Menu
Menu
Menu
Menu
Background: In this study, we aimed to explore the rural-urban disparities in the magnitude and determinants of missed opportunities for vaccination (MOV) in sub-Saharan Africa. Methods: This was a cross-sectional study using nationally representative household surveys conducted between 2007 and 2017 in 35 countries across sub-Saharan Africa. The risk difference in MOV between rural or urban dwellers were calculated. Logistic regression method was used to investigate the urban-rural disparities in multivariable analyses. Then Blinder-Oaxaca method was used to decompose differences in MOV between rural and urban dwellers. Results: The median number of children aged 12 to 23 months was 2113 (Min: 370, Max: 5896). There was wide variation in the the magnitude of MOV among children in rural and urban areas across the 35 countries. The magnitude of MOV in rural areas varied from 18.0% (95% CI 14.7 to 21.4) in the Gambia to 85.2% (81.2 to 88.9) in Gabon. Out of the 35 countries included in this analysis, pro-rural inequality was observed in 16 countries (i.e. MOV is prevalent among children living in rural areas) and pro-urban inequality in five countries (i.e. MOV is prevalent among children living in urban areas). The contributions of the compositional ‘explained’ and structural ‘unexplained’ components varied across the countries. However, household wealth index was the most frequently identified factor. Conclusions: Variation exists in the level of missed opportunities for vaccination between rural and urban areas, with widespread pro-rural inequalities across Africa. Although several factors account for these rural-urban disparities in various countries, household wealth was the most common.
Data for this cross-sectional study was obtained from Demographic and Health Surveys (DHS), which are nationally representative household surveys conducted in low- and middle-income countries. This study used data from 35 recent DHS surveys conducted between 2007 and 2016 in sub-Saharan Africa available as of December 2017. The DHS uses a multi-stage, stratified sampling design with households as the sampling unit.24 Within each sample household, all women and men meeting the eligibility criteria are interviewed. Because the surveys are not self-weighting, weights are calculated to account for unequal selection probabilities as well as for non-responses. With weights applied, survey findings represent the full target populations. The DHS surveys collects data using a household questionnaire. For eligible individuals within households, interviews are conducted using a woman’s or man’s questionnaire. DHS surveys are implemented across countries with standardized interviewer training, supervision, and implementation protocols. We used the World Health Organization (WHO) definition of missed opportunity for vaccination (MOV) as the outcome variable, defined as a binary variable that takes the value of 1 if a child who is eligible for vaccination, 12–23 months had any contact with health services, which does not result in the child receiving one or more of the vaccine doses for which he or she is eligible. Contact with health services was defined using the following five variables: skilled birth attendance, baby postnatal check within 2 months, received vitamin A dose in first 2 months after delivery, has a health card and received medical treatment for diarrhea/fever/cough. Place of residence which was categorised as rural or urban areas. The following factors were included in the models: child’s age, sex (male versus female), high birth order (>4 birth order), number of under five children in the household, maternal age in completed years (15 to 24, 25 to 34, 35 to 49), employment status (working or not working), maternal education (no education, primary or secondary or higher) and media access (radio, television or newspaper). Media access was assessed using the following indicators: access to information measured via frequency of watching television, listening to radio, and reading newspapers/magazine. To allow meaningful analysis, we dichotomized the response levels “less than one week”, “at least once a week”, and “almost every day” as one group and the response level “not at all” as the other group. We then created an additive media access variable (from 0 to 3) that counted the number of media type each respondent had access to. Wealth index was used as a proxy indicator for socioeconomic position. The methods used in calculating DHS wealth index have been described elsewhere.25,26 Briefly, an index of economic status for each household was constructed using principal components analysis based on the following household variables: number of rooms per house, ownership of car, motorcycle, bicycle, fridge, television and telephone as well as any kind of heating device. From these criteria the DHS wealth index quintiles (poorest, poorer, middle, richer and richest) were calculated and used in the subsequent modelling. The analytical approach included descriptive statistics, univariable analysis and Blinder-Oaxaca decomposition techniques using logistic regressions. We used descriptive statistics to show the distribution of respondents by the key variables. Values were expressed as absolute numbers (percentages) and mean (standard deviation) for categorical and continuous variables respectively. In the descriptive statistics the distribution of respondents by key variables were expressed as percentages. All cases in the DHS data were given weights to adjust for differences in probability of selection and to adjust for non-response in order to produce the proper representation. Individual weights were used for descriptive statistics in this study. We calculate risk difference in missed opportunities between the two group, living in rural or urban areas. A risk difference greater than 0 suggests that missed opportunities is prevalent among children living in rural areas (pro-rural inequality). Conversely, a negative risk difference indicates that missed opportunities for vaccination is prevalent among children living in urban areas (pro-urban inequality). Finally, we adopted logistic regression method using the pooled cross-sectional data to investigate the urban-rural disparities in multivariable analyses adjusted for explanatory variables. The Blinder-Oaxaca decomposition was a counterfactual method with an assumption that “what the probability of missed opportunities for vaccination would be if children living in rural areas had the same characteristics as their urban counterparts?”.27,28 The Blinder-Oaxaca method allows for the decomposition of the difference in an outcome variable between 2 groups into 2 components.27,28 The first component is the “explained” portion of that gap that captures differences in the distributions of the measurable characteristics (referred as “compositional” or “endowments”) of these groups.27,28 This illustrates the portion of the gap in missed opportunities for vaccination that is attributed to the differences in observable, measurable characteristics between the two groups. Using this method, we can quantify how much of the gap the “advantaged” and the “disadvantaged” groups is attributable to these differences in specific measurable characteristics. The second component is the “unexplained” part, meaning the portion of the gap due to the differences in the estimated regression coefficients and the unmeasured variables between the two groups.27,28 This is also referred to in the literature as the “structural” component or the “coefficient” portion of the decomposition. This reflects the remainder of the model not explained by the differences in measurable, objective characteristics. The “unexplained” portion arises from differentials in how the predictor variables are associated with the outcomes for the two groups. This portion would persist even if the disadvantaged group were to attain the same average levels of measured predictor variables as the advantaged group. The DHS stratification and the unequal sampling weights as well as household clustering effects were considered in the analysis to correct standard errors. All tests were two tailed and p < 0.05 was considered significant. Regression diagnostics were used to judge the goodness-of-fit of the model. They included the tolerance test for multicollinearity, its reciprocal variance inflation factors (VIF), presence of outliers and estimates of adjusted R square of the regression model. We checked for multi-collinearity among explanatory variables examining the variance inflation factor (VIF),29 all diagonal elements in the variance-covariance (τ) matrix for correlation between −1 and 1, and diagonal elements for any elements close to zero. The largest VIF greater than 10 or the mean VIF greater than 6 represent severe multicollinearity. None of the results of the tests provided reasons for concern. Thus, the models provide robust and valid results.
N/A
