2014.ISDS.Abstracts.Final.pdf


ISDS Annual Conference Proceedings 2014. This is an Open Access article distributed under the terms of the Creative Commons Attribution-
Noncommercial 3.0 Unported License (http://creativecommons.org/licenses/by-nc/3.0/), permitting all non-commercial use, distribution, 
and reproduction in any medium, provided the original work is properly cited.

ISDS 2014 Conference Abstracts

Epidemiological inferences using public information, 
influenza H7N9 epidemic in China

Eric H.Y. Lau*1, Jiandong Zheng2, Tim K. Tsang1, Qiaohong Liao2, Bryan Lewis3, John S. 
Brownstein4, 5, Sharon Sanders6, Sumiko R. Mekaru4, Caitlin Rivers3, Gabriel M. Leung1, 
Luzhao Feng2, Benjamin J. Cowling1 and Hongjie Yu2

1The University of Hong Kong, Hong Kong, Hong Kong; 2Chinese Center for Disease Control and Prevention, Beijing, China; 
3Virginia Tech, Blacksburg, VA, USA; 4Boston Children’s Hospital, Boston, MA, USA; 5Harvard Medical School, Boston, MA, USA; 
6FluTrackers International Charity, Winter Park, FL, USA

Objective
This study described the strength and limitation of using line lists 

that built on publicly available data in various types of epidemiological 
inferences during the H7N9 epidemic in China, 2013.

Introduction
The influenza A(H7N9) virus emerged in early 2013 in China, 

with more than 130 laboratory-confirmed cases identified within a 
short period of about three months. Evidence-based public health 
response is essential for effective control of the disease, which 
relies on epidemiological and clinical data with good quality and 
timeliness. Publicly available information from sources such as 
official health website, online news, blogs or social media has the 
potential of rapid sharing of data to a wide community of experts for 
more comprehensive analyses. In our study we described the strength 
and limitation of these data for various types of epidemiological 
inferences.

Methods
We obtained historical archives of five line lists for influenza 

A(H7N9) cases based on publicly available information, created 
by Boston Children’s Hospital (HealthMap), Virginia Polytechnic 
Institute and State University, FluTrackers, Bloomberg News and 
School of Public Health, University of Hong Kong, and also the official 
line lists from Chinese Center for Disease Control and Prevention, 
covering the period from early April to end of May. Demographic 
and epidemiological variables such as sex, age, province, city, date of 
illness onset, hospital admission, discharge, death and health status 
were available in most of the line lists.

We described the potential types of epidemiological inference that 
could be made for real-time severity and transmissibility assessment of 
the influenza A(H7N9) epidemic. Distributions of basic demographic 
and epidemiological variables were compared. We also compared the 
estimated onset-to-admission, onset-to-discharge and onset-to-death 
distributions and hospital fatality risk (HFR) inferred from different 
line lists, comparing to the official line list. Impact of live poultry 
market closure on H7N9 infection was also estimated in cities which 
implemented the intervention.

Results
Estimated age and sex distributions, epidemic curves, geographical 

spread, onset-to-hospitalization and onset-to-death distributions from 
line lists using public information were similar to those estimated 
from the official line list, at different time points over the course of 
the H7N9 epidemic. Live poultry market closure was found to be 
significantly associated with a lower H7N9 incidence in Shanghai, 
Nanjing and Hangzhou.

We estimated a shorter onset-to-discharge period from the line lists 
based on public information, compared to the official line list. The 
estimated HFRs from these line lists did not converge to the final 

estimate towards the end of the epidemic, when the outcomes of most 
hospitalized patients were available.

Conclusions
The results suggest that public information is able to provide 

accurate information on demographic characteristics and case counts. 
They also tend to report more detailed information for severe and 
death cases. However, information on patient status, hospitalization 
or discharge was less reliable. Publicly available information is likely 
to provide reliable information for inference of transmissibility or 
geographical spread but limited information on disease severity. It 
will be beneficial to public health by developing a protocol to setup 
an open-access minimum dataset based on publicly available data, 
with key epidemiological variables of standardized format and 
definition, along with a time stamp for each variable. This will allow 
a larger group of experts to carry out more timely and comprehensive 
epidemiological analyses when a novel disease emerges.

Keywords
Epidemiological inference; severity; transmissibility; line list; 
influenza H7N9

*Eric H.Y. Lau

E-mail: ehylau@hku.hk    

Online Journal of Public Health Informatics * ISSN 1947-2579 * http://ojphi.org * (1):e143, 201