Incomplete immunization among children aged 12–23 months in Togo: a multilevel analysis of individual and contextual factors

A descriptive and analytical cross-sectional study was carried out from February 27th to March 5th 2017 in the six health regions (HR) of Togo: Lomé, Maritime, Plateaux, Centrale, Kara, and Savanes. The World Health Organization (WHO) cluster survey methodology was used for the immunization coverage survey [14]. The sampling frame consisted of all the neighborhoods / villages / avenues of each selected prefecture. Thus, 30 clusters per health region were randomly selected and distributed by prefecture and canton. In each cluster, seven households were chosen by lot and interviewed. Overall, for the six health regions in Togo, the sample should include at least 1260 households. Therefore, the household was used as the main sampling unit.

All children aged 12–23 months living in selected households and with one of their parents or guardians were included in the study. A sample of 1128 children aged 12–23 months was calculated using a single proportion population formula with a 95% confidence level, 7% margin of error and 60% estimated immunization coverage rate in the study area [13]. A 10% non-response rate was considered and the minimum number of respondent-child couples required in each health region was estimated at 207.

A pre-tested quantitative questionnaire was administered during a face-to-face interview to the person who usually takes the child to the immunization center. The questionnaire included questions on socio-demographic characteristics of the respondent and the eligible child, possession of a vaccination card for the child, vaccines administered to the child (based on immunization records or respondent’s declarative statements), knowledge, attitudes and practices regarding immunization of children, and accessibility to immunization services. Data collection was carried out by investigators who participated in a two-day training session in order to master data collection tools and techniques for locating survey areas. All data collection forms completed by the interviewers were double checked by study supervisors for completeness and consistency. Validated forms were then sent to the EPI department for data entry.

Data entry and cleaning were carried out by a team of ten data entry operators, a supervisor and a data transcriber previously trained and under the supervision of a database manager. A database was designed with the Epidata software version 3.1. Descriptive analyses were performed and results were presented as proportions (% of children with complete immunization) with their 95% confidence interval. Comparisons of the qualitative variables were performed using chi-square or Fisher’s exact tests and comparisons of medians were performed using nonparametric Kruskall Wallis or Wilcoxon tests. Univariate and multivariate multilevel logistic regression analyses were performed to study the relationship between the dependent variable ‘incomplete immunization coverage’ and the individual and contextual variables. The dependent variable ‘incomplete immunization coverage’ was coded as 0 for fully immunized children and 1 for unimmunized or partially immunized children. The explanatory variables were presented into two categories for regression analyses: individual factors (socio-demographic characteristics of the parents/guardians, knowledge of mothers/guardians that vaccine is free of charge, possession of immunization cards) and contextual factors (walking time to immunization centers and health regions). All analyses were carried out using STATA® software version 11.0 (StataCorp, College Station, Texas, USA).

This study was approved by the National Ethics Committee of the Ministry of Health in Togo and all respondents signed an informed consent form before enrolment in the study.

Respondents in households were predominantly female (91.91%), married (93.42%) and only 22.76% had secondary or higher education. In most households (94.69%) there was only one child aged 12–23 months. In the Savanes-HR, 18.1% of parents or guardians reported having more than one child aged 12–23 months in their household compared with 1.90% in Lomé-HR and Centrale-HR, 2.39% in Kara-HR, and 2.80% in Plateaux-HR and Maritime-HR. Respondents’ sociodemographic characteristics and household characteristics are presented in Table 1.

Nine in ten children (90.72%) were immunized against tuberculosis. The proportion of children who received the first dose of pentavalent vaccine was 83.90%, but this proportion decreased gradually from the first dose to the third one which was 81.21%. The same trend was observed for the oral polio vaccine (OPV) with a proportion of children vaccinated dropping from 88.26% for the first dose to 86.52% for the second dose and then to 85.25% for the third dose (Table 2). Lomé-HR was the health region with the best immunization coverage. All children were immunized against tuberculosis and 99.05% received the first dose of vaccine against polio and pentavalent vaccine. Immunization coverage varied between 96.83% for yellow fever vaccine and 99.63% for BCG in children with an immunization card. In the Centrale-HR, all respondents whose children had an immunization card received BCG, three doses of oral poliomyelitis vaccine, three doses of pentavalent vaccine, three doses of pneumococcal vaccine, two doses of vaccine against Rotavirus, and vaccines against measles and yellow fever. For children without an immunization card, immunization coverage varied between 64.58% for yellow fever vaccine and 86.30% for BCG.

In the enrolled households, 912 children were fully immunized, yielding a complete immunization coverage of 72.32% (95% CI 69.76–74.77), irrespective of the availability of immunization cards. Among the 1076 children with immunization cards, 872 (81.04%) were fully immunized. Among those who did not have an immunization card, 77.42 and 41.18% of the children were partially immunized in Maritime and Lomé HR respectively. Despite the absence of an immunization card, it was found that 86.67% of children in the Centrale-HR and 52.94% of children in the Lomé-HR had received all their vaccines.

Main barriers to children’s immunization reported by respondents were: long distance to the health center (32.28%), poor road conditions (13.32%), lack of means of transport (10.55%) and lack of time (9.04%). The regional analysis shows that in the Lomé-HR, the main barrier was the lack of time (23.22%) while in the other regions, the remoteness of the health center was the main reason mentioned.

Table 3 summarizes univariate analysis. Four factors were associated with incomplete immunization: Monthly income (p = 0.035), marital status (p = 0.008), knowing that immunization is free of charge (p = 0.032), availability of immunization card (p < 0.001) and the health region (p < 0.001).

Multilevel multivariate logistic regression analyses results are presented in Table 4. The empty model (Model 1) showed that there was significant variability in the probability of children with incomplete immunization across districts [variance = 1.07, p < 0.001]. The inter-class correlation (ICC), 59.37% of variability in the probability of a child to have incomplete immunization was related to district-level factors. The variation in incomplete immunization in the other models (Model 2, Model 3 and Model 4) remained significant. In the Model 2, 62% of the variability was attributed to district-level factors. In addition, the proportional change in variance (PCV) for individual-level factors was estimated at 8.41% which means that, in the model, 8.41% of the variability in the probability of a child to have incomplete immunization across districts was explained by individual-level factors. With the full model (Model 4), we found that if a parent moves to another district that has a higher probability of incomplete child immunization, the likelihood of having his child not completely immunized would increase by 2.78 (median odds ratios (MOR) for the districts was 2.78). In the model 4, after controlling for both individual and contextual level variables, children whose parents or guardians attended secondary school or above were 33% (adjusted Odds Ratio (aOR) = 0.67, CI [0.47–0.94]) less likely to have an incomplete immunization than their counterparts with no educational background. The likelihood of incomplete immunization in children in this study decreased with increasing household’s income (aOR = 0.73, CI [0.58–0.93]). Conversely, children who did not have an immunization card (aOR = 13.41, CI [9.19–19.57]) and those whose parents did not know that children immunization is free of charge (aOR = 1.82, CI [1.00–3.30]) were more likely to have an incomplete immunization. In addition, children whose parents or guardians walked between half an hour and 1 h to go to the health center were 57% (aOR = 1.57, CI [1.15–2.13]) more likely to have an incomplete immunization compared to those whose guardians have to walk less than half an hour.