Background: The expanded Programme on Immunization (EPI) is one of the most cost-effective interventions to reduce childhood mortality and morbidity. However, determinants of childhood immunization have not been well studied in Senegal. Thus, the aim of our study is to assess routine immunization uptake and factors associated with full immunization status among Senegalese children aged 12-23 months. Methods: We used the 2010-2011 Senegalese Demographic and Health Survey data. The DHS was a two stages cross-sectional survey carried out in 2010-2011. The analysis included 2199 children aged 12-23 months. The interviewers collected information on vaccine uptake based on information from vaccination cards or maternal recall Univariate and multivariable logistic regressions models were used to identify the determinants of full childhood immunization. Results: The prevalence of complete immunization coverage among boys and girls based on both vaccination card information and mothers’ recall was 62.8%. The immunization coverage as documented on vaccination cards was 37.5%. Specific coverage for the single dose of BCG at birth, the third dose of polio vaccine, the third dose of pentavalent vaccine and the first dose of measles vaccine were 94.7%, 72.7%, 82.6%, and 82.1%, respectively. We found that mothers who could show a vaccination card [AOR 7.27 95% CI (5.50-9.60)], attended at least secondary education level [AOR 1.8 95% CI (1.20-2.48)], attended four antenatal visits [AOR 3.10 95% CI (1.69-5.63)], or delivered at a health facility [AOR 1.27 95% CI (1-1.74)] were the predictors of full childhood immunization. Additionally, children living in the eastern administrative regions of the country were less likely to be fully vaccinated [AOR 0.62 95% CI (0.39-0.97)]. Conclusions: We found that the full immunization coverage among children aged between 12 and 23 months was below the national (> 80%) and international targets (90%). Geographic area, mother’s characteristics, antenatal care and access to health care services were associated with full immunization. These findings highlight the need for innovative strategies based on a holistic approach to overcome the barriers to childhood immunization in Senegal.
We conducted a secondary analysis of the Senegal 2010–2011 DHS data. The objectives,organization, and sample design of the DHS are described elsewhere [9]. Briefly, the 2010–2011 DHS was a nationally representative household survey implemented across all the 14 administrative regions of Senegal between October 2010 and April 2011. During the 2010–2011 DHS, a total of 392 clusters were selected from urban and rural strata with a sampling probability proportional to the population size, with 147 clusters selected for urban areas and 245 clusters for rural areas. For the urban strata, the number of clusters per region varied from a maximum of 30 in Dakar to a minimum of 6 in the Kaffrine region. For the rural stratum, the number of clusters ranged from 4 in the Dakar region to 23 in the Diourbel region. At the second stage of the sampling, interviewers randomly sampled households within each cluster. A total of 8212 households were selected across the 14 regions in Senegal. All women aged between 15 and 49 years old who were either a permanent resident or a visitor present of the household on the night before the survey were eligible for the surveys. However, the overall sample size was only representative of the national population and couldn’t be disaggregated by region. Our analysis included women who had a live birth within the 2 years preceding the survey and with a living child aged 12–23 months. Eligible women were interviewed using a Women’s Questionnaire, including maternal and child parameters [10]. The 2011 DHS collected information on vaccination coverage from two sources: the vaccination card shown by mothers to interviewers and the mother’s recall of vaccination. If the health card was available, information regarding the date of administration was directly collected from the vaccination card which normally records dates of all routine vaccinations. If no card was presented, the interviewer asked the mother to recall all vaccination received by their child and when appropriate, the number of doses received without asking for the dates. During the 2011 DHS, out of a total of 2199 women surveyed, 31.33% (689/2199) did not show a vaccination card and therefore reported on children vaccination by recall only [10]. Evidence on the quality of data from maternal recall is documented from previous studies in sub-Saharan Africa and LMICs which indicate that maternal recall is almost similar compared to data collected from the health card [11, 12]. We used the Children’s record dataset for those aged between 12 and 23 months. According to WHO guidelines, a fully immunized child is a child in the 12–23 months old age group who has received a single dose of BCG vaccine, three doses of Pentavalent vaccine (which contains five antigens against diphtheria, tetanus, pertussis, hepatitis B and haemophilus type b), three doses of polio vaccine (excluding the dose given shortly after birth) and the first dose of measles vaccine. In this study, our definition of full immunization did not include vaccines introduced after 2012 such as rotavirus vaccine, pneumococcal vaccine and the second dose of measles-contained vaccine. Previous studies on child immunization coverage have used the same definition for full childhood immunization [13–15]. Thus, the percentage of children aged between 12 and 23 months who received all the specified vaccines according to the information on vaccination card or by mother’s recall represents the study outcome. In the children’s dataset, the outcome “complete immunization” had five categories of response: No (o), vaccinations dates on card (1), vaccinations reported by mothers (2), vaccinations marked on card (3) and don’t know (4). We recoded the three categories “vaccination date on card” (1), “vaccination reported by mothers” (2) and “vaccination marked on card” (3) as “1” and recoded all the remaining categories as “0” and labelled “not received”. For the nine vaccine doses, we first recoded to reflect “vaccinated” or “not vaccinated” for each dose and combined them to reflect “completely vaccinated”. We selected 17 co-factor variables potentially associated with child immunization. The WHO framework on epidemiology of the unimmunized child [6] describes the different factors affecting child’s immunization into four main categories: health care immunization system, communication and information, family characteristics, and parental attitudes and knowledge. In our study, the immunization system category included the distance to health facility, and the need to take transportation. The communication and information category included: use of mass media according to the levels of access and source (radio, TV and newspapers), family characteristics included the followings variables: mother’s and father’s education level, mother’s age at child birth, marital status, household level of poverty assessed by the wealth quintile, ethnic group, religion, child gender, birth order, urban/rural residence (urban/rural), and region of residence. Variables on familiarity and use of other health care services such as antenatal care during pregnancy and the relative distance to the closest health center represented the parental attitudes/knowledge. Finally, we included the gender relationship such as the involvement of women in household decision making. We summarized continuous variables using means with standard deviations and summarized categorical variables with frequencies and percentages. We conducted bivariate analysis and binomial logistic regression. Variables significant at p-value ≤0.25 were included in the multivariable logistic regression models. Vvariables that did not have a significant regression coefficient were removed and a smaller model set up. To assess for confounding, we compared for each variable the estimated coefficient in the smaller model with the previous values in the larger model. Variables, when excluded, that changed the coefficient of remaining variables of Δβ > 20%, were considered as potential confounders and were added back in the model [16]. Variables that were not significant at the univariate analysis were added back to the model and their significance assessed in the presence of other significant variables. Finally, we added demographic characteristics such as age, sex, and predictors well known from previous research but not significant in our model. Subsequently, the goodness of fit of our final model was tested using the Hosmer-Lemeshow test [17]. All data management procedures and statistical analysis were done using STATA software version 13. Due to the complex sampling design, we used the Svyset command to account for inverse probability weighting (IPW), clustering, and stratification to provide unbiased estimates of the population parameters.
N/A