Hrev_master [Healthcare in Low-resource Settings 2016; 4:5447] [page 17] Family factors associated with immunization uptake in chil- dren aged between twelve and fifty-nine months: a household survey in Kakamega Central district, Western Kenya Joram L. Sunguti,1 Penny E. Neave,1 Steve Taylor2 1Department of Public Health; 2Department of Biostatistics and Epidemiology, School of Public Health and Psychosocial Studies, Auckland University of Technology, New Zealand Abstract In this study, we assessed immunization uptake and identified family factors associated with immunization in children aged between 12 and 59 months in Kakamega Central, Western Kenya. A cross sectional study was conducted in 13 sub-locations between June and July 2013. Data on 577 children were col- lected from their respective caregivers, by trained research assistants. The proportion of fully immunized children was 80.9% (95% con- fidence interval 76.9-85.3%). Immunization coverage was higher among caregivers who had completed secondary school (88%), those who had attended antenatal care clinics (81%) and children born in a health facility (85%). Some evidence was seen of increasing cover- age with increasing socio-economic status. No evidence for a gender difference in coverage was seen. In the logistic regression model, the risk factors for incomplete immunization were: low educational level of the caregiver [adjusted odd ratio (AOR)=0.25; P<0.005], never attending any antenatal care (ANC) (AOR=0.14; P<0.05) and delivery outside of health facilities (AOR=0.40; P<0.005). Further inquiry is required into this area to fully com- prehend the inextricable linkage between fac- tors affecting immunization. Introduction In 1974, the World Health Organization (WHO) launched the Expanded Program on Immunization (EPI) initiative. Its aim was to ensure that children aged below 5 years in all countries benefited from vaccination against diphtheria, pertussis, tetanus (DPT), poliomyelitis, measles and tuberculosis. In some countries, more vaccines have now been added to the schedule including hepatitis B, haemophilus influenza type B and yellow fever.1 Despite this, in 2013, an estimated 14% of the infants (mostly from low income coun- tries) failed to access three of these vaccines (DPT) during their first year of life.2 In Kenya, the Ministry of Health is charged with the delivery of efficient immunization services, through the Division of Immunization (DVI) department. Within one year of birth, each child should receive one dose of Bacillus Calmette-Guérin (BCG) as protection against tuberculosis, three doses of vaccination against DPT, four doses of oral polio vaccine (OPV), three doses of hepatitis B vaccine (HBV), three doses of haemophilus influenza type B vaccine (HIB), three doses of pneumococcal conjugate vaccine and one dose of measles vaccine. The DPT, Hepatitis B and Haemophilus influenza type B vaccines are administered as a pentavalent vaccine.3 Despite the aim to vaccinate all children, vaccine-preventable disease outbreaks have been recorded,4 indicating that this is not being achieved. The most recent official esti- mates support this, with the national coverage being approximated as: BCG 79%; DPT-3 76%; OPV 82%; HBV 83%; HIB 83%; and measles 79%.5 Variations in immunization uptake have been documented in different areas of Kenya, with the highest rates being in Nandi County (93.9%) and the lowest in Mandera (27.7%).6 However, it is acknowledged that there are lim- itations to the accuracy of all official estimates with possible variations between 8% and 16%.5 A number of factors have been associated with immunization uptake. These include maternal education or literacy,7-9 maternal age at birth,10,11 paternal education level12 and ante- natal care utilization during pregnancy.13 Children born in a health facility have been found to be more likely to be immunized than those born at home,14,15 but there is no strong evidence that a child’s sex is associated with vaccination uptake.16,17 Household characteris- tics that have been documented to correlate with immunization include socioeconomic sta- tus,18 proximity to a health facility19 and whether the household is located in a rural or urban area.20 Despite studies showing association between socio-demographic factors and immu- nization uptake, this relationship is not con- clusive. A study in Ethiopia failed to show any significant association between immunization and socioeconomic status, maternal age, total number of children, age of the father, educa- tion level of the father and sex of the child.13 In Kenya, the following factors were not associat- ed with immunization; maternal age, socio- economic status, partner’s education level, sex of the child and place of delivery.9,19,21,22 This highlights the need for further studies to understand these associations. Kakamega County is in a predominately rural area of Kenya. The district consists of 13 administrative units, called sub-locations. The average population in each sub-location is 13,000. The main language is Luhya followed by Swahili and English. The majority of the population is subsistence farmers with a small number of business people working in an urban center. Like the other 47 counties, Kakamega County has a devolved governance system. Each county draws revenue from the central government allocation and levies taxes at the county level. Immunization uptake in Kakamega County is estimated to be 62.2%.6 However, as is the case with national vaccination uptake esti- mates, these may not be accurate. Clearly, there is a need to gather accurate estimates of complete immunization and the factors associ- ated with this. Understanding factors associat- ed with immunization are important in informing stakeholders to implement key Healthcare in Low-resource Settings 2016; volume 4:5447 Correspondence: Joram Luke Sunguti, Department of Public Health, School of Public Health and Psychosocial Studies, Auckland University of Technology, 55 Wellesley Street East, Auckland Central, New Zealand. Tel: +64.9.9219779. Email: jsunguti@yahoo.com Acknowledgements: we acknowledge the partici- pants from Kakamega who took time to take part in this study. We also acknowledge the District Medical Officer of Health, the Public Health Officer and all the research assistants from Kakamega Central for their invaluable support during the research period. We are grateful to the New Zealand AID Foundation for their funding support. Key words: Immunization; Vaccination; Factors; Kenya. Contributions: JLS, study conception, data acqui- sition and drafting of manuscript; ST, design, data analysis and interpretation; PEN, critical revision and final approval. Conflict of interest: the authors declare no poten- tial conflict of interest. Funding: this work was supported by New Zealand AID Foundation. Received for publication: 16 July 2015. Revision received: 7 February 2016. Accepted for publication: 8 February 2016. This work is licensed under a Creative Commons Attribution 4.0 License (by-nc 4.0). ©Copyright J.L. Sunguti et al., 2016 Licensee PAGEPress, Italy Healthcare in Low-resource Settings 2016; 4:5447 doi:10.4081/hls.2016.5447 Non co mmerc ial us e o nly [page 18] [Healthcare in Low-resource Settings 2016; 4:5447] interventions aimed at improving immuniza- tion uptake. The purpose of this study therefore was to measure completeness of immunization uptake and the factors associated with this in children aged between 12 and 59 months living in Kakamega County, Western Kenya. Materials and Methods A cross-sectional study design was used, specifically a stratified survey of households. It was determined that a sample size of 520 chil- dren (40 per sub-location or stratum) was required, based on the recommended single proportion formula for immunization, with a 95% confidence level, 5% margin of error and assuming 80% immunization coverage rate.23 A 5% non-response rate and a design effect of two were considered. Inclusion criteria were: a caregiver to a child aged between 12 and 59 months who had lived in the caregiver’s home in Kakamega Central District for at least six months. In households with two or more children qualify- ing for inclusion, the youngest was selected. In houses where twins lived, the tossing of a coin determined for which child the information should be collected. All those not meeting the inclusion criteria were excluded from partici- pation. Three weeks before data collection began, six research assistants from the district public health office were trained by the principal investigator on the rationale for the study, eth- ical issues, inclusion and exclusion criteria, the study method and how to record and return the information gathered. Five research assis- tants were public health officers while the other was a health records and information officer. All had attained tertiary level training (diploma and/or degree). Two weeks before the study, chiefs, village elders and research assistants made announcements about the study in local schools, churches and market places and encouraged participation. The selection of participants was done through stratified sampling followed by simple random sampling of households within strata. Kakamega Central district consists of 13 administrative units, called sub-locations. Each sub-location constituted a stratum from which households were drawn for the survey. The first household to be visited within each sub-location was selected randomly from a sampling frame listing all households available from the Ministry of Provincial Administration. The person who answered the door was informed about the study and asked if a child in the household met the inclusion criteria. The next house to be visited was the nearest household, which met the inclusion criteria. For those who met the inclusion criteria, one caregiver was interviewed with a short struc- tured questionnaire. Information was collected on: caregiver’s relationship to the child; moth- er’s age at delivery of the child; age and level of education of the principal caregiver and part- ner; the number of antenatal visits made; place of delivery; the birth order and sex of the child and the number of immunizations for the child. Socio-economic status was measured using a principal component analysis used in other household surveys in Kenya.23,24 In addi- tion to responding to the questionnaire, all study participants were asked to produce the child’s vaccination card, national identifica- tion cards of the caregiver, birth certificates and academic certificates. These were used to corroborate the information given by the care- Article Figure 1. Baseline characteristics of the survey. Table 1. Summary of the survey data per stratum. Strata Population distribution Sample distribution Survey results Weights N* % N* % p̂� i Var (p̂� i) Emukaya 1235 3.3 33 5.7 0.818 0.00465 0.57 Lurambi 1130 2.9 45 7.8 0.844 0.00299 0.38 Eshisiru 989 2.6 48 8.3 0.896 0.00198 0.32 Indangalasia 1566 4.2 44 7.6 0.432 0.00571 0.54 Shibuli 2417 6.4 38 6.6 0.868 0.00310 0.97 Shirakalu 1173 3.1 46 8.0 0.935 0.00135 0.39 Shiyunzu 1919 5.1 45 7.8 0.956 0.00096 0.65 Sichilayi 10,475 27.8 48 8.3 0.771 0.00376 3.34 Shirere 7738 20.5 46 8.0 0.870 0.00251 2.58 Township 2691 7.2 39 6.8 0.846 0.00343 1.06 Matioli 1387 3.7 49 8.5 0.673 0.00458 0.43 Murumba 2104 5.6 48 8.3 0.646 0.00487 0.67 Mahiakalo 2865 7.6 48 8.3 0.896 0.00198 0.91 Total 37,689 100 577 100 *N refers to number of households from which respondents were picked. Non co mmerc ial us e o nly [Healthcare in Low-resource Settings 2016; 4:5447] [page 19] giver. Respondents’ information was coded by the primary researcher into numerical responses and double-entered in excel before being exported to SPSS (IBM Corp. Released 2011. IBM SPSS Statistics for Windows, Version 20.0. Armonk, NY, USA). Statistical analysis was conducted using the software SPSS v20 for Windows, with an alpha value of 0.05 used to indicate significance. Data were initially checked for consistency and outliers through use of tables, histograms and box plots. Mean, median and standard deviations were used to describe continuous data, while frequencies were used for categorical data. A multiple logistic regression model was used to estimate associations and check for potential confounders among variables. To ensure accu- rate estimation of immunization in Kakamega Central district, each sample proportion (with the respective 95% confidence intervals) was weighted (Table 1). The sample weights, wi were derived from the formula: Ethical approval for the study was obtained from Auckland University of Technology Ethics Committee. Permission to proceed with the study was also obtained from the Kenyan Ministry of Health. Results Baseline characteristics After visiting 649 households (oversampling was done due to availability of more household for interviews), caregivers from 577 house- holds were interviewed, translating to a response rate of 90.1% (Figure 1). The mean age of the caregivers was 27.6 years, whilst that of the children was 24.8 months, with slightly more than half being boys (53.2%). Table 2 summarizes the sample characteris- tics. Immunization coverage Among the households visited, the propor- tion of completely immunized children was 80.2%. Adjusting for the stratified design, the estimated coverage for the district was 80.9%. Every child had received at least one form of vaccine against the diseases in the Kenyan immunization schedule. The vaccination cov- erage rates for BCG, the third polio dose (OPV- 3), pentavalent 3 and measles were 99.5, 85.1, 94.5 and 90.8% respectively. Bivariate analyses The coverage was higher (88%) among caregivers who had completed secondary school than among those who had not (74%), P<0.001. A similar result was seen for the edu- cation level of the partners. Although there were relatively few caregivers who did not attend any antenatal care (ANC) visits, there was evidence of a significantly lower coverage (54%) for them, compared to those who had attended ANC (81%), P<0.001. No evidence for a gender difference in coverage was seen, P=0.74. Children born in a health facility had greater coverage (85%) than those who were not (71%), P<0.001. Some evidence was seen of increasing coverage with increasing socio- economic status. Coverage decreased for chil- dren born into larger families, down to 69% for children with a birth order of six or more (Table 3). Logistic regression analysis Complex samples logistic regression was performed to assess the impact of the factors measured on the likelihood that children would be fully immunized. Prior to interpreta- tion of regression coefficients, the model was tested to determine its fitness. The Hosmer- Lemeshow goodness of fit test indicated that the logistic regression model was fit to test the association between socio-demographic vari- ables and immunization uptake. After backward stepwise elimination, the final model contained three explanatory vari- Article Table 2. Sample characteristics (n=577). Variable N % Caregiver’s relationship to the child Mother 547 94.8 Father 13 2.3 Other 17 2.9 Marital status of the caregiver Married 483 83.7 Single 70 12.1 Divorced 12 2.1 Widowed 12 2.1 Caregiver’s age (years) ≤20 71 12.3 21-30 357 61.9 31-40 123 21.3 40+ 26 4.5 Caregiver’s higher school level