. being treated in a self-care units was associated with lower mortality, even after taking into account age and comorbidities. after adjustment, home dialysis mortality did not differ from the mortality rate for hospital-based haemodialysis. however, the small number of patients with home dialysis has not allowed us to make an in-depth analysis so far. other factors (such as living conditions, delay in alerting and other home-based care) which are not available in our registry, need to be explored. although incomplete, the illness trajectory seems to show rapid worsening and a slow healing process. the short lapse of time before death could corroborate the physiopathology with the delay in host inflammatory response phase reported 7 to 10 days after the initial infection (23). this rapid negative development raises the question of reinforced surveillance at home, during dialysis sessions and, why not, preventive hospitalization in a safe environment. our definition of recovery should be taken with precaution since the definition of recovery is still under debate. some patients were maintained in hospital under isolation for 15 days. very soon after the start of the epidemic, the french-speaking society of nephrology, helped by infectious disease specialists made recommendation that for each dialysis patient with fever, a viral syndrome, pulmonary symptoms or diarrhoea, a ct scan should be prescribed as well as a pcr on a nasopharyngeal swab. contact subjects were also tested in the later period explaining the occurrence of few asymptomatic patients. these recommendations applied to all hospitalized patients and outpatients as well throughout the whole country. however, due to possible variations in diagnosis strategies, day 1 for each patient may vary from one unit to another. access to intensive care units was a concern for nephrologists in certain areas. some tensions could be noted in highly affected regions but, in general, dialysis patients could be transferred to intensive care as required depending on their age and comorbidities. the strength of this study is its national scale, including the whole population of french dialysis patients. however, these results must be interpreted bearing the following limitations in mind. various screening strategies may influence the detection of the disease. this could be the case especially for patients treated at home or asymptomatic patients or sudden death, but mild cases and hospitalized patients can be considered as being exhaustive. non-systematic screening favours the collection of more severe cases and leads to an overestimation of lethality. the second limitation is the lack of granular data on clinical presentation, laboratory results and treatment and the precise protective strategy implemented in the units. our study is based on a registry, which gives an exhaustive national overview but with a limited dataset -not on medical records, which could give more detailed data on treatment and clinical presentation but on a limited number of patients with a risk of selection bias. third, due to the confinement of registry research assistants, the data quality control procedure was limited. post hoc controls will be taking place to complete the data. fourth, the total number of patients tested and not considered as covid positive is unknown. as in the general population, the true lethality of covid-19 in infected dialysis patients needs to be confirmed by a longer follow-up and deployment of screening methods. despite the difficulty to have a "true" estimation, this preliminary report of the french registry shows a relatively low frequency of covid-19 among dialysis patients contrary to what might have been feared but, as in the general population, the epidemic did not evenly affect the whole territory. mortality in diagnosed cases (21%) is, associated with the same causes as in the general population, namely, high age, frailty and comorbid conditions. the french rein registry is intended to include all end-stage renal disease (esrd) patients on renal replacement therapy (rrt) living in france, including overseas territories. patients with a diagnosis of acute renal failure were excluded, i.e. those who recovered all or some renal function within 45 days or were considered by experts to have acute failure when they died before 45 days. the details of organizational principles and quality control are described elsewhere (24) . the rein network includes nephrologists, nurses, patients, public health representatives and epidemiologists coordinated within regional and national steering committees. the national coordination center is based at the agence de la biomédecine, a public health agency that oversees the activity of organ and tissue procurement and transplantation. the clinical characteristics at last follow-up included age, gender, comorbidities, mobility status (walks without help, needs assistance for transfers, or is totally dependent for transfers), body mass index (bmi), tobacco use, haemoglobin and serum albumin, dialysis technique (haemodialysis or peritoneal dialysis) and location (hospital-based, out-centre, self-care unit, home). this study analysed 10 comorbidities: diabetes, congestive heart failure, ischemic heart disease, peripheral vascular disease, aortic aneurysm, cerebrovascular disease, dysrhythmia, active malignancy, cirrhosis, and severe behavioural disorders (defined as including dementia, psychosis, or severe neurosis that may have affected the functional status or adherence to treatment). the last residence and last dialysis unit before february 15, 2020 were taken into account to avoid misclassification of patients transferred to another dialysis centre due to their infection status. the clinical characteristics of patients were expressed as frequencies and percentages for qualitative variables and medians with interquartile ranges for quantitative variables. the percentage of infected patients in the dialysis units of each region was adjusted on age (indirect standardization) to take into account the underlying age distribution of the dialysed patients. the crude ratio and the standardised ratio are presented on a map, according to the patients' area of residence. to give an overview of the epidemic in france, hospital mortality due to covid-19 on april 2020, extracted from the platform of the national public health agency, santé publique france: https://geodes.santepubliquefrance.fr/#c=indicator was reported. we also presented the cumulative number of infected patients on a day-to-day graph for the whole country. to describe the characteristics of infected patients, we compared this population with two control groups. the first one included all the dialysis patients in france who were not infected. the second, to take into account the heterogeneity of the distribution of the epidemic in the country, included only patients treated in the dialysis units where at least one infected patient had been declared. risk factors associated with being a case in those units were analysed by logistics regression with a stepwise selection of variables. the final model is based on complete data (no imputation). a p-value of <0.05 (two-sided) was considered statistically significant. results are reported as odds-ratios (ors) with their 95% confidence interval. lethality was estimated from the proportion of deceased patients among the diagnosed cases. to identify the risk factors associated with death in sars-cov-2 dialysis patients, a logistics regression with stepwise selection of variables was used. interactions between age and other factors were explored. a p-value of <0.05 (two-sided) was considered statistically significant. results are reported as odds-ratios (ors) with their 95% confidence interval. sensitivity analyses were made including the region of treatment, either as a fixed effect or with a random intercept. finally, when available, the course of illness was represented on a graph to describe the process of care for patients who died and for those who recovered. for each transition between the various care statuses, the number of patients and the median duration before transfer were calculated. j o u r n a l p r e -p r o o f recommendations for the prevention, mitigation and containment of the emerging mitigating risk of covid-19 in dialysis facilities covid-19 and the inpatient dialysis unit: managing resources during contingency planning pre-crisis consensus recommendations for the care of children receiving chronic dialysis in association with the covid-19 epidemic covid-19 from the nephrologist's point of view pandemic uncertainty: considerations for nephrology nurses how we mitigate and contain covid-19 outbreak in hemodialysis center (hd): lessons and experiences a case of novel coronavirus disease 19 in a chronic hemodialysis patient presenting with gastroenteritis and developing severe pulmonary disease coronavirus disease 2019 (covid-19) pneumonia in a hemodialysis patient first reported nosocomial outbreak of severe acute respiratory syndrome coronavirus 2 (sars-cov-2) in a pediatric dialysis unit peritoneal dialysis during the coronavirus 2019 (covid-19) pandemic: acute inpatient and maintenance outpatient experiences an analysis on the clinical features of mhd patients with coronavirus disease 2019: a single center study deisolation of covid-19-positive hemodialysis patients in the outpatient setting: a singlecenter experience management of patients on dialysis and with kidney transplant during sars-cov-2 (covid-19) pandemic in brescia covid-19: clinical course and outcomes of 36 hemodialysis patients in spain presentation and outcomes of patients with eskd and covid-19 lessons from the experience in wuhan to reduce risk of covid-19 infection in patients undergoing long-term hemodialysis hemodialysis with cohort isolation to prevent secondary transmission during a covid-19 outbreak in korea covid-19 and smoking: a systematic review of the evidence spatial analysis of case-mix and dialysis modality associations case-fatality rate and characteristics of patients dying in relation to covid-19 in italy opensafely: factors associated with covid-19-related hospital death in the linked electronic health records of 17 million adult nhs patients. | medrxiv siddiqi hk, mehra mr. covid-19 illness in native and immunosuppressed states: a clinical-therapeutic staging proposal the renal epidemiology and information network (rein): a new registry for end-stage renal disease in france covid-19 therapeutic trial synopsis we gratefully acknowledge all participants of the rein registry, nephrologists and research assistants alike, especially at this very particular time. the centres participating in the registry are listed in the rein annual report: http://www.agence-biomedecine.fr/le-programme-rein.we also thank teresa sawyers, medical writer at nîmes university hospital for her help in editing the text. the authors of this manuscript declare that they have no competing financial interest and no conflict of interests to disclose. the registry is supported by the agence de la biomedicine, france. key: cord-264203-7dnz9yaa authors: coldefy, magali; curtis, sarah e. title: the geography of institutional psychiatric care in france 1800–2000: historical analysis of the spatial diffusion of specialised facilities for institutional care of mental illness date: 2010-10-10 journal: soc sci med doi: 10.1016/j.socscimed.2010.09.028 sha: doc_id: 264203 cord_uid: 7dnz9yaa as in other european countries, specialised psychiatric hospitals were established throughout france during the 19th century. the construction of these hospitals can be considered as the concrete expression of a therapeutic innovation which recognized insanity as an illness that could be treated in such specialised institutions. the spatial diffusion of these innovative institutions through 19th and 20th century france is analysed and we explore how far this can be understood through theories of diffusion of innovations including geographical models of hierarchical and expansion diffusion (or whether other conceptual models are more appropriate). the research reported here particularly focuses on the period 1800–1961. it involved the construction of an original historical database of both psychiatric hospitals and information on the cities where these institutions were located. this was used to examine and interpret the different phases of development of psychiatric institutions and the parts of the country and types of geographical setting where they were concentrated. a multiple correspondence analysis was then performed to examine the connections between different aspects of the diffusion process. the study shows the limitations of classical models of spatial diffusion, which are found to be consistent with some, but not all aspects of the development of psychiatric institutions in france. an alternative political ecology approach seems more appropriate to conceptualise the various processes involved; national policies, social representations, medicalisation of care of mental illness, and urban and economic growth all seem to be associated with the emergence of a variable and complex pattern. this paper also opens a large field of research. compared with other western countries, the geography of french psychiatric care is relatively under-researched, although there has been a strong spatial dimension to mental health policy in the country. this analysis provides a context for studies of more contemporary processes of french deinstitutionalisation, which is strongly structured by the past heritage of these large asylum facilities. this paper critically explores the relevance of innovation diffusion theories to the geographical development of psychiatric asylums in 19th and 20th century france. since hagerstrand's pathbreaking work in the 1950s, geographers have emphasized the role of spatial structures in processes of innovation diffusion. from various case studies, hagerstand highlighted temporal and spatial regularities in diffusion processes (hägerstrand, 1953) . spatial diffusion of an innovation expresses both the conservation and transformation of geographical spatial structures (saint julien, 1985) . innovation spread is driven by dynamic spatial interaction. two models are classically presented: the hierarchical diffusion model and the contagious diffusion model. the first takes into account the functional hierarchy of settlements. innovation spreads between densely populated urban areas with a high level of interaction and subsequently filters down to smaller, less influential areas. size and rank in the urban spatial system are therefore determining criteria of the hierarchical diffusion process. the second model is based on effects of 'distance decay' and contiguity in the spatial diffusion process and involves 'contagious' spread to areas in close proximity. an innovation will tend to spread within neighbourhoods close to its point of adoption (daudé, 2001) . hagerstrand showed that in most cases, diffusion is achieved through a combination of 'hierarchical transmission' and 'neighbourhood contagion'. empirical observations of innovation diffusion processes have systematically demonstrated that vertical diffusion down through the urban hierarchy has been dominant in a large number of diffusion processes, accompanied by 'horizontal', contagious diffusion around the larger centres (pumain & saint julien, 2001) . innovations first appear in large cities before spreading into the whole urban system. however, despite the rather general relevance of these diffusion models, not every diffusion process can be described solely using these concepts. according to saint julien (1985) , other factors can interact with diffusion flows, such as: chance events, market characteristics independent of the urban hierarchy, effects of the existence of a centralised or decentralised management of the diffusion process or the competitive or noncompetitive nature of the system. in health geography, research on spatial diffusion has mainly focused on the diffusion of infectious diseases, especially nonvectored infectious diseases giving rise to epidemics through human contact (meade & earickson, 2000) . there is a long history of research to describe and predict how epidemics spread geographically, providing information for action to anticipate, treat and perhaps prevent epidemics. since pioneering work in the 18th and 19th centuries (currie, 1792; currie, 1811; snow, 1854; webster, 1799) , the emergence of new infectious diseases at the end of the 20th century has given a new impetus to research in this field. hiv/ aids, for example, was largely studied in the late 1990s (amat-roze & remy, 1990; bastos & barcellos, 1995; dias & nobre, 2001; gould, 1993; kearns, 1996; shannon, 1994; wallace & wallace, 1995; wood et al., 2000) . recent research has also focused on ancient epidemics, like the plague (especially the second pandemic) or influenza (spanish influenza for instance) (anatra, 1987; hunter & young, 1971; lemey, suchard, & rambaut, 2009; merler & ajelli, 2010; sabatini, 1987; smallman-raynor, johnson, & cliff, 2002; tuckel, sassler, maisel, & leykam, 2006) . more recently, numerous studies have focused on the international diffusion of severe acute respiratory syndrome (sars) and h5n1 avian influenza (souris, gonzalez, shanmugasundaram, corvest, & kittayapong, 2010) , which are examples of diseases presenting new challenges to public health in this era of more pronounced globalisation (affonso, andrews, & jeffs, 2004; bowen & laroe, 2006; meng, wang, liu, wu, & zhong, 2005; shannon & willoughby, 2004; smallman-raynor & cliff, 2008; wang, christakos, han, & meng, 2008) . less attention has been paid to the diffusion of medical innovations and new types of care structure within health care delivery systems. however, research of this type can be helpful in formulating and evaluating policies aiming to improve the provision of care, suggesting what factors may help or hinder the dissemination of good practice and how effectively new initiatives are introduced throughout a health system. for example, complex systems such as health services rely on large and expensive infrastructures and on the provision of trained staff that are difficult to move once they are in place, and considerable effort is often required to achieve universal changes in professional practice. services therefore develop in a way that is 'path dependent'; the history of development of a service can influence the potential for new development in the future. investigations of this type include studies of the diffusion of: tomography scanners in the us (baker, 1979) ; abortion facilities in the north-eastern us (henry, 1978) ; the administration of antipsychotic olonzapine to urban and rural children in michigan (penfold & kelleher, 2007) ; alternative chiropractic and naturopathic practices in canada (williams, 2000) , and the international diffusion of yoga (hoyez, 2007) . diffusion of innovations in health policy and health promotion has also been studied from a geographical perspective. shannon, bashshur and metzner (1971) analysed the spatial diffusion of a prepaid group practice health plan and nykiforuk, eyles, and campbell (2008) studied the diffusion of smoke-free spaces in canada using roger's (2003) framework for the diffusion of innovations and classic geographical diffusion models. some geographers have also studied the spatial diffusion of hospitals and (most pertinent here) the evolution of national systems of psychiatric hospitals from an historical perspective. these are interesting for the way that they demonstrate the growth of medical power and influence as well as changes in access to care. they also provide the context for studies of more contemporary processes of deinstitutionalisation that have often retained vestiges of the older health care system, still influencing the way care is provided today. jones (1999) compares implementation and spatial aspects of mental health policy reforms in united kingdom and italy since the 1950s, noting that in italy, the diffusion of reform was spatially uneven. it was more advanced in the industrialised and urbanised north of the country while in the poorer, more rural south, development was retarded and mainly left to the management of voluntary and religious sector organizations (galzigna & terzian, 1980) . jones suggests that in the uk, psychiatric hospitals developing through the larger urban centres eventually led to a more equitable distribution in the national space. for jones, this was due to the strong intervention of the british government in the implementation of a national system of institutions, but the dynamic process of diffusion is not detailed in her paper. in a particularly comprehensive discussion, philo (2004) gives an account of the development of mental asylums in england and wales up to 1860, which suggests that various forces came into play. debate and rivalry among medical professionals were important in the early phases of development. philo also points to developments from the late 18th to the middle of the 19th centuries, when initiatives to locate asylums in what were thought to be more humane and therapeutic settings outside major cities became increasingly influential. it seems that trends depend on national context since a rather contrasting american study (hunter, shannon, & sambrook, 1986) , reports the emergence and diffusion of public 'lunatic asylums' in the united states during the 19th century, demonstrating how over time the establishment of these facilities spread from the north-east to the west of the country. further research conducted by bretagnolle, giraud, and mathian (2008) on american urbanisation allows us to draw a parallel with the diffusion of the railway network, suggesting that in america, the diffusion of institutions for mental health care (as well as other services) followed geographical processes of colonization and social and economic development taking place at the time. the role of railways and transport networks on the spread of disease and health care has been examined by hogbin in south africa during the first part of the 20th century (hogbin, 1985) . it is thus clear that a good deal can be learned from studies of diffusion of mental health care institutions in the 18th, 19th and early 20th centuries. it shows the interdependencies between socio-economic development and health care developments across national spaces. this diffusion of institutional structures is interesting in that it also represents the concrete implementation of ideas about appropriate models of psychiatric care. the emergence and dissemination of an idea concerning psychiatric care is not necessarily perfectly matched by the implementation of the idea through construction of the specialised psychiatric hospitals that are of interest here. in this study we are particularly concerned with the diffusion of this concrete expression of a new care model through the modification of the psychiatric infrastructure, since it is at the point of construction of these new facilities that changes in provision of psychiatric care will have started to have an impact on the care environment for people with mental illness. the innovation diffusion model also raises some interesting issues concerning whether or not there is a specific 'tipping point' in time and space at which an innovation begins to spread, or whether change is influenced by more continuous processes of path-dependency whereby past actions and thought influence present patterns of change. in this paper we contribute to the international discussion concerning the importance of national context in the history of psychiatric care provision by considering the development of psychiatric institutions in france during the study period. the analysis aims to determine the relevance of 'classic' diffusion models in this process (which might suggest psychiatric care development was part of socio-economic growth and development in france, as in america). following philos' and jones' european examples, we also seek to identify other key processes, associated with professional medical influence and governmental health care policy at the time, that also appear to have driven the growth of the system. the lunatic asylum as an innovation in 18th and 19th century france: from the 'alienist' perspective on government policy here we briefly summarise the processes that influenced the diffusion of the 'lunatic asylum' as a model of psychiatric care in france during the 18th and 19th centuries. prior to these developments, no specific medical or health care response was proposed for people with mental disorders. they were placed in institutions for the indigent and criminals. hitherto 'insanity' had not been understood as a treatable illness, so the aim was to restrain people identified as 'mad' and prevent them from disturbing public order, not to try to cure them. foucault (1961, chapter ii) suggests that in france, and particularly in paris, this approach was clearly illustrated in institutions called hôpitaux généraux created in 1656 (imbert, 1982) to implement this policy described by foucault (1988 edition, p. 38e64) as 'the great confinement'. the lamentable conditions of their confinement were already being identified in the 18th century (colombier & doublet, 1785) . at around the start of the 19th century, in france as in other countries, we begin to see the seeds of innovation: insanity began to be interpreted as an illness that could be cared for in specialised institutional settings. this therapeutic innovation was rooted in the emergence of the philanthropic and humanist ideals of the 18th century. these were associated with a shift away from demonological interpretations of madness and the growing pre-eminence of naturalistic explanations. the idea of the curability of mental illness and the legitimacy of the physician's role in the social management and treatment of madness also contributed to the emergence of this innovation gauchet & swain, 1980) . these ideas were promulgated through the 'alienist' school of thought, calling for the separation of 'mad' people into specialised, therapeutic settings as recommended by pinel (1801) in france, and tuke (1813) and browne (1837) in the united kingdom. they were developing the concept of mental 'alienation' (mental illness viewed as a person's inability to integrate in society), arguing that a mental disorder inhibited the sufferer's feelings to such an extent that eventually, both the self and the external world seemed unreal. for the french alienist pinel (1801) , the asylum was the only suitable place for 'moral treatment' requiring the patient's isolation from society as a whole, as well as from other groups who were seen as 'deviant' and dangerous to society. the lunatic asylum thus became the preferred therapeutic instrument of this moral treatment, secluding mentally ill people from the stresses of mainstream society and family life and incarcerating them in a secluded place, ideally situated in tranquil countryside where a strict moral framework was imposed. foucault (1988 edition, 259) argues that '.the asylum becomes, in pinel's hands, an instrument of moral uniformity and social denunciation.'. 'place' has considerable significance in this model; physicians aimed to put the mentally sick in a new situation, removed from places, objects, people and circumstances that shaped their usual relationships and behaviour. at this period, well before the introduction of psychotropic drugs, 'moral treatment', acting on intellect and feelings, also marked a move away from physical treatment by traditional methods of blood-letting and purges applied to the patient's body (goldstein, 1997) . pressure of opinion was building in france in favour of extensive reform and was beginning to be felt by both the government and the medical profession. in a report on institutions for the 'insane' presented to the french interior minister in 1819, the alienist physican esquirol wrote: 'these unfortunate people are treated worse than criminals and reduced to a worse condition than animals'. it was at around this period that esquirol introduced the term 'asylum' to distinguish psychiatric care institutions from both the 'hôpital général' carceral regime and 'hôtel-dieu' hospitals for paupers, since these earlier types of institution were considered oppressive, arbitrary in their treatment of mentally ill people, and likely to exacerbate their condition (lantéri-laura, 2001). esquirol wrote: 'i would like us to give these facilities a specific name which does not bring to mind a painful image; i propose we name them asylums' authors' translation from (esquirol, 1818, p. 26) . the term 'lunatic asylum' was still used as late as 1937 when it was replaced by 'psychiatric hospital'. by then, psychiatry had become an established practice within the medical profession and psychiatric institutions had entered into the clinical domain. the later phase of our study period thus arguably represents the shift to a different model of care associated with a new phase in the diffusion of changing ideas about psychiatric treatment that were expressed in the new facilities built most recently. the governmental response to the 'alienist' model, promoting the asylum as an institutional model, was the 1838 lunacy act. this required that every french département (representing the local administrative tier of national government in france) provide a 'facility dedicated to host and care for lunatics'. promulgated under the july monarchy, the lunacy act continued to influence the provision of care for mental disorders, for over 150 years as it was only revised on june 27th 1990 with the 'act relative to the rights and protection of people hospitalised because of mental disorders and to their hospitalisation conditions'. the lunacy act of 1838 instituted the mandatory provision of mental health care in each administrative département either by the creation of at least one asylum or by contracting with an authorized voluntary hospital to do so. this legislation could therefore be expected to have had a significant impact on the geographical diffusion of this type of institutional structure, albeit that the 'lunatic asylum' was not specified as the model on which these facilities were to be built. individuals that were to be housed in these new facilities were nevertheless described as 'lunatics' rather than 'insane' or 'agitated' which suggests that the law makers were influenced by the 'alienation' paradigm proposed by pinel. while the 1838 lunacy act did not include direct guidelines on the type of site that should be preferred for asylum facilities, psychiatric ideas on the subject had already been clearly expressed in france. the esquirol (1838) thus specified that asylums should be built outside cities for economic and therapeutic reasons. the following quote illustrates alienist ideas that dictated 19th century views of what might constitute (or undermine) a therapeutic setting for care of mental illness: "most lunatic asylums are located in cities, a few in the countryside, in the plains or on the heights. in cities, space is lacking, the sick are excited by the hubbub and the noise of the population; visits are more numerous and more frequent; nurses are more distracted, more inclined to leave, while in countryside, there is more space, the sick enjoy more peace and quiet, can go out for a walk in tranquil surroundings or engage in gardening; they have fewer visitors and finally, there are economic advantages. buildings on a high plateau are more favourably situated but when the plateau is not sufficiently extensive, buildings cannot develop on the same level or be sufficiently spaced out; terraces and steps are then required because of the uneven ground." he specifically cites the example of antiquaille hospital in lyon: "located at mid-altitude on the fourviere mountain, it is built on the ruins of an ancient roman construction. this choice of location was unfortunate. it was impossible to design suitable buildings: yards are too narrow, promenade galleries are missing, the ground is arid, and vegetation cannot improve the view or refresh the air. water is not very abundant whereas it is required in such a house. views are certainly very extensive, but the insane can constantly see their fellow citizens coming and going on the banks of the saone river and in the neighbouring streets. they hear the hubbub of the city; is that not sufficient to provoke feelings of irritation likely to increase and to maintain delirium?" (translated by the authors from esquirol, (1838, p. 463)). in the following analysis we shall treat the establishment of 'asylum' facilities as a 'proxy marker' for the implementation of a major innovation in the care of people with mental disorders in france. these asylums constituted a new type of clinical and therapeutic environment for care of mental illness as conceived by pinel. in the following discussion we use the term 'asylum' to refer to state sponsored, specialised psychiatric hospitals in france that were either established following the 1838 lunacy act, or preexisting facilities, including voluntary or religious institutions, recognized by the government as meeting the requirements of the act. other institutions providing mental health care (in multispecialty general hospitals or independent institutions not recognized by the state) are not included in this category, although their contribution in the general context of care provision is taken into consideration in our analysis. in this study, 'asylum' therefore refers to an administrative category of residential institution. these institutions did not all systematically incorporate every aspect of pinel or tuke's asylum model of care, and it is likely that the care provided over the period covered, varied from one institution to the next. however, one aspect of asylum design does become apparent in this analysis; the preference for a rural or semi-rural location as an ideal site. our analysis indicates that this had a significant and lasting influence on the geographical development of psychiatric care in france and contributed to the specific geographical pattern of diffusion of asylum facilities around the country, as will be discussed below. our focus on the establishment of institutions corresponds to the schumpeterian definition (schumpeter, 1912 (schumpeter, , 1939 of an innovation, which is distinct from an invention as it describes the process by which a new idea is effectively adopted by society (dortier, 2004) . the lunatic asylum can also be considered as an 'institutional innovation' according to the pederson's (1970) definition, because it does not directly apply to individuals or households (as in 'individualistic' or 'domestic' innovation), but involves the introduction of a collective service. this is underlined by the way the innovation was not left to develop randomly, or under the sole influence of the medical profession. government legislation was introduced as a means of organizing and centrally coordinating the even spread of asylums to every part of the country. using the functionalist perspective proposed by brown (1981), we consider to what extent the diffusion of 'lunatic asylums' in france corresponded to a 'decentralised process' (spreading autonomously throughout the national space) or a 'centralised process' propagated under the control of a national agency or policy, which determines diffusion conditions (daudé, 2001) . centrally managed innovation diffusion may follow different time space paths than individualistic or decentralised processes. in this case, the adopter of the innovation was central government, aiming to influence the process of innovation through local administrative and geographical levels of government throughout the country. the government of the day was keen to demonstrate the effectiveness of this recently created government structure, inspired by the egalitarian and republican goals of the french revolution (1790). in 1838, legally assigned with new powers in terms of resources, broader responsibilities and greater facilities, french départements provided a conduit for central power to all parts of france, ensuring the management of national space in line with central government policy (burguière & revel, 1989) . this was paralleled by increasing spatial accessibility of most parts of the country, particularly in the first part of the 19th century with the expansion of the railway network (suggesting interesting potential parallels with bretagnolle's study mentioned in the introduction). these processes might have been expected to encourage homogenization and evenness in social and economic development across all french départements, though they might also have tended to encourage early adoption of the new model of psychiatric hospital in the geographical centres of central governmental control in paris, the capital city, and in provincial centres of government. this review of the processes influencing asylum diffusion through french national space suggests it can be viewed as an example of an innovation diffusion process in which the original innovation took place through an informal network of reformers, (which might have produced rather randomly distributed sites for the very first asylums), but that after 1838, the leading adopter was a collective (state) agent, operating through a highly structured geographical and administrative hierarchy. the following analysis explores how these processes influenced the diffusion of asylums in 19th and 20th century france. the state hierarchy was strongly centred in the capital city and its regional seats of government, and had the potential to control the pattern of spreading the innovation through the national space. this could lead one to expect an even, more or less simultaneous geographical diffusion of asylum facilities designed to ensure provision in each département. in many other cases of innovation diffusion, the largest urban centres are most likely to be the sites for early adoption. however, in this case, the diffusion phase dominated by the 'alienist' model of care could be expected to result in the early establishment of asylums in rural or semi-rural settings close to major towns and more particularly, in the proximity of regional administrative centres. to investigate spatial diffusion of psychiatric hospitals in france from the 19th century to the present day, the initial task involved building an original historical database of psychiatric hospitals, their location and date of establishment, indicating the points at which, in different parts of france, asylum facilities were first adopted as innovative care institutions for the mentally ill. this was achieved using a number of different data sources. these data were then analysed and interpreted in the light of the conceptual frameworks and the historical context discussed above. data from the national register of health and social facilities (fichier national des etablissements sanitaires et sociaux, finess) were employed. this is based on information provided by local agencies of the ministry of health and social affairs. created in 1979, the finess inventory made it possible to precisely locate existing facilities and the date on which establishments set up since 1979 became operational. however, it does not allow us to reconstitute the history of hospital development prior to 1979; hospitals which closed before 1979 do not appear in the register, and the date of establishment for older facilities is not included. this inventory is therefore not sufficient for our purpose but is useful to supplement and consolidate historical information from other sources. archival data were used for the earlier period. the french national statistics service (la statistique générale en france sgf) published data on asylums from 1835 to 1942. in the introduction to the volume covering the period 1854e1860, the minister of agriculture, commerce and public works indicates to 'his majesty the emperor' that 'this work not only allows us to appreciate the administrative situation of our asylums and its degree of development; it also contains a certain amount of strictly medical information, which appears to be helpful for the very delicate and difficult study of one the cruellest human infirmities' (translated from statistique de la france, 1865, p. 10). the format of this publication was modified over time. while finess was produced as a register in list format, the sgf provided more comprehensive statistical data on hospitals presented by département and by year. these data allow us to correctly date the creation of asylums established between 1835 and 1942. complementary information was sought in historical studies on french psychiatry. two main archival sources were used. the first was the website created by dr caire on the french history of psychiatry (http://psychiatrie.histoire.free.fr/). this site constitutes a rich documentary database on psychiatric hospitals. hospitals are presented by département with the date of creation when known. in addition, personal communication with the author made it possible to enhance the information available from this source. the other useful source was found in the paper by longin (1999) , which presents historical periodisation of the construction of psychiatric hospitals. institutions can be dated and located within départements. most of the data were taken from official reports (constans, lunier & dumesnil, 1878; esquirol, 1818) . to analyse the spatial diffusion of psychiatric institutions at departmental level, a temporal and geographical database was constructed showing french departmental boundaries and the associated resident populations for each period. first drawn up in 1790, the boundaries of french départements were modified throughout the 19th and 20th centuries, partly because of modifications to national borders (such as germany's annexation of the alsace and moselle regions during 1870e1918) and partly because of changes within the national space due to demographic and urban growth during the 19th century. rapid and spatially uneven population growth since the 19th century led to increasing disparities in population size between départements. various base maps were constituted for different years from a historical database of french towns and their attribution to départements (http:// cassini.ehess.fr). demographic data used to assess the scale of urban development were collected from different sources: ined-insee census demographic tables (croze, 1988) for the period 1861 to 1982, and the royal almanach for the years 1801 and 1816 (http://sref.free.fr, http://splaf.free.fr/). for local analyses, another database comprised of historical data on french cities initially produced by pumain (pumain & riandey, 1986) and completed by guerin and paulus (guérin-pace, 1990; guerois & paulus, 2002; pumain & riandey, 1986) was used. this database contains city population figures for the period 1831 to 1999, and a classification of cities distinguishing between: urban centres (most populated parts of urban agglomerations), isolated cities (urban areas bounded within a single urban space), suburban areas and rural areas. this morphological definition of cities, taking into account both population size and continuity of built up areas, combined with information on the dates urban areas first developed, reflects the structure of the french urban system at different time periods. more detailed information was compiled for each asylum analysed and for locations in which they were located (table 1) . the analysis was designed to explore whether the geographical pattern of the diffusion of asylums in france seemed consistent with the processes thought to be influencing this diffusion, as reviewed above. the analysis proceeded by first trying to establish whether the 1838 legislation provided a major impetus to the development of asylums throughout the country, which would be consistent with the idea of a centralised institutional innovation. then, at different historical phases of development of asylums, the analysis investigated the parts of the country where they were set up and the types of geographical setting where they were concentrated. in order to model the neighbourhood diffusion process, a contiguity matrix of french départements was created and in each département, euclidian distance was calculated between asylum locations and the city where the departmental administrative centre, (representing the local seat of government power), was located. finally, to bring all this information together, a multiple correspondence analysis (mca) was performed on the dataset to examine the connections between different aspects of the diffusion process. at this point we were also able to explore the possible significance of independent and religious institutions that were not recognized by the state as psychiatric 'asylums', but which may have influenced the spread of alienist ideas. descriptive analysis was carried out using the sas statistical package, mca was carried out with spad software (morineau & aluja-banet, 2000; morineau & morin, 2000) . mca is a useful tool to identify the main dimensions of a spatiotemporal diffusion process (saint julien, 2001) . it allows us to highlight key components of the diffusion process and to analyse their interactions. of the various techniques for multivariate analysis available, mca (or 'homogeneity analysis') (everitt & dunn, 2001) was selected because it can include categorical variables (lebart, morineau & piron, 1998; volle, 1997) . alternative methods also considered were multiple factorial analysis (mfa) (escofier & pagès, 1998) or mixed data factor analysis (mdfa) (pagès, 2004) . however, mca was preferred since it is widely used and understood, as well as being the most likely to offer statistically robust results. to carry out this mca, quantitative variables were converted into nominal categories, choosing a classification which would generate similar numbers of categories as were present in the qualitative variables. if the variables in mca differ significantly in the number of categories, this will tend to distort their impact on the analysis. this is because variables with a large number of categories will carry disproportionate weight in the resulting dimensions. table 1 variables characterising the asylum facilities and the places where they were located, incorporated in the mca. the units of analysis for the mca are the lunatic asylum locations created in french départements since the 17th century. variables used in the mca are listed in table 1 . the full range of variables are only included in the mca for asylum institutions of interest here (state-sponsored psychiatric hospitals that were recognized by, or were established in response to, the 1838 lunacy act). however some information relating to private and non-specialised institutions that were not in this category are also projected on the mca plots as illustrative individual cases, which may influence the pattern of relationships in the rest of the analysis. for example, a psychiatric ward in a general hospital, could have constituted an 'acceptable' way of caring for people with mental health problems in a département and such provision may have resulted in a delay in the establishment of a dedicated asylum facility in that area, or progressive independent institutions may have played a role in the dissemination of alienist ideas in psychiatry. the mca includes information on the position of départements within each of 8 statistical regions in france, (using the territorial units for statistics nomenclature defined for the member states of the european union (nuts1)). this gives an indication of the geographical position of the département where innovation took place at different time points, and the category of settlement in which the asylum was located. components of the mca were then used to build a classification of different types of lunatic asylum locations. this cluster analysis was based on a hierarchical ascendant classification using 'ward criteria' aimed at both maximising inter-group inertia and minimising intra-group inertia. to optimize cluster homogeneity, we used the 'dynamic nodes' method, a consolidation procedure involving aggregation around moving centroïds. temporal-spatial trends in the adoption of lunatic asylum facilities in french départements from 1617 to 1981 fig. 1 shows the time trend in the proportion of départements adopting the asylum model of care (i.e. for each year, the proportion of départements that had established at least one asylum). the diffusion of these psychiatric institutions through france lasted almost 400 years, from 1617 to 1981. as shown on fig. 1 , the diffusion process is still incomplete, because eight départements out of 95 have never had a specialised, public sector psychiatric hospital, whereas the 'deinstitutionalisation' of psychiatric care began in the 1960s with the introduction of acute psychiatric units for the provision of care within general hospital structures. four of these eight départements had been accommodating people with mental disorders in specialised wards in general hospitals since the 19th century. the other four départements had never previously provided a specialised public hospital service for mentally ill patients but currently provide acute psychiatric beds in multi-specialty hospitals. the general form of the curve is consistent with the typical pattern of development of innovation diffusion processes. the curve is similar to an 's-shaped' logistic form, apart from a perturbation caused by the resumption of new adoptions after the 1940s, as registration of new hospitals recommenced in france after a hiatus during world war ii. particularly notable is the absence of any change in the trend associated with the introduction of the 1838 act. the rate of innovation had started to progress most rapidly well before this date, and the rate of diffusion of the asylum model across départements in france actually slowed down shortly after the 1838 act was passed. therefore, it seems that at most the act only confirmed a pre-existing trend of introducing the process, but there is no evidence that it led to its acceleration. as is typical of diffusion processes, four main phases in the introduction of french asylums can be identified, similar to the stages of 'emergence', 'expansion', 'consolidation' and 'new expansion' proposed by hägerstrand (1953) . these are marked on fig. 1 and the départements involved in each phase are plotted on the maps in fig. 2 . details of the type of locality in which the new institutions were set up are also given in table 2 . phase 1: emergence of asylum institutions (17th and 18th centuries) initial innovation during the 17th and 18th centuries commenced in certain geographically dispersed centres around the 1645 1659 1673 1687 1701 1715 1729 1743 1757 1771 1785 1799 1813 1827 1841 1855 1869 1883 1897 1911 1925 1939 1953 1967 1981 emergence of the innovation expansion consolidation new expansion country (in some départements in the north and west of france and in dispersed locations in the south and west e see fig. 2 ). in 1800, only 10 out of the 86 existing départements at that time had a public or voluntary lunatic asylum. the voluntary sector, rather than the state, was the predominant early adopter (59% of the new establishments). as shown in fig. 2 , earlier adopters of the innovation appeared in diverse regions of france. most of the earliest adopters (before 1800) were départements located in northern france. in the 18th and 19th centuries, the north of france had higher level of education and industrialisation than the south (furet & ozouf, 1977; pumain, saint julien, & ferras, 1990) . these wealthier northern départements were also privileged areas for exchange and production (pumain et al., 1990) . it is interesting that paris, as the governmental and cultural centre of france, was not among the first to establish asylum facilities. although two hospitals with psychiatric wards and one private asylum existed during the 19th century, no state lunatic asylum was established in paris during the first part of the period. it was not until 1867 that it opened its first lunatic asylum 'sainte-anne'. asylums for the curable and incurable would be built outside the city at a later date (lamarche-vadel & préli, 1978) . this may have been because of rigidities in the system of institutional provision that already existed in paris, where the hôpital général had become firmly established. it would also be consistent with the preference for locating 'lunatic asylum' facilities in less urban settings. although pinel developed the 'lunatic asylum' concept through his observation and critique of conditions in the hôpital général setting, his ideas were initially concretised in new institutional facilities elsewhere in the country. with the exception of two départements, the early adopters were also generally more populated than non-adopters (398,000 inhabitants on average for this group of early adopters vs. 302,000 on average for non-adopters). apart from the striking absence of the parisian capital at the emergent phase of the process, this gives the impression of a hierarchical diffusion process, with the innovation spreading initially in the more populated and economically advanced areas and later reaching the more sparsely populated and economically 'backward' regions (saint julien, 1985) . this may have been a simple effect of the pressure of potential demand (which would be greatest in populated areas). however there may have been qualitative differences in the propensity for innovation and the availability of resources for new developments so that areas that were socially and economically more dynamic (pumain, 2006) led the way in adopting the new style of asylum. this was a period of very rapid industrial and economic growth in the north associated with the exploitation of coal and the industrial revolution, so that the region saw rapid urbanization and population growth and was relatively wealthy at this time with sufficient community resources for new health care development. the diffusion process advanced rapidly throughout most of the 19th century. an increasing rhythm of change is observed after 1808e1810, and well before the 1838 lunacy act. this may have been due to state intervention preceding legislation, and was probably also strongly influenced by the alienist network of reformists. during the years preceding the act, the question of care for the 'insane' was on the government agenda. the french alienists pinel and esquirol, both parisian doctors, were disseminating their ideas about treatment for the insane. the influence of pinel's report entitled medico-psychological treatise for mental alienation published in 1801 and reprinted in 1809, reached beyond the medical and bureaucratic fields (goldstein, 1997) . it is very likely that this original paradigm shift in psychiatry provoked by pinel (and by his colleagues in other countries such as william battie and william tuke in england) initiated the lunatic asylum diffusion process rather than national government policy (see philo, 2004, and foucauld, 1988) . with his theory on mental alienation and moral treatment, pinel laid the foundations of french psychiatry through the diffusion of his ideas. on the eve of the 1838 act, 38 départements out of 86 had already developed asylums to implement the innovations he proposed. voluntary initiatives remained numerous during this phase. if there was a 'tipping point' at which innovation started to escalate, it occurred prior to 1838. the legislation appears to have simply taken up and officially endorsed a previously established movement by encouraging the diffusion of the innovation throughout the national space. the political and economic context may also have played a role in these developments. the french government, under the imperial regime (1804e1814) and the following restoration (of monarchical sovereignty) until the 1830s, brought a degree of political stability and economic expansion favouring hospital development (longin, 1999) . psychiatric establishments created by the state became predominant after the 1838 act. between 1838 and 1899, 67% of the new asylums were the result of public initiatives (as opposed to institutions set up by voluntary bodies and recognized by the state after their inception). at the eve of the 20th century, 61 out of the 86 départements in existence by that time had adopted the innovation by constructing an asylum facility. the maps for 1838 and 1900 in fig. 2 show that more central and southern parts of the country had begun to establish asylums. the geographical pattern of asylum development also suggests that the 'contagion model' of diffusion is also relevant throughout the 19th century. this is confirmed by the finding that a département was more likely to be an adopter when a neighbouring area had already established asylum facilities. among the neighbouring départements bordering earlier (pre-1838) adopters, 49% had adopted the innovation during the period 1838e1899 versus 17% of départements not neighbouring previous adopters. this contagious diffusion model seems to be more relevant in the north of france. this might be associated with the more advanced development of communication networks in the north of france at this time. in the south, the innovation seems to have been taken up more spontaneously and randomly in space (fig. 2) . the rate of the new establishment of asylum facilities slowed in the 20th century, as most départements that had not already done so adopted this type of mental health care facility. the apparent acceleration in 1940 is the result of a bias in the data, noted above; some psychiatric hospitals established by 1940 could not be precisely attributed to the preceding years. then we see a period of relative stabilisation until the 1960s when most of the 'late adopters' made some provision of this type. during the first part of the 20th century, the drive to expand provision seemed to focus particularly on rural and less populated départements. the north-east and south-west of france constitute areas with high levels of adoption during this period (see fig. 2 ). classic models of the diffusion of innovations would predict a slowing down in the third phase of a diffusion process, but longin (1999) suggests a further explanation of this relative stabilisation at the beginning of the 20th century, linked with the development of secularism. the 1905 act on separation of church and state prevented any denominational private enterprise. furthermore, damage during the first world war strongly affected some asylums. closures and transformations were considered in some cases. this is a period of rehabilitation and repair rather than of new construction of asylums. the second world war resulted in less destruction of hospitals but more than 40,000 patients died in french psychiatric hospitals during this period. concern over conditions in asylum facilities, the discovery of neuroleptics, together with changes in the economic and political situation after wwii, subsequently led to a new mental health strategy: the 'sectorisation' policy. sectorisation interrupted the classical diffusion process proposed by hagerstrand, so that the final phase of complete saturation predicted by his model did not occur. instead, sectorisation introduces a new paradigm in psychiatric care, with a shift from large residential institutions to community-based services. in france, this process of deinstitutionalisation was initially planned on the basis of a territorial strategic framework. the 'psychiatric sector' was defined as a geodemographic area of around 70,000 inhabitants, for which a specialised team and a range of community-based services were dedicated. at this point the incentive to provide psychiatric hospital facilities was no longer driven by the aim of providing institutions for long term care, but by the need to convert and redevelop the service infrastructure to provide acute inpatient services as part of a deinstitutionalised model of care. to be able to implement this new policy, départements who had not yet followed the trend to build asylums had to create such acute facilities. the last départements to build inpatient psychiatric services often set up psychiatric hospitals or psychiatric wards in the multi-specialty hospitals serving the area, resulting in a more rapid period of growth in the number of psychiatric facilities after the 1960s. arguably, these were part of a new phase of innovation in psychiatric care, rather than the last stages of the diffusion of asylums. however, they can also be seen as a continuation of psychiatric service infrastructure development that had been set in motion though the asylum development process, since in the french case the original asylums have often been retained and converted to the community care model. choice of location within départements: distancing the 'mad' from the city by shifting the scale of analysis to a more local one, one can also discern what seem to be the effects of changing ideas about what constituted both a therapeutic setting and an appropriate location for a 'lunatic asylum'. the communes where new asylums were located were examined in terms of their population size, their general position relative to urban areas and their distance from the administrative centre for the département where the prefecture (government headquarters for the département) was located. as discussed above, the lunacy act gave départements no precise guidelines as to the ideal setting for such facilities but alienist theories suggested that a rural location was preferable to an urban setting. at present, more than half of the 144 public or integrated psychiatric hospitals created throughout the period are located in what are now urban centres; 16% in suburban areas, 17% in isolated towns and 14% in urban areas (french population census, 1999) . however, it must be borne in mind that the urban geography of france has developed over time so that a third of the asylums that were initially established in rural settings were later absorbed by urban sprawl and are now suburban areas, while some previously rural settings have become urban centres or small towns. table 2 thus shows that historically, 48% of the psychiatric hospitals were initially located in urban centres, 36% in rural areas, 12% in isolated towns, and 4% were initially located on the outskirts of cities. table 2 also shows how, from the early 1800s to the first part of the 20th century, the distribution of new asylum locations shifts over time from predominantly urban to more rural and semi-rural settings. a different perspective on the geographical position of these asylum buildings is provided in table 3 , which shows their average distance from the administrative centre of the département and the proportion of the asylum buildings that were located within the city centre where the prefecture (local seat of state government) was located. the pioneering asylum facilities set up before 1800 were on average located 7 km from the main city centre (60% were within the main administrative city centre). this is consistent with the theory that dominant centres in the urban hierarchy adopt the innovation before smaller centres. the mean distance to the prefecture is greater for the asylums set up during the 19th and early 20th centuries (20e30 km from the main urban centre of the département). this may be a reflection of the diffusion of alienist ideas concerning the appropriate setting for an asylum. after 1960 the average distance to the main city centre declined to 18 km indicating a growing proportion of more urban locations for more recently established facilities. however, by this phase the geographical pattern of diffusion became quite complex. table 2 also shows that in the second part of the 20th century, the positions of new asylums were more widely distributed in suburban settings and in more isolated urban centres. while rural locations were still often selected, we can also observe a larger number of psychiatric hospitals being set up in large cities (table 2) . while urban areas with populations of over 50,000 adopting this innovation slowly decreased until the 1960s, the trend reversed in the last part of the 20th century. this should be considered in the 1960's context of deinstitutionalisation associated with a psychiatric paradigm shift. the aim in this most recent period was no longer to isolate and distance people with mental health problems, but to integrate them into the community and to bring the health care facilities closer to population centres. multiple correspondence analysis: relationships between different temporal-spatial trends these trends are summarised in a multiple correspondence analysis (mca) to explore the complex associations between the attributes of asylum facilities, listed in table 1 . a scree plot analysis of eigenvalues showed that the first four dimensions from the factorial analysis account for 37% of the variability. fig. 3 presents the first two dimensions (which together explain about 22% of the variability) and the size of the data points indicate their influence on the pattern of correspondence between the different variables. the first dimension on the horizontal axis is strongly structured by characteristics of the places where hospitals were sited. it clearly opposes hospitals located in middle to large-sized urban centres or main administrative centres (to the left of the diagram) to sparsely populated rural areas or semi-rural settlements (less than 5000 inhabitants) more distant from the main cities (to the right). by projecting the temporal dimension on this factorial component (the phase in the diffusion process when the hospital was established, marked as a jagged line in fig. 3) , a path emerges demonstrating a strong relationship between location and time. thus middle to large-sized urban centres are more likely to be the settings of pioneer lunatic asylums, created before the lunacy act, while less central and more rural locations are more often sites table 3 mean distance of asylum locations from the city where the prefecture (regional government office for the département), was located, according to phase of diffusion. for 20th century establishments before the 1950s. this seems consistent with the idea of a hierarchical diffusion trend, with early adoption of the asylum model in larger centres and later adoption in smaller settlements. the second dimension, on the vertical axis, is essentially structured by the relationship between the locations of public sector institutions of special interest here and the presence of facilities provided by independent charitable organizations. this dimension opposes (at the bottom of the diagram) voluntary hospitals (frequently of religious origin, often located in places with no preexisting provision) to places with existing, 'embryonic' provision in a public and secular hospital (at the top of the plan). this is consistent with a theory of path-dependency in service development, later hospital developments being associated with earlier patterns of development. the most densely populated départements also appear in the upper part of the diagram, suggesting a longer history of provision of public facilities in these départements. the third and fourth axes are not illustrated. the third axis opposes suburban asylum locations, often close to the main urban centre of a département, to hospitals located in isolated places, further from the main cities. the former group was more likely to be in the voluntary sector, whereas the latter group was more often public sector facilities. the projection of the temporal dimension of the diffusion process provides a clear pattern; suburban locations (which were on the city fringes when the asylum was established) were more common for pioneer establishments while isolated locations were later developments. the fourth axis showed suburban locations were more common in rather sparsely populated départements while very rural locations were more often chosen in more populated départements. this may suggest that in rather urbanised and industrialised settings the move toward tranquil rural settings promoted by the alienist movement was particularly strong. our study was faced with several challenges, so that the conclusions are subject to several caveats. the first of these was the question of how to analyse a hierarchical diffusion hypothesis when the urban hierarchy was changing significantly through the period studied. the dramatic modification of the french national urban hierarchy during the 19th century due to the unprecedented urban growth biases the hierarchical diffusion model. the départements' mean population density increased by a factor of 1.8 during the 19th century with a good deal of local variability, which radically altered their demographic ranking. the urban hierarchy also became more differentiated: in 1801, for the least and the most populated départements, the population density ratio was around one to seven; by 1982, the ratio was one to 34. this unstable urban hierarchy makes the hierarchical hypothesis difficult to apply, even if we hypothesise that the greatest population growth was associated with a greater probability of adopting the lunatic asylum innovation. similar challenges also face other studies of diffusion over extended time periods. we also note that percentages in the tables and the results of the mca need to be interpreted with caution due to the relatively small numbers of data points. this makes it especially difficult to assess the later stages of the diffusion of the asylum model. it would be interesting to have more information on the capacity of the institutions and the numbers of patients using them, in order to assess the extent to which provision was related to likely demand in départements of varying population size. bearing in mind these limitations, this analysis of spatial diffusion of asylum facilities as an innovation has shown the limits of the relevance of classical models of spatial diffusion. the contagious diffusion model, arguing for diffusion governed by geographical proximity, does not seem very appropriate for our case study, except during the early expansion phase in the 19th century. in addition, its applicability remains limited to the northern part of france, perhaps because of the stronger economic, industrial and demographic development of this french region at that period. elsewhere in france, it is difficult to distinguish the effect of contagious diffusion specific to psychiatric hospitals. while the hierarchical diffusion model seems to be more relevant in our study, it nevertheless proves to be inadequate in explaining the whole process of diffusion and location of lunatic asylums. some large urban centres, and namely the capital city, paris, delayed in the construction of asylum facilities despite the fact that it had been the centre of emergence of the clinical and therapeutic ideas about the moral treatment and the need for such asylums. this delay in the creation of a lunatic asylum in paris was criticized at the time and interpreted as an administrative failure. for example, semelaigne wrote in 1860: 'in france, several large cities already have model establishments, and rival improvements are developing in foreign countries. in paris, however, through a regrettable anomaly, the bicetre and salpetriere hospitals are not affected by this trend, as indicated by both their imperfections and gaps in their scientific progress and actual achievements. this immobility, in a centre from which fruitful initiatives usually emanate, cannot continue. the capital city is embarrassing itself. a special commission has been established to consider the changes required in this situation'. (translated from a quote from daumézon (1959) ). this phenomenon reflects both the social rejection of people with mental illness and the facilities to treat them, and processes operating in landscapes of power as defined by dear and wolch (1987) . while paris was a centre of psychiatric knowledge, it may have been slow to establish asylums because of the effort required to reorganize the existing provision in 'hôpital général' facilities, and there may also have been motives to distance people with mental illness from the capital by devolving provision for 'the insane' to the provincial départements. this analysis therefore differentiates between the site of innovation in the sense of development of a new model of care (the alienist approach) and the diffusion of the concrete expression of this model: i.e., special purpose residential care facilities designed to deliver this psychiatric care. while diffusion of alienist ideas may have followed the classical hierarchical diffusion model from paris to other large urban centres (and other countries), and then to smaller urban centres, the diffusion of asylum facilities was influenced by other factors which will have favoured or impeded their establishment. among these we can include the local influence of organizations prepared to create unconventional facilities, (these were apparently often not-for-profit private associations or religious institutions), and the economic and social dynamism of communities within some départements. of course, there exist alternative interpretations of pinel's ideas and taking them into consideration highlights how change in health care systems is multi-faceted and complex. foucault, for example, interprets the diffusion of the therapeutic benefits of the asylum model in terms of the diffusion of growing power and discipline exerted by the medical profession in psychiatry (foucault, , 1975 . other authors have interpreted the development of psychiatric medicine less in terms of repression and punitive action towards mentally ill people and more in terms of innovative knowledge of the social and psychological determinants of mental illness progressively leading to a long term shift towards new models of care and risk governance (gauchet & swain, 1980; quétel, 2009; swain, 1977) . either interpretation is particularly interesting in the french context in that it is, arguably, rather unusual in france for professional associations, rather than the state, to determine national policies and welfare strategies. apart from the situation in paris, the hierarchical diffusion model seems to apply to our case study quite well, particularly before the 1838 act. except for some rural départements where religious communities initiated asylum development, pioneer départements were usually densely populated and tended to be relatively advanced both socially and economically. the average size of new adopters (absolute and relative to the period) tended to decrease until 1838, which suggests that the innovation was filtering down the urban hierarchy. having said this, there were some inconsistencies in the general trends; certain départements with small populations established asylums quite early, while some more populated areas were slower to set up asylums. to some extent, the 1838 act, requiring each french département to have an asylum, disrupted the hierarchical diffusion process, imposing a more universal diffusion of asylum development. although the 1838 act did not proactively initiate the trend to set up asylums, it nevertheless framed the later stages of the process and may have influenced its course of development. likewise the 'sectorisation' policy introduced in 1960, in the wake of deinstitutionalisation of psychiatric care, also 'interrupted' the final stages of the diffusion process, as discussed above. our findings therefore raise questions concerning the relevance of classical diffusion models for the interpretation of this example of health system development and argue for an approach based on more complex models. a more relevant conceptual framework might be political ecology, involving the exploration of large-scale social, economic and political influences that shape the local context (e.g. described by mayer (1996) , richmond,c., elliott,s., matthews,r., & elliott,b. (2005) ), as well as locally specific factors that influence the trajectory of development of health care systems. such a perspective would also place more emphasis on mental health system development in its wider social, economic and political context, including the evolution of the social representations and medical knowledge of mental illness, the political management of the diffusion of these innovations, and also the profound changes in the urban hierarchy of the country through the 19th century. a conceptual framework based on political ecology would also make it possible to consider that decisions regarding the development of psychiatric care were being made simultaneously at different geographical scales, from local to national level. furthermore, it would allow us to emphasize the importance of a historical perspective, stressing the path-dependency that helps us to understand how historical conditions influenced the dynamic processes of innovation considered here. ideas of path-dependency also continue to be particularly relevant for french mental health provision because of the continuity between past and present in the geography of the provision of services. contemporary processes of french deinstitutionalisation are strongly structured by the past heritage of asylum institutions. unlike the situation in the united kingdom or the united states, french deinstitutionalisation has not led to the mass closure of psychiatric hospitals. to date in 2010, no psychiatric hospital closure has been registered in france following the deinstitutionalisation principles. the present psychiatric sectorisation policy therefore has to adapt to this pre-existing asylum geography. this creates issues of accessibility and problems of rehabilitation and transformation of parts of disused buildings, often costly to maintain and difficult to convert to other uses, especially when buildings are classified as historical monuments. this paper also opens up a large field of research, since the geography of french mental health care has not been previously studied, despite the strong spatial dimensions of mental health care planning, as enshrined in the 1838 act, in the 1960 policy for geographical division of the country into psychiatric sectors, and, more recently, in the 'area health plans' (projets médicaux de territoire) aiming to facilitate and coordinate primary and hospital care, social and health services. the urban geography of sars: paradoxes and dilemmas in toronto's health care la géographie du sida en afrique. cahiers géos 1e37 the plague of 1647e1658 in the western mediterranean: the italian side diffusion of high technology medical innovation -computedtomography scanner example the social geography of aids in brazil airline networks and the international diffusion of severe acute respiratory syndrome (sars) la mesure de l'urbanisation aux etats-unis, des premiers comptoirs coloniaux aux metropolitan areas (1790e2000). cybergeo, systèmes, modélisation, géostatistiques what asylums were, are and ought to be histoire de la france. l'espace français instruction sur la manière de gouverner les insensés, et de travailler à leur guérison dans les asyles qui leur sont destinés rapport sur le service des aliénés en 1878 tableaux démographiques. la population en france: histoire et géographie historical account of the climates and diseases of the united states of america a view of the diseases most prevalent in the united states of america analyse de processus centralisés de diffusion spatiale: le cas des établissements des réseaux de services aux entreprises rapport présenté à la séance du 24 novembre 1959 de la commission de santé mentale landscapes of despair: from deinstitutionalization to homelessness analysis of spatial diffusion patterns for aids cases in some brazilian states dictionnaire des sciences humaines analyses factorielles simples et multiples. dunod des établissements consacrés aux aliénés en france et les moyens de les améliorer des maladies mentales considérées sous les rapports médical, hygiénique et médico-légal applied multivariate data analysis histoire de la folie à l'âge classique surveiller et punir. naissance de la prison new york: virgin books. translation to english by richard howard of foucault lire et écrire: l'alphabétisation des français de calvin à jules ferry l'archivio della follia la pratique de l'esprit humain. l'institution asilaire et la révolution démocratique consoler et classifier: l'essor de la psychiatrie française. les empêcheurs de penser en rond la dynamique d'un système de peuplement: évolution de la population des villes françaises de 1831 à 1982 commune centre, agglomération, aire urbaine: quelle pertinence pour l'étude des villes? cybergeo innovation diffusion as a spatial process diffusion of abortion facilities in northeastern unitedstates railways, disease and health in south-africa the 'world of yoga': the production and reproduction of therapeutic landscapes diffusion of influenza in england and wales rings of madness: service areas of 19th century asylums in north america histoire des hôpitaux en france community-based mental health care in britain and italy: geographical perspectives aids and medical geography:embracing the other? eléments pour une histoire de la psychiatrie reconstructing the initial global spread of a human influenza pandemic: a bayesian spatial-temporal model for the global spread of h1n1pdm petite histoire des hôpitaux psychiatriques francais en anglaisa short history of french psychiatric hospitals the political ecology of disease as one new focus for medical geography medical geography understanding the spatial diffusion process of severe acute respiratory syndrome in beijing the role of population heterogeneity and human mobility in the spread of pandemic influenza analyse en composantes principales (avec illustrations spad) smoke-free spaces over time: a policy diffusion study of bylaw development in alberta and ontario analyse factorielle de données mixtes innovation diffusion within and between national urban systems use of surveillance data in developing geographic dissemination strategies: a study of the diffusion of olanzapine to michigan children insured by medicaid a geographical history of institutional provison for the insane from medieval times to the 1860s in england and wales: the space reserved for insanity traité médico-philosophique sur l'aliénation mentale ou la manie hierarchy in natural and social sciences le fichier de l'ined: "urbanisation de la france les interactions spatiales: flux et changements dans l'espace géographique histoire de la folie de l'antiquité à nos jours the political ecology of health: perceptions of environment, economy, health and well-being among 'namgis first nation data on the diffusion of the american plague in the province of torino italy processus de diffusion spatiale et modélisations du changement théorie de l'évolution économique the slow plague -a geography of the aids pandemic -gould severe acute respiratory syndrome (sars) in asia: a medical geographic perspective the geographical spread of avian influenza a (h5n1): panzootic transmission the spatial anatomy of an epidemic: influenza in london and the county boroughs of england and wales, 1918e1919. transactions of the institute of british geographers retrospective space-time analysis of h5n1 avian influenza emergence in thailand statistique des asiles d'aliénés pour les années 1854 à 1860. strasbourg: imprimerie adminsitrative de veuve berger-levrault le sujet de la folie the diffusion of the influenza pandemic of 1918 in hartford description of the retreat analyse des données u.s. apartheid and the spread of aids to the suburbs: a multi-city analysis of the political economy and spatial epidemic threshold data-driven exploration of 'spatial pattern-time process-driving forces' associations of sars epidemic in beijing a brief history of epidemic and pestilential diseases the end of the line: has rapid transit contributed to the spatial diffusion of hiv in one of canada's largest metropolitan areas? the authors would like to thank pr. denise pumain for her support and advice during the research process and dr. michel caire for his valuable assistance in the building of the database. this study was funded by an industrial agreement for training through research (cifre contract), through the national agency for technical research. key: cord-337744-g17qe8fi authors: pullano, g.; valdano, e.; scarpa, n.; rubrichi, s.; colizza, v. title: population mobility reductions during covid-19 epidemic in france under lockdown date: 2020-06-01 journal: nan doi: 10.1101/2020.05.29.20097097 sha: doc_id: 337744 cord_uid: g17qe8fi on march 17, 2020, french authorities implemented a nationwide lockdown to respond to covid-19 epidemic emergency and curb the surge of patients requiring critical care, similarly to other countries. evaluating the impact of lockdown on population mobility is important to help characterize the changes in social dynamics that affected viral diffusion. using travel flows reconstructed from mobile phone trajectories, we measured how lockdown altered mobility patterns at both local and country scales. lockdown caused a 65% reduction in countrywide number of displacements, and was particularly effective in reducing work-related short-range mobility, especially during rush hours, and recreational long-range trips. anomalous increases in long-range movements, localized in both time and space, emerged even before lockdown announcement. mobility drops were unevenly distributed across regions. they were strongly associated with active population, workers employed in sectors highly impacted by lockdown, and number of hospitalizations per region, and moderately associated with socio-economic level of the region. major cities largely shrank their pattern of connectivity, reducing it mainly to short-range commuting, despite the persistence of some long-range trips. our findings indicate that lockdown was very effective in reducing population mobility across scales. caution should be taken in the timing of policy announcements and implementation. individual response to policy announcements may generate unexpected anomalous behaviors increasing the risk of geographical diffusion. on the other hand, risk awareness may be beneficial in further decreasing mobility in largely affected regions. our findings help predicting how and where restrictions will be the most effective in reducing the mobility and mixing of the population, thus aiding tuning recommendations in the upcoming weeks, when phasing out lockdown. french authorities responded to the rapid growth of covid-19 cases by imposing heavy restrictions on mobility, as many other countries in europe and beyond 1 . lockdown was enforced on march 17, 2020, and helped slow down infection rates and limit the strain on the healthcare system 2 . accurately measuring changes in human mobility under these restrictions is essential to (i) quantitatively determine how imposed measures and recommendations (e.g. regarding telework where possible, ban of leisure trips) translated into reduced mobility at specific scales and times, (ii) inform models estimating the effectiveness of the ongoing lockdown in reducing the epidemic spread 3,4 , (iii) help devising social distancing measures needed for the post-lockdown phase. accessing human mobility data to measure these changes is now possible at several spatial and time scales, and often in nearly real-time. these data have been proven useful in many epidemiological contexts 5 -including for example the west africa ebola epidemic 6 -and are being used now for covid-19 pandemic in many countries 7 -namely, belgium 8 , germany 9 , india 10 , italy 11,12 , poland 13 , spain 14 , uk 15, 16 , usa [17] [18] [19] . mobile phone records are one of the main sources of mobility data. they describe travel flows among the different locations of a country. these flows can be analyzed over time to study population patterns, with no information on individual users, safeguarding privacy 7, 20, 21 . in this report, we used data provided by orange business service flux vision, and studied how mobility in france changed before and during lockdown. we broke down our results by trip distance, user age and residency, time of day, and analyzed regional data and spatial heterogeneities. we investigated behavioral responses to announcements of interventions, and to the epidemic burden, as well as associations of mobility reduction with demographic and socioeconomic indicators. considering the network of travel connections among french locations, we also identified the most vulnerable and most resilient connections to the mobility shock induced by lockdown, with a specific focus on main french cities. mobile phone data were provided by the orange business service flux vision. they comprised origin-destination travel volumes among ~1,500 geographic areas of mainland france, which group municipalities at the 2018 epci (établissements publics de coopération intercommunale) level. the average distance between the centroids of two adjacent areas is 22 km. travel volumes were computed on-the-fly from signals exchanged between mobile phones and the mobile network, which contain information about the identifiers of the mobile phone and of the antenna handling the communication. knowing the spatial localization of the antennas allows reconstructing the approximate position of the device in communication. this was then used to compute aggregated travel volumes among locations, with no residual information tracing back to the individual users. data provided the number of displacements (or trips) observed between any two consecutive locations where the user spent at least 1 hour. for each pair of locations and any given day, data were provided stratified by age class. travel flows were adjusted by orange to be representative of the general population. regional hospitalization data were obtained from santé publique france 22 . from them, we extracted as indicator the cumulated number of covid-19-related hospitalizations per 100,000 inhabitants at a given date, for each region. on april 5, 2020, grand-est . cc-by-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. (which was not certified by peer review) the copyright holder for this preprint this version posted june 1, 2020. population data and regional socioeconomic indicators were obtained from the french national statistical institute (insee) 23 . we used the following indicators: i) fraction of population in the age range 24-59, corresponding with the peak of activity 24 . île-de-france had the highest value (47.7%), bourgogne had the lowest value (42.4%). the sample standard deviation across regions was 1.5%. ii) 90 th percentile of the regional standard of living (niveau de vie), defined by insee 23 as the household's gross disposable income divided by the number of consumption units (measuring the size of the household -one unit for the first adult, 0.5 units for each additional person over 14 years of age and 0.3 for each child under 14 years of age). île-de-france had the highest value (46,607 euros), hauts-de-france the lowest (33,548 euros). the sample standard deviation across regions was 3,449 euros. employment data were obtained from insee 25 and from the report of the french ministry of labor on the impact of restrictions on economic activities 26 . as indicator, we used the fraction of employees in the sectors mostly affected by lockdown. these are the sectors in which at least 50% of employees stopped working (hotels, hospitality, food services, and construction), or had been working remotely (finance, insurances, it). île-de-france had the highest value (22.91%), bourgogne the lowest (11.70%). the sample standard deviation across regions was 3.10%. ethics. mobile phone data were previously anonymized in compliance to strict privacy requirements, reviewed and approved by the french national commission for data protection 27 (cnil, commission nationale de l'informatique et des libertés), ruling on all matters related to ethics, data, and privacy. timeline fit and prediction. to fit and forecast time series we used the forecasting procedure prophet by facebook open source 28 . we enforced weekly seasonality, and used school holidays by region 29 as additional (additive) regressors. trip analysis. our analyses were performed on all trips and on trips whose geodesic distance between location centroids is longer than 100 km (long trips). the cutoff of 100 km effectively discards commuting, as ~95% of daily work-related trips are shorter than 100 km 30, 31 . we distinguished between residents, i.e., users with french sim cards, and foreigners. we broke down data in three age classes: young (younger than 18 y.o.), adults (18-64 y.o.), and seniors (65+ y.o.). we classified tips by their time of day: daytime (7am-7pm), nighttime (7pm-7am), and distinguished between weekdays and weekends. during weekdays we also considered rush hours (7am-9am, 5pm-7pm). mobility reduction during lockdown. mobility reduction during lockdown was computed in a case-crossover framework by comparing the week starting monday april 6, 2020 (3 weeks into lockdown), to the week starting monday february 3 (control week). the latter was chosen as being before school holidays, and after strikes of public transport. all statistical analyses were performed in r, version 3.6.1. two-sided significance of pearson coefficients was determined at a level of 0.05. network analysis. nodes in the networks represent the geographic locations in which we divided mainland france, and links represent trips between locations. links are directed (trips have origins and destinations), weighted (by the number of trips linking . cc-by-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. (which was not certified by peer review) the copyright holder for this preprint this version posted june 1, 2020. . https://doi.org/10.1101/2020.05.29.20097097 doi: medrxiv preprint two locations), and evolve in time. to handle and analyze networks we used standard python libraries, among which networkx. to smooth spatial data, we used a standard gaussian kernel with fixed characteristic distance, and adjusted locations by their population (see appendix). the radius containing 95% of outgoing traffic from a city was computed by considering all mobility links that start from that city, each with its geodesic distance. they were included incrementally from the shortest to the longest (in terms of geodesic distance), until the cumulative sum of the weights of the included links reached 95% of the total outgoing traffic. changes in circle radius capture changes in the geographic pattern of outgoing mobility. if mobility is reduced homogeneously across distances, the radius will remain constant. if reduction increases with the distance, the radius will decrease (and vice versa). three phases have marked the french response to covid-19 epidemic (figure 1 ). phase 1 started in early january and can be identified with the first publication of covid-19 case definition by santé publique france 32 . its aim was to detect imported cases as quickly as possible and conduct case-contact epidemiological investigations to identify possible local transmissions and isolate cases. phase 2 started on february 29, 2020 upon appearance of localized clusters, and featured the same measures of phase 1 coupled with targeted social distancing interventions (e.g. school closure, gatherings and public transport bans) to stop possible transmission in the community. during this phase, two clusters were identified, in oise and haute-savoie. phase 3 was declared on march 14 when the virus was recognized to actively circulate in the territory. starting few days before phase 3, a set of announcements were made by french authorities that progressively led to the lockdown on march 17, 2020 ( . cc-by-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. (which was not certified by peer review) the copyright holder for this preprint this version posted june 1, 2020. while no observable change in mobility occurred during phase 1 and 2 of the epidemic, the start of phase 3 on march 14 had a substantial impact on mobility in france (fig. 1a) . this transition occurred prior to the announcement (march 16) and implementation (march 17) of lockdown measures, and saw nationwide mobility go from ~60m trips per day down to ~20m trips after lockdown entered into effect. the shock in mobility spread out over a transition period lasting almost a week. to study in detail this transition, we quantified the deviation of measured traffic flows from the predicted evolution of traffic over time. predictions were obtained from fitting is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted june 1, 2020. . https://doi.org/10.1101/2020.05.29.20097097 doi: medrxiv preprint mobility data from january 6, 2020, to monday, march 9 (training set, fig. 1) , and assuming no perturbation due to covid-19 and associated interventions after march 9. total flow was significantly below predictions starting march 14, as a likely consequence of the start of phase 3. mobility further decreased on sunday, march 15, when local elections took place. instead, an anomalous rise in traffic took place on the day before lockdown enforcement, which had higher volume than the surrounding days, whereas still lower than the predicted baseline. long trips (>100 km) were also significantlyalbeit slightly -below the predicted baseline during the weekend (march 14, 15). they however went back to seemingly normal values on march 16 -i.e., in agreement with the unperturbed prediction -, and near-to-normal values on lockdown day. however, this country-level behavior hid anomalous deviations from the predicted mobility behavior in specific locations, as fig. 1b shows. spikes in outgoing traffic are distinctively visible in île-de-france (the region of paris) and, at the same time, in incoming traffic in normandy and bretagne. they measure the pre-lockdown exodus out of paris occurring before lockdown took effect 33, 34 . analyses at finer scales within île-de-france revealed that anomalous outgoing traffic concentrated in the paris area, and western île-de-france. similar spikes of outgoing and incoming traffic were also visible in the south east, close to the alps, as reported previously 33 . the transition starting with phase 3 reshaped weekly patterns compared to those measured in the unperturbed mobility. before the mobility shock, a stable weekly pattern was observed, with peaks on fridays and troughs on sundays for all trips, and peaks on both fridays and sundays for long trips. during the transition, no weekly pattern was recognizable, as mobility was perturbed in different ways across several days. following the transition, patterns no longer featured peaks in mobility on fridays (for all trips) or on sundays (for long trips). mobility patterns quickly entered a new equilibrium after lockdown enforcement, marking the end of the transition period. using a crossover framework (see methods), we found that lockdown decreased the overall number of trips by 65% (figure 2a) . reduction was stronger for trips made by foreigners (~85%), suggesting that the enforcement of lockdown disrupted tourism and impacted more the mobility of foreign nationals in the country 35 . their number of trips was however very small even before lockdown compared to french residents (3%), therefore we excluded them from the rest of the analysis as their contribution is negligible. long-range traffic (>100 km) was disrupted more severely than average (85% reduction, fig. 2a) . this was likely associated with a disruption of long-range transportation (trains, flights), and the ban of leisure-related trips, also confirmed by the almost disappearance of long trips during the weekend (see below). . cc-by-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted june 1, 2020. . https://doi.org/10.1101/2020.05.29.20097097 doi: medrxiv preprint figure 2 . mobility reduction during lockdown across user type, age and time of day. reduction is computed as the average over the week starting monday april 6, with respect to the average over the first week of february (starting monday february 3). a), b) and c) show the relative reduction broken down by residents/foreigners, age classes, and times of the day. they also show statistics for all trips (orange) and long trips (green), defined as trips with geodesic distance longer than 100 km. horizontal orange and green lines indicate relative reduction on all residents (all, long, respectively). mobility reduction in total trips was homogeneously distributed across age classes (fig. 2b) . when considering only long trips, reduction instead increased with age, as seniors reduced their trips above 100 km by ~ 90%. drops in mobility were uneven across the time of the day (fig. 2c) . movements during rush hours were the most disrupted, indicating that the combined effect of school closure and telework led to a ~75% reduction. daytime movements during weekends also exhibited a higher-than-average decrease, hinting at a successful reduction of recreational activities. nighttime movements during weekdays instead recorded the lowest reduction, well below average. they might be related to unavoidable workrelated mobility, whose impact is however likely to be limited, as these movements make up for only ¼ of the total. long-range mobility almost completely stopped during weekends (around 95% decrease). regional heterogeneities in mobility reduction during lockdown. traffic reductions were not homogeneous across the 13 regions of mainland france. reduction of internal traffic was above average in 4 regions (île-de-france, auvergne-rhône-alpes, grand est, provence-alpes-côte d'azur), whereas markedly below average in bourgogne-franche-comté, centre-val de loire, and normandy (figure 3) . similar fluctuations were visible in outgoing traffic (coefficient of variation equal to 8.4% compared to 8.0% for internal traffic). île-de-france, hauts-de-france and grand est all experienced above-average reductions in outgoing mobility, as high as 80% for île-de-france. corse also exhibited a reduction comparable to île-de-france, showing a clear disruption of the long-range connections linking the island to mainland france. similar reductions were obtained with incoming fluxes in the regions (not shown). . cc-by-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted june 1, 2020. the impact of nationwide lockdown in the reduction of outgoing mobility per region was strongly associated with the fraction of the population in the most active age range (24-59 y.o.) 24 (pearson r = 0.91, p < 0.01) and the fraction of workers employed in sectors that substantially modified their organization during lockdown, due to telework, partial or full closure of activities (pearson r = 0.80, p < 0.01) (table 1 and figure 4) . it was moderately associated with the standard of living of the region (pearson r = 0.63, p = 0.02). regional drops in mobility in a given week (april 6-12, 2020) were strongly associated with covid-19 hospitalization rates registered and communicated in the week before (april 5) (pearson r = 0.73, p < 0.01; figure 4) . . cc-by-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted june 1, 2020. similar results were obtained for drops in mobility within the region, except for the association with the hospitalization rate per region, which however showed a similar, though non-significant, tendency (table s1 and fig. s1 in appendix). taking out the data point of île-de-france as the region mostly affected by a departure of inhabitants for relocation in other regions led to similar results ( table s2 in appendix). disruption of mobility connections. shifting the focus from overall traffic reductions to mobility connections between locations, we found that some connections completely disappeared, as individuals stopped going from one location to another (figure 5) . the probability that a mobility connection observed in the control week (week of february 3, 2020) was also observed when interventions were announced and after they entered into effect (persistence probability, fig. 5b ) decreased steadily during the transition period (67% of connections surviving in the week of school closure and nonessential activity closure announcements, march 9 to 15; 50% in the week of announcement and implementation of lockdown, march 16 to 22) to stabilize in the first . cc-by-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted june 1, 2020. . https://doi.org/10.1101/2020.05.29.20097097 doi: medrxiv preprint full week of lockdown (34% of connections surviving, march 23) and beyond. long connections were less resilient than average, as only 1/4 of them survived lockdown. after lockdown effects stabilized (e.g. starting the second full week of lockdown, march 30), connections usually characterized by small traffic prior to restrictions (weak connections) were the most likely to disappear, with 70% of them corresponding to 100 trips per week (fig. 5c) . the traffic lost on these connections however barely contributed to total traffic reduction (3% contribution). restricting the analysis to long mobility connections (> 100 km), the fraction of the weak connections disappearing slightly increased (from 70% to 89%), however with a reduction of 47% of the traffic. the disruption in connections occurred with a certain delay compared to reductions in traffic. for example, on monday march 16 -the day before lockdown -traffic was reduced by 30% with respect to the previous monday, but the number of connections went down by 4% only. one week after (march 23), traffic drop was 64% and the drop in the number of connections was 55%. is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted june 1, 2020. . https://doi.org/10.1101/2020.05.29.20097097 doi: medrxiv preprint lost on connections which used to have at most a certain weight in the baseline week. for all panels: orange: all links, green long links (longer than 100 km). restrictions on mobility during lockdown had an uneven impact on the 10 most populated french cities. the circle containing 95% of outgoing traffic from each city decreased after lockdown took effect for all cities, indicating that long-range mobility was disrupted more than short-range one (figure 6) . but reductions varied from more than 80% (paris, bordeaux, nice) to 60% (strasbourg, lille), mainly due to different patterns of commuting and connectivity characterizing the mobility of each city. in normal circumstances, paris is connected to almost the rest of the country, whereas the other cities have a more localized pattern of mobility with fewer long-range connections. once lockdown was implemented, surviving mobility shrank around the cities. connections among main cities disappeared too. considering the 4 connections per city with highest traffic that are compromised by the lockdown, we no longer detected mobility from bordeaux, montpellier, and nantes to lyon, or from montpellier to strasbourg (figure 6 ), compared to pre-emergency situation. . cc-by-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted june 1, 2020. . cc-by-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted june 1, 2020. using travel flow data extracted from mobile phone trajectories, we documented a large drop in both short-range and long-range population mobility following lockdown enforcement in france. overall, trips were reduced by 65%, similarly to reductions found in belgium 8 , spain 14 , and italy 11 during lockdown, albeit different data sources, spatial resolutions, and definitions of mobility proxies prevent direct numerical comparisons. the transition signaling the drop in mobility lasted almost a week, anticipating the enforcement of lockdown and creating opposite mobility behaviors. individuals started spontaneously reducing their mobility on saturday following the announcement of school closure, likely because of fear of the growing epidemic and heightened risk awareness 36-40 generated by the first governmental decision on nationwide interventions. the weekday-to-weekend pattern was disrupted, with overall mobility on monday following the closure of all non-essential activities similar to the preceding saturday. at the same time, fear of an imminent change in policy imposing stricter restrictions -as already implemented in italy, spain, austria 41 pushed individuals to relocate themselves even to farther away regions where to spend the period of lockdown, if put in place. the exodus, largely covered by the press 33,34 , occurred already before the announcement of lockdown and led to anomalous increases in mobility flows out of certain regions (e.g. île-de-france) and incoming in others (e.g. normandy). such behavior was similarly reported in china (from wuhan to other regions), in italy (from the north to the south) prior to the implementation of lockdown, and in india 10 . it demonstrates that the timing at which a policy is announced might disrupt social dynamics as much as the direct effect of the policy, at least in the short term. increased caution should therefore be considered in the period from announcements to enforcement to avoid unpredictable behaviors that may result in unwanted seeding of the epidemic to other areas. no increase in viral circulation became then visible in the receiving regions in the following weeks, as lockdown strongly suppressed epidemic activity in all regions 3, 4, 42, 43 . seeding events due to relocations may however become more important in phasing out the lockdown, as less strict social distancing measures may prevent such suppression. region-specific interventions may increase this risk by inducing similar behavioral responses. new york state reported for instance increased mobility in counties with no imposed lockdown 19 . in this perspective, nationwide interventions and restrictions limiting displacements were adopted by several countries 44, 45 to prevent compensation effects and reduce the possible geographical spread of the epidemic. once lockdown entered into effect, population mobility reductions were heterogeneous across regions. larger reductions were measured in regions more severely hit by the epidemic, with an estimated 1% decrease in regional mobility every 10 additional hospitalizations (per 100,000 inhabitants). this suggests that individuals witnessing a larger covid-19 burden on the hospital system in their region may have further limited displacements compared to those living in less affected regions. media largely communicated on the epidemic, also providing early on region-specific information on hospitalizations and mounting pressure on the healthcare system. exposure to this information likely triggered a behavioral response increasing compliance to movement restrictions. a similar, though stronger, behavior was observed during a 3-day national lockdown enforced nationwide in sierra leone in march 2015 in an effort to control ebola epidemic 6 . the correlation remains significant even taking out the region of île-de-france, which experienced a reduction in population due to relocation of individuals. . cc-by-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted june 1, 2020. . https://doi.org/10.1101/2020.05.29.20097097 doi: medrxiv preprint clearly, other factors may have come into play to differentiate drops in regional mobility. lockdown restrictions had a severe impact on jobs and the organization of work. regions with the higher proportion of activity sectors mostly impacted by the lockdown (due to telework, but also to complete or partial closure of sectors, such as tourism, entertainment, food services, and construction) also experienced larger drops on mobility. a smaller fraction of active individuals continued to go to work, while the others limited their displacements respecting lockdown mobility restrictions. indeed, regions with larger portions of the population in the most active age range (24-59y) 1,2 were also the ones where lockdown had the largest effects. besides the displacements to go to work, active population is also highly mobile for leisure activities, which were completely banned by restrictions (with short exceptions to do sport once a day for at most 1 hour). uneven mobility drops were also associated with socioeconomic disparities. increasing evidence points at different socioeconomic strata getting uneven shares of the covid-19 burden 46 . higher income jobs can often be performed remotely, in confinement, whereas lower income jobs cannot. a survey in france reported that 39% of low income workers were still going to their workplace during lockdown, against only 17% of high income workers 47 . also, wealthier population strata weather short-term financial losses better, making them more prone to stop working and stay at home if they are afraid or sick. at the same time, they can afford more leisure activities and have a more varied social network 48, 49 , leading to a higher rate of leisure-related mobility in normal circumstances. wealthier populations then likely experienced a larger mobility reduction because of the possibility to work remotely or stop working, as well as for the imposed ban on leisure activities. a strong response was documented in the older age class, which is at highest risk of developing severe forms of covid-19 if infected. seniors almost stopped taking trips longer than 100 km, likely to avoid leisure activities and family trips, as recommended by authorities. the most effective reduction in overall mobility occurred during rush hours, associated with a disruption of commuting patterns. this reduction alone likely boosted the role of mobility restrictions in suppressing viral diffusion, as mounting evidence shows that public transportation is a main risk factor for transmission 50, 51 . lockdown had a different impact on mobility depending on distance, causing larger disruptions on long-range mobility, as also reported in belgium 8 and italy 11 . short-range and long-range mobility flows play different roles in the spread of an infectious disease epidemic. short-range connections are mainly responsible for local diffusion in the community within and around a metropolitan area, whereas long-range connections drive the spatial spread of the epidemic, acting as seeding events to otherwise unaffected or weakly affected areas 52 . mobility flows out of the city of wuhan were shown to have seeded other prefectures in china in the early phase of the epidemic before travel restrictions and substantial control measures were implemented [53] [54] [55] . a delayed response or less efficient lockdown would have likely led to a larger outbreak increasing its geographical range. coupled with social distancing interventions, longrange mobility restrictions are therefore critical to geographically contain the epidemic, especially when epidemic activity is largely heterogeneous at the spatial scale, showing a patchy geographical pattern observed in many affected countries including france. banning trips above 100 km as announced by french authorities 44 will continue breaking the spreading pathways along which the epidemic could spread and reducing the locations reachable by the virus, as observed during the lockdown. nonetheless, in the lockdown phase we documented that some long-range mobility connection, among the ones with highest traffic, survived the restrictions -namely, from paris to . cc-by-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted june 1, 2020. . https://doi.org /10.1101 /10. /2020 montpellier, and from the other most populated cities to paris (except toulouse and strasbourg). these movements should be carefully accompanied by strict hygienic and preventive measures to avoid re-seeding events from visitors or returning residents, as discussed above. the largest reduction of mobility across distance was reported for paris. before lockdown, 95% of outgoing traffic reached destinations within 200 km from the city center, approximately the distance between paris and lille, close to the belgian border. after lockdown, this radius reduced to 29 km, the distance from the city center to disneyland paris. achievable distances from large cities shrank during lockdown, even in absence of explicit limitations on distance, also reducing the number of reachable destinations. mobility became more localized and restructured around metropolitan areas, serving the needs of individuals who continued their daily displacements associated to work, e.g. in essential professional categories. a similar geographical fragmentation induced by restructured local job markets was also observed in italy 12 . our analysis offered plausible interpretations on how the labor market, demographic and socio-economic indicators, and awareness of increased epidemic risk might have shaped the reduction in mobility, confirming evidence observed in previous 6,37 and current 47,56 outbreaks. being observational in nature, the study does not allow us to identify causal relationships; also, confounding effects among the covariates may be expected, but the available sample was too small to take this into account. focusing on the reduction in mobility during lockdown and its association to hospitalizations in the same time period, our study did not aim to assess the role of mobility in shaping the epidemic spread, but to investigate a behavioral response likely induced by risk awareness. associations were robust also removing the data point for île-de-france, the region mostly affected by the exodus of individuals relocating in other locations. this suggests that associations are not biased by a change in population size of the region. regional variations in mobility may be induced by differential restrictions based on estimated epidemic activity in the region 10 . however, this was not the case in france, where a nationwide lockdown was applied uniformly in the country. local authorities additionally imposed heavier restrictions in certain areas over time, like curfews in cities in hauts-de-france and in the south of france 57 . we did not consider these additional restrictions as possible factors in our analysis. we expect them to result in smaller effects, likely not visible at the resolution scale under study here. using aggregated flow data extracted from mobile phone trajectories, we documented the large impact that lockdown had on reducing mobility in france. different effects were observed across scales, with larger disruptions on long-range connections leading to a localization of the mobility. factors related to demography, professional categories, and socio-economic level were all associated with the reduction in mobility. uneven drops in population movements by region may also be explained by a different behavioral response linked to the perceived risk of the epidemic in the region. our findings may help predicting how and where restrictions will be the most effective in reducing the mobility and mixing of the population, thus aiding tuning recommendations in the upcoming weeks, when phasing out lockdown. . cc-by-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted june 1, 2020. . cc-by-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. (which was not certified by peer review) the copyright holder for this preprint this version posted june 1, 2020. . https://doi.org/10.1101/2020.05.29.20097097 doi: medrxiv preprint let ! be the value in location of the quantity we want to smooth. let ! be the population in , and ∆ !" the geodesic distance between locations , . then, the smoothed value is ! #$%%&'() = ∑ " " *+ ∆ !" -. where is the characteristic distance parameter. the sum runs over all the locations. table s1 . correlation coefficients. the table reports the correlation coefficients and their p-value for the four indicators considered and internal and outgoing regional mobility. . cc-by-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. (which was not certified by peer review) the copyright holder for this preprint this version posted june 1, 2020. table s2 . correlation coefficients without île-de-france. the table reports the correlation coefficients and their p-value for the four indicators considered and internal and outgoing regional mobility, computed excluding île-de-france. . cc-by-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. (which was not certified by peer review) the copyright holder for this preprint this version posted june 1, 2020. . https://doi.org/10.1101/2020.05.29.20097097 doi: medrxiv preprint figure s1 . reduction in internal mobility for the week april 6-12, 2020 vs. epidemic, socioeconomic, and demographic indicators. the following plot is the equivalent of figure 4 for internal mobility. correlation is evaluated between outgoing traffic and the four considered indicators: a) the population in active age (24-59 years old), b) the fraction of employees in the sectors mostly affected by lockdown. c) the 90 th percentile of the regional standard of living. pearson correlation coefficients and their p-values are reported, d) the cumulated number of covid-19 hospitalizations per 100,000 inhabitants on april 05, 2020. . cc-by-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. (which was not certified by peer review) the copyright holder for this preprint this version posted june 1, 2020. . https://doi.org/10.1101/2020.05.29.20097097 doi: medrxiv preprint how will country-based mitigation measures influence the course of the covid-19 epidemic? santepubliquefrance.fr. covid-19 : point épidémiologique du 16 avril 2020 expected impact of reopening schools after lockdown on covid-19 epidemic in île-de-france expected impact of lockdown in île-de-france and possible exit strategies human mobility: models and applications population mobility reductions associated with travel restrictions during the ebola epidemic in sierra leone: use of mobile phone data mobile phone data and covid-19: missing an opportunity? covid-19: belgium analyses telecom data to measure the impact of confinement covid-19 mobility project in germany mapping the lockdown effects in india: how geographers can contribute to tackle covid-19 diffusion covid-19 outbreak response: first assessment of mobility changes in italy following lockdown mobile phone data analytics against the covid-19 epidemics in italy: flow diversity and local job markets during the national lockdown community mobility changes due to the coronavirus (covid-19) outbreak in poland informe sobre los cambios de movilidad en españa debido a las medidas de confinamiento contra la extensión del covid-19 oxford covid-19 impact monitor analysis of human mobility in the uk during the covid-19 pandemic assessing changes in commuting and individual mobility in major metropolitan areas in the united states during the covid-19 outbreak effect of social distancing measures in the new york city metropolitan area mapping county-level mobility pattern changes in the united states in response to covid-19 protect privacy of mobile data aggregated mobility data could help fight covid-19 indicateurs : cartes insee. bilans économiques 2018 des régions françaises activité et conditions d'emploi de la main-d'oeuvre pendant la crise sanitaire covid-19 coronavirus : le grand exode des citadins confinement : plus d'un million de franciliens ont quitté la région parisienne en une semaine cnews. coronavirus : quel impact sur le tourisme epidemiological determinants of spread of causal agent of severe acute respiratory syndrome in hong kong what can we learn about the ebola outbreak from tweets? outpatients' behavior of seeking medical care after onset of severe acute respiratory syndrome and community control measures in beijing public perceptions, anxiety, and behaviour change in relation to the swine flu outbreak: cross sectional telephone survey survey on the likely behavioural changes of the general public in four european countries during the 2009/2010 pandemic covid-19 epi dashboard estimating the burden of sars-cov-2 in france covid-19: one-month impact of the french lockdown on the epidemic burden stratégie de déconfinement fase 2 american inequality meets covid-19 premiers de corvée et premiers de cordée, quel avenir pour le travail déconfiné network diversity and economic development using big data to study the link between human mobility and socio-economic development the subways seeded the massive coronavirus epidemic epidemiology and transmission of covid-19 in 391 cases and 1286 of their close contacts in shenzhen, china: a retrospective cohort study multiscale mobility networks and the spatial spreading of infectious diseases population flow drives spatio-temporal distribution of covid-19 in china the effect of human mobility and control measures on the covid-19 epidemic in china effect of non-pharmaceutical interventions to contain covid-19 in china location data says it all: staying at home during coronavirus is a luxury this study is partially funded by: anr projects evalcovid-19 (anr-20-covi-0007), sphinx (anr-17-ce36-0008-05) and dataredux (anr-19-ce46-0008-03); eu h2020 grants recover (h2020-101003589) and mood (h2020-874850); reacting covid-19 modeling grant. we thank luca ferreri, sylvain bourgeois, erwan le quentrec, zbigniew smoreda, chiara poletto, pierre-yves boëlle for useful discussions. key: cord-021158-075vh5jg authors: fortané, nicolas title: antimicrobial resistance: preventive approaches to the rescue? professional expertise and business model of french “industrial” veterinarians date: 2020-01-06 journal: nan doi: 10.1007/s41130-019-00098-4 sha: doc_id: 21158 cord_uid: 075vh5jg this article focuses on the development of veterinary medicine in the industrial pig and poultry production sector. in the current context of controversies over the public problem of antimicrobial resistance (amr), the veterinary profession is tending to promote a model of preventive medicine that is supposed to reduce the use of antibiotics in livestock farming. however, veterinarians specializing in pig and poultry production (“industrial vets”) have in fact been adopting such approaches to animal health for several decades. based on 28 interviews with pig and poultry veterinarians practicing or having practiced in western france between the 1970s and the 2010s, the article aims to understand how such a form of professional expertise has developed, and the business model that underpins it. contrary to public discourses which promote preventive approaches as a way to diversify professional expertise and to disconnect veterinary incomes from drug sales, it is indeed this economic model that has allowed the development of such approaches within industrial livestock farming. modern strategies for reducing antibiotic use should therefore seek less to renew the professional expertise of veterinarians than to find new ways to valorize it economically. the french veterinary profession is currently undergoing major changes; or at least it tends to see itself as being at the heart of a period of major challenges that are pushing it to reinvent itself. this is not the first time that it has had to face such a reflexivity test, even in recent history (in the british case, some historians even see a cycle-woods 2011a), but recent literature produced by professional veterinary organizations shows the importance of what is currently perceived as a need for self-analysis and change (ondpv 2018; vetfuturs france 2018) . there are several issues that might explain why this period is favourable to such a prospective assessment of veterinary futures. one of the most important concerns the controversies and public policies that have developed over recent years with regard to the issue of antimicrobial resistance (amr), which has directly challenged the economic and professional model of farm animal veterinary practices that were setting drugs (mostly antibiotics) up as a cornerstone of veterinary activity, as a source of both income and professional expertise. cross-fertilization of research on the veterinary profession and drug regulation is not common. although veterinarians have aroused the interest of certain historians and sociologists of professions, this has essentially been in relation to the analysis of this social group's process of professionalization (berdah 2012; mitsuba 2017) , its role in animal health or food safety policies (woods 2011b; enticott et al. 2011; fortané 2016, 2018) , the dynamics that contribute to its specialization (gardiner 2014) or feminization (surdez 2009 ), or finally to knowledge and professional practices in farm (shortall et al. 2016; ruston et al. 2016 ) and small animal medicine (sanders 1994; morris 2012) . as for the regulation of veterinary drugs, there are also several studies by historians on the vaccination of animals against major zoonoses or epizootic diseases (woods 2004; berdah 2018) , sometimes on the veterinary pharmaceutical industry (corley and godley 2011) , and more recently on the amr issue (kirchhelle 2018 ). yet unlike the uses of human medicines that medical anthropology has been able to theorize and document for many years (whyte et al. 2002) , the uses of veterinary medicines, i.e. the conditions under which they are prescribed, dispensed and used, are rarely studied, except in interdisciplinary literature from the field of veterinary sciences (speksnijder et al. 2015; coyne et al. 2016) . this article tries to open a way to cross-fertilize these reflections. using recent debates on the amr problem, it proposes to examine the relationship between the development of professional veterinary expertise and of the drug market, based on the case of a specific segment of the profession, namely veterinarians specializing in industrial poultry and pig production in western france. it thus puts the amr issue under a broader lens as it analyses ongoing changes within the veterinary profession not as potential consequence of recent measures aiming to reduce antibiotic use, but rather as a reason for the way the problem is now framed. indeed, it is common to hear professional organizations or public authorities state that in order to reduce their economic dependence on antibiotic sales, vets must rethink their activity by favouring preventive approaches to animal health which would involve a diversified range of services and would contribute to placing vets in an advisory role with a holistic vision of livestock farming or even of the food supply chain (vetfuturs france 2018) . however, such a form of professional expertise, combined with a particular business model for the practices promoting it, is not fundamentally new. if it is at the heart of contemporary debates, it is because it is based and supported by far earlier dynamics that initially had nothing to do with amr, but which used the opportunity of current controversies surrounding antibiotic use, sale and prescription to reinforce and legitimize a certain vision of veterinary medicine, based on preventive approaches to animal health. poultry and pig medicine in industrial production is an especially interesting area for an analysis of these dynamics. firstly, because approaches to animal health that have developed in this field are very singular and characteristic of the intensive farming methods used in these sectors, particularly in the brittany and pays de loire regions where a large proportion of the production is located. indeed, a certain vision of preventive medicine was developed by pig and poultry vets in the 1970s and 1980s, even if the issue of this form of veterinary expertise has not always been raised in these terms. this article therefore aims to understand the professional knowledge, practices and economic model upon which this kind of expertise is based, and why the current amr context is an opportunity to expand it (or at least attempt to). secondly, pig and poultry vets make up an extremely small and autonomous segment of the profession (which makes it possible to draw up a fairly representative picture) although its homogeneity should not be overestimated. this article thus seeks to provide a thorough analysis of this very particular part of the veterinary profession, the specificities of which have almost never been addressed by the literature. despite their small numbers, "industrial" vets are nevertheless an essential component of the profession, because they manage the health of an economic sector which supplies a considerable share of national animal production. the article opens with a brief presentation of the political context and controversies surrounding the amr problem, and how the french veterinary profession has faced up to this by defending the preventive medicine model. it then describes this form of expertise in the professional segment studied here, showing why pig and poultry vets chose this specialization. the article then looks at the origins of these preventive approaches to animal health, both in terms of knowledge and practices, and the economic model associated with it. finally, it reviews the strategies currently developed by industrial vets to adapt to the constraints of increased control of antibiotic use in livestock. the amr problem and the preventive "solution" the problem of antibiotic use in livestock farming is not new. as soon as these molecules were introduced in agriculture in the late 1940s, there was controversy concerning the development of resistant bacteria in animals and food, and the risks of human contamination (bud 2007 ). yet for several decades, this issue has been eclipsed by the belief in a permanent renewal of the therapeutic arsenal, consisting in thinking that the continuous discovery of new antibiotics would compensate for the development of increasingly resistant bacteria (podolsky 2018) . after the swann report in 1969, a series of measures to control the use of antibiotics as growth promoters 1 was nevertheless adopted in europe, progressively separating the molecules used in agriculture and human medicine (kirchhelle 2018) . but 20 years later, during the avoparcin crisis 2,3,4,5 these measures were considered ineffective (in the sense that they did not prevent the transmission of resistant bacteria between humans and animals) and the use of antibiotics as growth promoters was finally banned in the european union in 2003 6 (kahn 2016) . the problem of antibiotic use in livestock farming as we know it today reemerged in the late 2000s, this time focusing on veterinary uses, i.e. on curative or preventive uses with veterinary prescription (fortané 2019) . veterinarians were directly accused of being responsible for the overuse and misuse of antibiotics (and therefore for the spread of resistant bacteria) on the basis of a fairly simple argument: their supposed professional "conflict of interest". indeed, in france, since the 1975 act on veterinary pharmaceuticals, vets have had a dual monopoly on the prescription and supply of medicines (hubscher 1999) . even if, in theory, delivery is shared between three beneficiaries (veterinarians, pharmacists and approved co-operatives), vets capture the vast majority of the curative drug market (of which antibiotics constitute the main category) (guillemot and vandaële 2009) . the argument that veterinarians over-prescribe antibiotics in order to increase their incomes then became the main framing of the amr problem. this construction of the problem was in reality carried by a coalition of human health actors (doctors, pharmacists, health administration) whose political agenda was twofold. on the one hand, they defended a measure that crystallized the debates around the years 2009-2013: the "decoupling" of prescription and delivery, which consists in applying the professional and economic model that prevails on the human drug market, i.e. reserving prescriptions to physicians and sales to pharmacists. decoupling basically means forbidding veterinarians from selling pharmaceuticals, as is the case in countries such as sweden or spain (fortané 2016) . on the other hand, this coalition supported the concept of "critically important antibiotics", the principle of which is to reserve certain molecules, in particular the latest generations of antibiotics, for human medicine. from a political and institutional point of view, this period was extremely interesting because it put the spotlight on definitional and jurisdictional conflicts between different social groups for the control of the legitimate use of antibiotics. it finally ended in a relative victory for veterinarians who succeeded, at the end of an unprecedented mobilization, in reversing the stigma that human health stakeholders assigned to them. indeed, vets have been able to impose the image of a profession that is not guilty of overusing antibiotics but which is instead accountable for their proper use. the notions of prudent, judicious, rational or responsible use, now widely used in amr debates, are thus a social construct produced by conflicts between social groups for the definition of the legitimate use of antibiotics (fortané 2019) . this veterinarian victory led to the withdrawal of the two emblematic measures (decoupling of prescription and delivery; ban on critically important antimicrobials) supported by the coalition of human health actors. in return, a stricter framework for the use of antibiotics in animal husbandry was implemented between 2014 and 2016: margins on the sale of antibiotics are now limited and the retail price of antibiotics must be the same for every client, 7 and antimicrobial susceptibility tests are mandatory for the prescription of critically important antibiotics. 8 but once again, the most important part of this victory certainly concerns the changes regarding the image of the profession. indeed, veterinarians not only reversed the stigma and positioned themselves as guardians of antibiotics, they were also able to re-appropriate the problem by highlighting the way they could solve it. without denying that the economic model of the profession was too financially dependent on antibiotic sales, and that antibiotic use may have been too prevalent in animal care in the past, vets started to defend the development of preventive approaches which would, according to them, be the only way to ensure the transition towards an economic and professional model guaranteeing responsible use of antibiotics. this view was reinforced and supported by farmers and public authorities who also called for the development of such approaches which are usually promoted by national amr policies ( fig. 1 ) in france and in other countries (badau 2016; piquerez 2019) . 9 at the heart of this new prospective narrative for the profession and its role in managing the amr problem, we can observe the construction of a (supposedly) new conception of animal health, which is not based on a strictly clinical approach to diseases but on a holistic vision of animals and livestock farming (biosecurity, hygiene, nutrition, good husbandry practices, etc.). in the posters above, veterinarians are portrayed as the gatekeepers of such an approach, through their transversal role as "health advisors". this professional model, based on the knowledge and practices of preventive approaches to animal health, goes hand 7 law for food, agriculture and forestry n°2014-1170 of october 13th, 2014. 8 decree no. 2016-317 of march 16th, 2016. 9 when public authorities and the veterinary profession began to conceive this campaign, a third poster was designed, saying: "my vet is much more than a mere drug supplier, he is also a teacher! he prescribes the medicines i need … but above all he talks to my farmer to make sure that i am perfectly fed, sheltered and vaccinated". however this poster was not retained for the campaign because of its relatively sensitive headline. in hand with an economic model where the incomes of veterinary businesses would be more diversified since these new "health advisors" would be able to monetize a wider range of goods and services (hygiene and nutrition products, bacteriological analyses, livestock audits, etc.) than just pharmaceuticals. could we look beyond the symbols and images of professional and political discourses and see whether these preventive approaches that might change the way veterinary medicine is practiced, and how antibiotics are used, rely on actual knowledge and practices and, if so, where and since when? the fact is that this model of preventive veterinary medicine seems in reality to be quite typical of a very particular segment of the veterinary profession, the one this article proposes to describe, namely industrial vets. the remainder of the article thus seeks to address the following two questions: is this form of professional expertise really perceptible in the field, and is its development truly linked to the global context of the amr problem and to recently implemented policy measures, or does it have other origins and raisons d'être that might actually 0 when public authorities and the veterinary profession began to conceive this campaign, a third poster was designed, saying: "my vet is much more than a mere drug supplier, he is also a teacher! he prescribes the medicines i need … but above all he talks to my farmer to make sure that i am perfectly fed, sheltered and vaccinated". however this poster was not retained for the campaign because of its relatively sensitive headline. these two posters were used in the french 2016 amr policy ("plan ecoanɵbio"). they are a perfect illustraɵon of how the veterinarian's role was reframed towards prevenɵve approaches to animal health (in parɵcular vaccinaɵon), as a means to reduce the use of anɵbioɵcs. the first poster (on the leō) says: "my vet is far more than a mere emergency doctor, he is an expert contribuɵng to good husbandry pracɵces". the second poster (on the right) says: "my vet is far more than a mere hands-on man, he is an advisor, always there to prevent and vaccinate". both conclude with: "ask your vet for advice" 9 . when poultry or pig vets are asked why they chose this specialty, they often point out a huge contrast between what they do and the way they perceive cattle (i.e. "rural") or companion animal vets. they feel that being an "industrial vet" essentially relates to four characteristics. firstly, veterinarians specialized in poultry or pig medicine attach importance to working at the heart of the agri-food system, with livestock farming professionals. many of them have agricultural family origins and believe that they chose this profession in order to maintain a strong link with the rural world. they perceive farmers as animal experts, unlike pet owners, and see their activity as teamwork alongside skilled professionals, with whom it is easier to interact and who can also provide them with knowledge about animals. their clients are therefore also their partners in animal health management: they can trust them, rely on them and delegate tasks to them. secondly, pig and poultry vets consider their work to be a permanent renewal, as opposed to the repetitive and sometimes boring work of cattle vets who constantly reproduce the same gestures and who are rarely motivated or intellectually stimulated by new situations and new challenges. most of them, whether they are young vets or already have 15 or 20 years of experience, tend to compare their professional activity with the image of the "emergency vet" (or "fire brigade" vet), available day and night to care for sick animals. from their point of view, this traditional activity is typical of cattle vets, corresponds to the past and relates to a type of work which is limited to clinical diagnosis, drug prescription and/or surgery (e.g. midnight calving). in their opinion, the only objective of such a way of working is to ensure that clients are satisfied with an occasional intervention and that they will once again call upon veterinary services the next time health problems occur. the point here is not to claim that this is what cattle vets are actually doing, 10 but simply to note that this narrative is widely used to define, by contrast, the professional identity of industrial vets. i am indirectly from a rural family, that is to say that my grand-parents, my uncles, they all worked as farmers, in mixed farming and mixed animal farming, from both sides of my family. it is just my parents who had access during the postwar period to national education programmes and became teachers or researchers. so i am from the third generation but i spent a lot of time at the farm. 10 although some of the characteristics that pig and poultry vets tend to associate with cattle vets have actually been observed in other studies, in particular the fact that cattle vets do not have a high opinion of the technical expertise of the farmers they deal with (shortall et al. 2016) . these opinions of their clients nevertheless depend to a certain extent on the type of farm and farmers they relate to: cattle vets have better professional relationships with "commercial farmers" (managers of large, modern and business-oriented farms) as the latter tend "to understand the need to use the vet as a disease prevention consultant rather than to treat individual sick animals: i.e. part of vets' desired move from a 'test and treat' to a 'predict and prevent' model of veterinary intervention" (shortall et al. 2018, p. 597) . in addition, this shows that a tension between "fire brigade" work and advisory veterinary work is also present within a certain section of the cattle vet profession. i became a vet to take care of farm animals because i like being outside and having an intellectual occupation. so that is the reason why. so at vet school, you didn't think about working with dogs? certainly not (laughs)! that was my nightmare! when you work in the pig sector, what is good is that you work 100% on farms. because when you work with cattle, most of the time you have to do some small animal work as well. and what i want is to work with professionals who are producing the animals we eat. pig farmers (well, not all of them but most of them) have very good technical skills. pig farming is a very dynamic sector, you never get bored! we often have to upgrade the farms to new standards so it's always evolving. research also moves very fast, the pharmaceutical industry innovates a lot so it's interesting. (…) and to be honest, what interests me the most is this kind of follow-up, not being a "fire brigade" vet. but i can't talk very knowledgeably about the "classic" rural vet as i never was one. but what i do think about this kind of practice is that there is a lot of emergency and "fire brigade" work and i don't think that is very challenging. in pig production, we have been evolving for quite a long time because if we only do "fire brigade" work… well, farmers expect a lot more than that! and for us, that is also what's interesting. pig farming is a batch production, so we have this tendency to always try to do something for the next batch, try to prevent future batches from getting the disease we have now. so having to try this and that is always very dynamic and challenging. so we know that there is always a new batch to come, and that is not the same in cattle farming. thirdly, pig and poultry vets describe their work as being part of an epidemiological rather than clinical approach to animal health, which they associate with the idea of a preventive rather than curative or therapeutic approach to diseases. this conception is closely linked to the two previous points. on the one hand, it refers to the professional proximity that veterinarians maintain not only with their clients but also with technical advisors, often employees of the co-operative to which the farmer belongs or sometimes of a feed mill company. indeed, in pig and poultry farming, technical advisors play an important role as they visit the farms much more frequently than vets. although their role should not involve animal health issues (but rather feeding, housing or husbandry practices), they tend to conceive these aspects as a whole, as do vets when they look at the technical factors (rather than biological and medical ones) of diseases. animal health is thus the work of a professional trio, whose theoretically distinct roles often overlap, encouraging situations of cooperation and also sometimes conflict (adam et al. 2017 ). on the other hand, and consequentially, this so-called epidemiological conception of animal health establishes a form of activity which is not limited to a clinical and individual approach to pathology but which, on the contrary, opens up professional expertise to areas often considered as not specifically veterinary, such as hygiene, nutrition, zootechnics or biosecurity. the knowledge and the toolbox of industrial vets are thus extremely varied and cannot be limited to surgery tools or prescription booklets. the veterinarians interviewed often emphasized the "evidence-based" side of their professional expertise (and legitimacy), which is based on the collection and analysis of various data and promotes intellectual stimulations regularly renewed by the diverse situations they must face. they believe all these elements to be part of preventive approaches to animal health and they tend to perceive themselves as health advisers, or "health managers" as thoms (2015) has shown in the case of german poultry vets. i graduated from veterinary school in 1985. then i got a degree in epidemiology. what i realized and particularly interested me during my studies was the difference between individual medicine and herd medicine. so with my education and also my family farming background i realized that individual medicine was more a cost than a profit. so i tried to develop this kind of herd care and preventive approach. i started my professional life in rural medicine in a region that was quite a pioneer in this type of approach, especially regarding reproduction. we have audit grids and we have exploration tools. for example, i have a device to measure co 2 , to dose carbon monoxide. i have a ph-meter, a conductivity meter, a laser. i have a burette to measure the water flow in the feeding system. i have a light meter in case i need it. i have a camera to explore water pipes. i always have smoke bombs in my car to check ventilation in the buildings. so really, we have exploration tools in the form of scissors and gloves to perform autopsies but also equipment for managing the buildings. but aren't you stealing work from the technicians? technicians do autopsies (laughs)! no, it's true that controlling ventilation doesn't mean that i am able to deal with the farm it system. so technicians keep their skills. but i consider that it is my job to explain to the farmer that if he has a colibacillosis it's not because a germ fell from the sky but because his ventilation system doesn't work the way it should. that's part of a vet's job, even though it is the technician who is able to fix the problem. we are doing what i might call an etiological or epidemiological diagnosis. fourthly, the characteristics of pig and poultry medicine must also be related to the type of animals with which these vets work on a daily basis. without necessarily following the epistemological principles of the "animal turn" promoted by certain branches of the social sciences (guillo 2015) , which consists in analysing the way in which animals themselves shape human interactions or the socio-technical devices within which they take place (notion of "animal agency"), it must be noted that taking care of chickens or pigs has different implications than is the case with cattle, dogs or cats. two elements seem central here and are directly related to the characteristics of pigs and poultry and the farming systems to which they belong. on the one hand, these animals have a relatively short lifespan: only about 30 days for a conventionally raised chicken (between 90 and 120 days in organic or label farming) and about 6 months for pigs (the sows being slaughtered after 2 or 3 years). it is therefore very different from a dairy cow that must remain healthy and productive for an average of 5 years, or pets that live 10 or 15 years, sometimes more. in these conditions, which obviously also depend on the economic structure of the agri-food industry, animal health becomes part of biological and temporal dynamics where the slightest disorder may have pathological consequences that might be seen as "just-in-time diseases", i.e. health issues directly related to, or shaped by, the sociotechnical and economic infrastructures (including the animal bodies themselves) within which they emerged (allen and lavau 2015) . on the other hand, the health of chickens and pigs is not considered individually, but rather from a population perspective. these are animals that are reared in batches and it is the group of animals that constitutes the epidemiological reference unit for both the vet and the farmer. in certain cases, particularly in poultry farms where the production cycle is very short, many decisions are therefore made not for the current batch but for the following one, in order to avoid repetition of the same problem. overall, all of these aspects are typical of industrial veterinary medicine and, therefore, of this form of expertise and professional legitimacy that can be qualified as preventive veterinary medicine. to sum up, and without prejudging whether or not these elements can be found in other segments of the veterinary profession, we can say that poultry and pig medicine is characterized by (i) a vocation; (ii) a technical and preventive approach; (iii) tripartite work; (iv) animals and farming systems that "call for", or co-construct, this form of expertise. however, when listening to veterinarians talk about their profession, this way of working on industrial farms does not seem to be linked to the rebuilding claimed by the profession since its incrimination in the amr problem, even though certain policy measures recently implemented may encourage the development of this preventive and evidence-based approach, such as the obligation to perform antimicrobial susceptibility tests before prescribing critically important antimicrobials (bourély et al. 2018) . indeed, the roots of this particular form of veterinary expertise can be traced back to the mid-1970s/early 1980s, when some pioneers began to specialize in pig and poultry farms-at that time during a massive industrialization and intensification process. the origins of preventive approaches in "industrial" veterinary medicine during the 1960s and 1970s in france, particularly in brittany and pays de loire, poultry and pig farms expanded rapidly (nicourt 2013) . the mixed crop-livestock model was gradually being replaced by specialist farms which were developing through a twofold process of intensification (increased herd size, confinement and containment of animals, rationalization and (bio)technicization of husbandry methods such as genetics, feeding and pharmaceuticals) and industrialization (concentration and vertical integration of the food chain's stakeholders, and taylorization of farm labour) (diry 1985) . this movement led to the emergence of new needs in terms of health management. firstly, at that time veterinarians knew little about these animals in terms of medical knowledge or techniques (poultry and pig medicine was rarely touched upon in veterinary schools back then). secondly, the confinement of animals in enclosed buildings led to outbreaks of disease, especially-but not only-infectious diseases which were unknown or at least whose management methods were no longer appropriate. a small number of vets then began to specialize in this type of production, viewing it as a promising and developing market. poultry farming was not taught at all. poultry diseases even less so and the prevention of poultry diseases even less than that. so my gateway to poultry farming was the technical-economic approach to things, and not the purely pathological approach, which i could have done at the time on sheep or cattle. except that there, i was in a completely capitalist system. as you can see, one can adapt to anything (laughs). in the pig sector, it was mainly vets employed by agricultural co-operatives who embarked on the adventure, supported in particular by the creation of the station de pathologie porcine (spp) (national veterinary laboratory specializing in pig health) in ploufragan (north brittany) in 1975 (fortané 2017) . in the poultry sector, which is structured more around industrial groups than around co-operatives, there were more independent vets who were orienting their practices towards this type of clientele. what these veterinarians had in common was the conviction that they could no longer do their job in the traditional way, i.e. as emergency "rural" vets, alone or in partnership in small practices serving a diversified clientele (cattle farmers and pet owners in particular, with a few occasional interventions on poultry or pig farms). moreover, they were convinced that their job was to accompany the development of intensive and industrial livestock farming by providing services adapted to the specific animal health issues of this sector, and that their own organization had to follow and mimic the development of the industry they were working for. the term "industrial vets" therefore refers not only to the kind of clientele they were (and still are) working for, but also to their own state of mind and conception of veterinary medicine, as a profession and as a business. in fact, we were the defenders, the propagators of intensive livestock production. we never denied that, that's what we were employed for. we were just saying "ok, but the conditions for success are this and this and this". at the time, there was a double-digit growth in the sector, so we said: "we must stick to the development of this sector". we were focused on poultry, we had almost given up on pigs. we said: "we have to stick to the leaders". very quickly, in 84-85, we said: "it's x [one of the biggest poultry industrial group], we'll stick to x, we'll stick to the growth of x, like a leech, we'll never let it go". that was our aim in 84-85 and it became concrete a little later when we were working for x, even though we weren't their official vets. (…) we had developed well in 84-85, so how could we consolidate our system, how could we really sell it, how could we duplicate it? this is where the notion of "global offer" began. here, you find three activities: the veterinary practice, including advice and training, the medicines and the analysis. when you come here, you have this "global offer": the lab part, the drug part and the advisory part. the way this pioneering generation of pig and poultry vets worked was characterized by their capacity and interest in conceptual innovation and do-it-yourself techniques for responding to the often unknown situations they had to face. on the one hand, it concerned the development of pharmaceutical products adapted to these animals and their husbandry conditions, at a time when the pharmaceutical industry was producing very few medicines for this type of livestock, not considering it to be a worthwhile market. it was therefore not uncommon for veterinarians to order pharmaceutical raw materials (rather than manufactured drugs) in order to themselves prepare a product that could be effective in terms of both pharmacological (e.g. combining an antibiotic and an anti-inflammatory) and galenic properties (e.g. designing a drug in the form of a powder to be spread in bedding rather than to be mixed with food-for dermatological infections in particular). in 83-84, we had outbreaks of staphylococci. obviously, the animals caught it very early because they already started to have small pimples at 8-10 days. so the mother was a carrier, so we thought: "what if we use an antiseptic powder?" then we thought: "we need a powder that is very powdery, that adheres well to the skin". we worked on the galenic, excipients and everything. we developed a powder and tried it and the farmers said: "this works well". who were you doing this with? we made the powder ourselves. we had a mixer. in our pharmacy, we had a manufacturing workshop. (...) it was common at the time. we had the right to make extemporaneous preparations that were the result of our prescription and our imagination. on the other hand, this broader conception of the veterinary role could be seen in the holistic vision of health that these professionals were promoting. most of the time, they combined therapeutic intervention with bacteriological analysis and research on animal nutrition (some of these vets were also employed by feed mills). all of the larger practices of this pioneering generation developed a laboratory, sometimes a makeshift one, in order to perform the autopsies necessary for bacteriological tests. x [a pioneering poultry vet] had a very solid clientele of horses and cattle at the time, but he was interested in birds. (...) when the first industrial poultry farms, the large flocks of 300 to 400 chickens, settled in, he moved towards that. very quickly, he realized that the bigger the batches of poultry became, the more the diagnosis had to go through the laboratory in order to be accurate -at least regarding bacteriology and parasitology. since he's a guy who doesn't want to do things by half, he thought: "i can't make a laboratory like that, we suck as vets, i need training". he said to his partner: "i will go to pasteur". 11 further education at that time was unimaginable. (...) then, when he came back, he started his own lab and that was really the beginning of the adventure. in the pig sector, it is interesting to note the extent to which the history of the ploufragan station is still rooted in the north brittany territory, as many generations of veterinarians, still today, have done part of their training there or continue to rely on the expertise of spp's epidemiologists or microbiologists in their daily activity (especially in the case of outbreaks of infectious diseases or to set up clinical trials). oh yes, with regard to emerging diseases, i often contacted the station when we saw pathologies that seemed new to us. this was the case for many diseases. there was streptococcal, there was respiratory coronavirus. at the spp, i called [the vet in charge] and the others, they were not convinced because at that time there was still pseudorabies in the farms. i had to talk to [a pharmaceutical company] about it and then the spp became interested. indeed, veterinary researchers from the spp had developed a preventive approach to animal health called "ecopathology" that had considerable success among pig vets and farmers in the 1980s (fortané 2017) . it was mostly based on an epidemiological conception of animal health (disease outbreaks are related to multiple variables, particularly "technical", i.e. non-clinical ones), and was adapted to the specific issues and husbandry conditions of intensive livestock farming. in this regard, the type of preventive veterinary medicine which was developed at that time in france would seem to have similar characteristics to that which had flourished in the uk slightly earlier (from the 1950s to 1970s) (woods 2013) : animal health has to be conceived at the scale of the herd (and not the individual animal), be articulated to non-medical matters such as feeding, housing and genetics, and integrate the economic issues of performance and profitability within the advisory support vets provide to their clients. the most important difference during the process of institutionalizing these preventive approaches in france and uk seems to be the role played by public authorities. while in the uk, the state firmly supported the development of preventive veterinary medicine as a way to accompany the modernization of british livestock farming, in france, except for the spp which was partially founded by local authorities and state veterinary services, most of the preventive approaches that flourished in the 1980s were supported by the private engagement of pioneering vets trying to respond to the challenges of a growing industry and, in the meantime, to capture the market of "industrial" veterinary services. 12 all in all, the point here is not to say that contemporary pig and poultry vets are nowadays using the exact same techniques and knowledge of the preventive approaches developed in the 1970s and 1980s, or even to suggest that all industrial vets of that time were doing exactly the same. 13 it is rather to demonstrate that this type of veterinary medicine has at least a 30 to 40-year history in these sectors and that some aspects of this can still be seen in the way that modern industrial vets continue to work (and to perceive their job and professional identity), such as the importance of technical expertise, epidemiological knowledge and bacterial laboratory-in other words, that a diversified form of veterinary expertise is required. even though this history has more or less vanished from the profession's official memory, a few unconscious vestiges of it can still be seen in recent amr debates. for example, during the campaign promoting the amr policy in the pig sector in 2014, an implicit link was made between this pioneering medicine and the kind of expertise which is now considered to be the future of the profession and through which veterinarians tend to legitimize their role as the guardians of antibiotics. this is demonstrated by the utilization of the ecopathology icon which expresses the holistic conception of animal health (namely the so-called "ploufragan hexagon") to promote good veterinary practices in antimicrobial use in livestock (fig. 2) . all these elements show that what is now called preventive veterinary medicine and which is perceived, within the political context of the amr problem, to be the model of professional expertise towards which veterinarians must turn, is not fundamentally new, at least within the knowledge and practices of pig and poultry vets. but what about the economic model on which this kind of expertise is based and how it is actually related to the use of antimicrobials? although it is in fine difficult to associate the development of preventive approaches with the reaction of veterinarians to their incrimination on the amr issue, contrary to what the sole examination of professional discourses in the press or in various 12 it should also be noted that in the uk the development of preventive approaches concerned farm animal medicine in general (so mostly cattle medicine), while in france this movement was located in "industrial" pig and poultry medicine -although due to local history it would seem that in some regions other forms of preventive medicine were also developed in a small number of cattle veterinary practices as from the late 1970's (combettes et al. 2012) . 13 this is indeed the purpose of talking about "preventive approaches" in the plural, i.e. to highlight the fact that these approaches could be rooted in different local histories or individual trajectories yet still share some common principles. while pig vets were mainly referring to ecopathology by virtue of their link with the spp, poultry vets did not use a specific term (preventive, holistic or epidemiological approach, global offer, etc.). nevertheless all industrial vets were (and still are) referring to what they were (are) doing with the same kind of contrasting schemes: epidemiological vs clinical, preventive vs curative, herd vs individual, technical vs medical, etc. which i consider to be the common ground of "preventive approaches to animal health" with, in addition (cf. next section), an economic model based on free-of-charge services funded by drug sales and aimed at "capturing" clients. institutional spaces relating to amr policy-making might suggest, nowadays this model has nevertheless become more visible and contributes to the re-legitimation of the veterinary profession (badau et al. forthcoming) . however, the idea that preventive approaches will make it possible to move away from the economic model placing antibiotic sales at the heart of veterinary income seems more dubious, as the development of said preventive approaches was in fact based on this very system. by the notion of economic model, we mean the way in which veterinary services are monetized. one of the key elements of the criticism against veterinarians with regard to the amr issue has been the fact that their sources of income are largely dependent on drug sales (antibiotics in particular). however, whereas professional discourses seek to explain that the development of a preventive medicine in which antibiotics no longer hold a central place will make it possible to change the "business model" of veterinary practices, it is on the contrary clear that this economic model based on drugs sales has in fact allowed these preventive approaches to flourish. even though it is difficult to concretely assess the accountability of veterinary practices due to a lack of available data (in particular if we wish to distinguish between different segments of the ploufragan hexagon: animal health is dependent on six variables (tillon et al., 1980) 1 prevenɵon campaign against anɵbioɵc misuse in pig farming, 2014 2 fig. 2 the common principles of past and present preventive approaches in industrial veterinary medicine. 1-ploufragan hexagon: animal health is dependent on six variables (tillon 1980) . this famous diagram represents the way animal health (and economic performance of the farm) is conceived within the ecopathological framework: it is correlated to six variables, namely the farmer, the animal, housing, feeding, microbes and husbandry practices. at that time, it was considered quite innovative to claim that veterinary expertise should take into consideration all these aspects and not just focusing on animals and microbes (fortané 2017 ). 2-prevention campaign against antibiotic misuse in pig farming, 2014. this poster used in the amr campaign perfectly reproduces the six "ecopathological" variables (here referred to as the six "pillars" of animal health). it says: "no more antibiotics than needed. with my vet, i manage the health of my animals while limiting the use of antibiotics". this poster therefore illustrates how preventive approaches within veterinary medicine are deemed to be a solution to reduce antimicrobial use profession), we can nevertheless rely on estimates from various literatures. a 1984 veterinary thesis on the structure of the drug market considered that drug sales were generating 30 to 40% of the income of farm animal practices (rivière 1984) . thirty years later, in the midst of the amr debates, a report by the french food, agriculture and rural areas council estimated this figure to be around 60% for farm animal vets in general, and up to 80% (including a high proportion of antibiotics) for industrial vets (dahan et al. 2013, p. 20) . so one cannot help but wonder what happened during the three last decades and to what extent this evolution of the economic model of veterinary activity has in fact supported the development of preventive approaches to animal health, contrary to the prospective narrative the profession has constructed to defend its role as the guardian of responsible use of antibiotics. the answer is actually quite simple, and is a classic case in agricultural economics: the sale of inputs funds the advice. in the case of industrial vets, the context in the 1980s and 1990s favoured the emergence of strong competition between practices seeking to capture the growing clientele of pig and poultry farmers. this competition was also heightened by the fact that the supply of veterinary goods and services was still highly heterogeneous, due to the non-existence of standardized pharmaceutical, hygienic or nutritional products adapted to the new health problems caused by the intensification of animal husbandry. consequently, pharmaceuticals, which are the only product on which veterinarians have a monopoly for prescribing (and therefore advice) and dispensing (and therefore sales), became the main means of ensuring client loyalty. since the 1975 act, the veterinary drug market has thus been developing into a captive market, where drug sales have gradually become the only monetized exchange between vets and farmers, funding all the services actually offered by veterinarians (visits, diagnosis, analyses, prescription, audit, etc.) (bonnaud and fortané 2018) . indeed, preventive veterinary medicine developed on the basis of this economic model which corresponded, on the one hand, to that of the few pioneering independent practices which owned bacteriology laboratories, small drug manufacturing factories or wholesale companies and, on the other hand, to that of the agricultural co-operatives which the 1975 act established as lawful drug suppliers and which, until 2007, employed veterinary practitioners in this regard. indeed, it was precisely by providing a whole range of "free of charge" services (visits, advice, vaccination, etc.) that these vets were able to disqualify their competitors, i.e. their "emergency" or "fire brigade" colleagues who offered little advice but were able to monetize their occasional interventions. yet were the farmers prepared to pay for their visits, as they would have done with rural vets at the time? because, even the independent vets specializing in group pathologies, so pig and poultry, like those from x [one of the pioneering veterinary practices in industrial medicine], most of the time, they did not charge for their visit, their visit was paid through drug sales. in fact, most farmers were smart enough to understand that if they were satisfied with the services of a vet or a veterinary practice, they should buy drugs to pay them indirectly, even if there were price differences. i think it was like this: "you do me a favour, i need you, you treat, you prevent, we work together, i buy drugs". it was indirect payment. (…) i've always thought that the best way to ensure farmers' loyalty is to be efficient. then they will not even talk about drug prices. there was a time when vets were very active on the farms. if farmers were happy with our services, [they didn't care about] the price of the drugs. at one point, we were a little cheaper but we weren't always cheaper. we were also known for being reasonably priced, so that there wasn't too much competition. sure, we weren't the cheapest on the market, but that's not the issue. but i'm convinced, and i know that's the way it was for a lot of people, the best way to keep them was to meet their expectations. their expectations were simple: "i have problems, i have to solve them, help me solve them, if it takes time ok it takes time". it was up to us to meet farmers' expectations in terms of health. if the farmer was satisfied, the rest would follow, he would take the feed, the genetics, the drugs. we had an indirect commercial role, but we could only do it if we were efficient. (…) but some of them stopped buying our drugs. they told me: "you're too expensive". so i answered: "if we are too expensive, you go elsewhere, but you no longer have the services". what does he want? a cheap price, but with a service that will be what it will be, it may be very good, ok, but the farmer is free to choose. however, he fully understood that he could not ask us to come ten times a year or twenty times and then buy zero medicines. moreover, i tell most people: "we provide you with important services, we know that you also like to work with other people from time to time, at x for example, so don't buy everything from us, but buy a little from us, as we come to see you regularly and we need a return". and then everything was fine, they bought some elsewhere and some from us. the construction of a captive market is therefore a central dimension of what i call here "the economic structures of professional expertise". the development of preventive approaches within industrial veterinary medicine is not only the result of the emergence of new demands or needs from farmers, but it is also and above all the consequence of the intra-and extra-professional competition that veterinarians have had to face after the 1975 act. prescribing and dispensing pharmaceuticals, combined with a holistic expertise, is a means of capturing a clientele through a simple and exclusive form of economic contracting (the sale of drugs) while providing a wide range of services (diagnosis, advice, etc.). it is indeed the articulation of these two dimensions that makes it possible to prevent, or at least reduce the risk, that clients are captured by a competitor. in this sense, preventive approaches to animal health must be read simultaneously as a professional and a business model, these two dimensions being constitutive of each other. such economic structures of veterinary expertise have been observed in another context, that of the usa in the 1950s and 1960s. smith-howard (2017) shows how american veterinarians, who did not have a monopoly on the sale of medicines, gradually established themselves as the main distribution channel by coupling their prescriptions with preventive services that provided real added value to farmers, compared with pharmacists or other retailers who were merely supplying the drugs. a linguistic distinction has even emerged between "dispensing" and "merchandising", contributing to (re)establishing the legitimacy of veterinarians on a new basis. in the end, it is therefore in the light of this link between professional expertise (in the sense of type of knowledge and services) and business model (in the sense of how these services are economically valorized) that we must question the current context and controversies surrounding amr. it seems finally a little too simplistic to support the thesis of the professional conflict of interest that human health actors have mobilized to point out veterinarians' responsibility in the amr problem, because of their dual monopoly on the prescription and supply of medicines. in fact, the structure of the veterinary drug market has set up the sale of pharmaceuticals as a quasi-unique way of making professional expertise profitable, even though this expertise was already based on preventive approaches (and thus diversified services) for quite a long time. the current challenge for the veterinary profession is therefore less the development of a preventive medicine than the renewal of the economic model on which it has been based until now. in this regard, it is actually interesting to note that the vets i met in this study clearly mention this issue and have already started to develop strategies to deal with it. that's exactly the issue we now have with my colleague. i'm right thinking this. historically in poultry, no charge is really made for visits. overall, payment is made through [the purchase of] medicines, which is, for me, not a bad thing if it is not excessive. at some point, when we take the example of human medicine, you pay a doctor, you pay a pharmacist, everyone is happy because there is healthcare security. the day there's no healthcare security, i don't know what we're going to do. so that's what i explain to farmers: "the day you have to pay a vet and a pharmacist it's going to cost double". so at the moment, with this system, you only have one cost, so in theory that's good. indeed, there can be misuse, someone who would systematically prescribe medicines even if there is no need. this is clearly an issue and it is difficult to explain, that's our problem. but honestly, every morning when i stand up, i don't say to myself: "i will prescribe 10 kilos there, 15 kilos there and 5 kilos there, and the day is done". that's absolutely not the way i think, but i can understand… and i understand when people are telling me: "yes, but if you sell, you earn". that's all there is to say. it's been on my mind for a year. firstly, particularly in a context where profits on medicines are now more strictly controlled (and therefore more limited), it seems essential for the veterinary profession to be able to charge for services that were previously funded by drug sales. this is the case in particular for visits, technical advice and bacteriological analyses (the latter were generally charged but often below the real cost-which could be high for the veterinary practice that has to employ laboratory technicians). secondly, there are strategies consisting in investing in hygiene and/or nutrition product factories (in particular disinfectants and food additives), in order to be able to sell products other than just pharmaceuticals. it is a strategy of diversification of the professional and economic activity that pioneering independent practices or certain cooperatives employing veterinarians have been using for many years, but which is now tending to become generalized among industrial vets due to the development of franchised practices. company x is a part of group y [a pig co-operative]. [we sell] zero drugs. there is hygiene, so disinfectants and detergents, and also nutrition and probiotics, i.e. products that can be added to food or that farmers can put in drinking water. you have suppliers for these products or you manufacture them yourselves? there are products that we make ourselves, otherwise we choose products from everything we know, and from people we work with. we are doing clinical and zootechnical trials as well. we've written some articles, i did one on swine haemorrhagic dysentery. we did an international article in which we showed that we can prevent it with a certain probiotic. it allowed me to go and see farms in portugal, spain and england as well. i even went to cuba for x. trying to sell products? to preach the good word! we hold meetings on a given theme, for example it might be digestive issues, and we talk about the ecopathological approach. rather than using this or that antibiotic, there are other things that work, and we know this because we have experience with farmers from y. thanks to this experience, and because there were disastrous situations where antibiotics no longer worked, we can tell other veterinary colleagues [that these products work well]. a third strategy consists in developing various forms of contractualization with the clients. for example, the vet interviewed below tries to imagine an annual flat-rate model that would cover all veterinary services, from advice to medicines. this would be based on a form of insurance, where some farmers would ultimately pay a higher price than the actual cost of the services they have received, while others, particularly those facing more health issues, would pay a lower price (although the reverse might be true from one year to the next). in fact, this system is similar to the one set up by the so-called "veterinary groups under contract" in central and south-eastern france, but which was only developed in areas where livestock farming is not widespread i.e. in small and medium-sized cattle or sheep farms connected to local markets; this system is mainly associated with alternative agricultural projects resulting from 1970s protest movements (combettes et al. 2012) . it would be interesting to know the extent to which such a model might be easily generalizable in the heart of an industrialized agri-food system where farmers are involved in more complex and constraining chains of interdependence with upstream and downstream industries. this trend resonates with certain dynamics currently observable in the uk where farm animal vets are also concerned about the sustainability of their economic model and are trying to develop similar forms of contractual veterinary services, although less than 10% of the clientele of the practices trying to develop these kinds of contractual schemes have adopted it so far (ruston et al. 2016 ). the economic model and the type of goods and services that veterinarians can provide are nevertheless now considered to be key criteria for the sustainability of veterinary businesses (henry et al. 2016) . so i've been thinking for a year. now i'm going to test a new system, a comprehensive flat-rate package. i'm testing it, it's brand new, with two or three farmers. this is a complete veterinary follow-up, which includes visits and medicines. we don't talk about fees anymore. [...] so i don't know yet. my package system might be good in theory, maybe not good if farmers... because you know, on the farms, there are major variations in antimicrobial consumption. at some point the package is going to be an average. some of the farmers will find my services a little bit too expensive whereas others will like it, but yes perhaps major antibiotic users so it might be a bad influence for them. i'm not claiming victory yet, but i'm trying to find another system. a fourth strategy consists in monetizing certain veterinary services no longer to farmers, but to co-operatives. this relates to at least two types of activity. first of all, it concerns the follow-up of "herd health plans" that co-operatives are obliged to set up if they want to be approved for selling veterinary medicines, in accordance with the 1975 act. it is of course an activity that has existed for a long time, but that many cooperatives delegated to their employed vets until 2007. the second activity relating to this strategy is linked with the development of new roles for veterinarians, in particular those of standards controllers or certifiers. indeed, with the development of "antibioticfree" labels, more and more co-operatives are asking vets to help them implement such standards. for example, vets have to perform "pharmaceutical audits" in an attempt to recruit farmers who can comply with such specifications, or to set up protocols to monitor antibiotic use and help farmers reduce it. when the co-op announced its project [antibiotic-free pigs], some farmers called me the day after, or i called them, to be part of this project. so for those who were still using antibiotics in feed there were still some stages to go through but when a farmer calls you, you think that you're going to need him for the project so you have to find alternatives. you have to know if he really needs antibiotics and how to reduce them. so this is where you have to set up a procedure, to find ways to identify the flaws… and in the end this is the perfect way to familiarize farmers with this issue. (…) well the farmer has to be motivated but this is where the bill of specifications helps, because there is an added-value. and there is also glory and self-satisfaction because the added-value isn't much, but there is both. yet these new roles for veterinarians, or more exactly these new ways of monetizing their expertise, are associated with a redefinition of their professional identity that vets are not always comfortable with. while being a health advisor rather than a clinician is perfectly in line with their preventive conception of veterinary medicine, becoming a sort of controller and sometimes even a sales representative in charge of enrolling and accompanying farmers in quality insurance schemes is not necessarily easy. this difficulty has also been noted among uk cattle vets who felt conflicted between their role of practitioner (within the framework of their relationship with their clients) and their role of "enforcers" of biosecurity practices (while implementing policy measures to prevent bovine tuberculosis), although in this case they were acting on behalf of the government and not for market schemes of private stakeholders (enticott 2012) . the vet interviewed below explains the practical and ethical adjustments he has had to make to accommodate this new way of valuing his professional expertise, and the limits he personally sets to remain in what he considers to be the role of a health advisor. so luckily we have some experience [with following up on bills of specifications], especially with welfare standards. in this standard, the farmers have to use the co-op's animal feed. so our experience allows us to recognize the feed bags. and one day i was on a farm visit, and i saw a feed bag. and i thought "but this is not the feed from the co-op?". because i know that it is normally smal