Incomplete vaccination and associated factors among children aged 12–23 months in South Africa: an analysis of the South African demographic and health survey 2016

Background: Socioeconomic and health inequalities remain a huge problem in post-apartheid South Africa. Despite substantial efforts at ensuring universal access to vaccines, many children remain under-vaccinated in the country. This study aimed to assess the prevalence and factors associated with incomplete vaccination in the first year of life, among children aged 12–23 months in South Africa. Methods: The study is a secondary analysis of the 2016 South African Demographic and Health Survey. A multivariable logistic regression model was applied to the data on 708 children aged 12–23 months. The study outcome, vaccination completeness, was assessed using a composite assessment of nine doses of four vaccines; Bacillus Calmette–Guérin (BCG) (one dose), Polio (four doses), diphtheria-tetanus-pertussis containing vaccines (DTP) (three doses) and measles-containing vaccines (MCV) (one dose). Children who received all the nine doses were categorized as completely vaccinated. Independent variables included child, maternal, and demographic characteristics. Variables were included in the model based on literature findings. Bivariate analyses were used to examine the crude association between each independent variable and incomplete vaccination, while the multivariable logistic regression model was used to examine the adjusted association after controlling for other variables. Measures of association were presented as odds ratios (OR) with their 95% confidence intervals (CI). Results: About two-fifths (40.8%) of the children were incompletely vaccinated. The prevalence of incomplete vaccination was significantly high among children whose mothers did not receive antenatal care (ANC) during pregnancy (57.5%), and children living in Gauteng Province (52.2%). From the bivariate analyses, the odds of being incompletely vaccinated were three times higher in children whose mothers did not attend ANC compared with children whose mothers attended ANC (crude OR = 2.93; 95% CI 1.42–6.03). The odds were about three times higher in children living in Mpumalanga province (OR = 2.58; 95% CI 1.27–5.25) and in those living in Gauteng province (OR = 2.76; 95% CI 1.30–5.91), compared with those living in Free State province. Conversely, the odds were 32% lower in children from rich households, compared with those from poor households (OR = 0.68; 95% CI 0.47–0.98). In the adjusted model, the higher odds of incomplete vaccination in children whose mothers did not attend ANC were maintained in both magnitude and direction (adjusted OR [aOR] = 2.87; 95% CI 1.31–6.25). Similarly, compared with children living in Free State province, the higher odds of a child being incompletely vaccinated in Mpumalanga (aOR = 2.30; 95% CI 1.03–5.14) and in Gauteng (aOR = 3.10; 95% CI 1.35–7.15) provinces were maintained in both magnitude and direction. Conclusions: There is a substantial burden of incomplete childhood vaccination in South Africa. Maternal ANC attendance during pregnancy and area of residence significantly influences this burden. Interventions that promote broader health service utilization, such as ANC attendance, can help improve the awareness and uptake of routine childhood vaccination. It is also imperative to take into consideration the provincial disparities in childhood vaccination completeness, in planning and implementing interventions to improve vaccination coverage in the country.

This is a cross-sectional analysis of nationally representative data from the South African Demographic and Health Survey (SADHS) 2016. The main purpose of the DHS was to provide up-to-date estimates of basic demographic and health indicators which include fertility levels, maternal and childhood mortality, immunization coverage, HIV testing and counseling, and physical and sexual violence against women. Another objective was to provide estimates of health and behavior indicators in adults aged 15 y and older. Administratively, South Africa is divided into nine provinces. The provinces are in turn divided into 52 districts: 8 metropolitan and 44 district municipalities. The district municipalities are further subdivided into 205 local municipalities.17 The SADHS is a composition of household surveys conducted in South Africa across provinces and districts using a multi-stage, stratified sampling design with households as the sampling unit.17 The sampling frame used for the SADHS 2016 is the Statistics South Africa Master Sample Frame (MSF), which was created using Census 2011 enumeration areas (EAs). In the MSF, EAs of manageable size were treated as primary sampling units (PSUs). Small neighboring EAs were pooled together to form new PSUs, while large EAs were split into conceptual PSUs.17 The MSF contained information about location (province, district, municipality), type of residence (urban or non-urban), and number of households for each PSU. The SADHS 2016 followed a stratified two-stage sampling design with a probability proportional to size sampling of PSUs at the first stage, and systematic sampling of residential dwelling units (DUs) at the second stage. One or more households may be located in any given DU; recent surveys have found 1.03 households per DU on average. The Census 2011 DU count was used as the PSU measure of size. The sampling was designed to provide estimates of key indicators for the country as a whole, for urban and non-urban areas separately, and for each of the nine provinces in South Africa. To ensure that the survey precision was comparable across provinces, PSUs were allocated by a probability proportional to size. Each province was stratified into urban, farm, and traditional areas, yielding 26 sampling strata, from which 750 PSUs were selected. DUs within each PSU were listed and this list served as a frame for sampling DUs. Data collection for the SADHS 2016 took place from 27 June 2016 to 4 November 2016. Data were collected using questionnaires administered by conducting face-to-face interviews. Information obtained from the women’s or caregiver’s questionnaires included vaccination status and maternal, child, and demographic characteristics. Information on vaccination was collected through the Road to Health booklet (RtHB) and mothers’ verbal reports. The RtHB contains children’s health records and is issued to all caregivers. It also contains checklists of appropriate child growth monitoring and developmental assessments that should be performed during each visit to the hospital by a child, including vaccination status assessment.18 Interviewers asked mothers to present the vaccination cards to obtain vaccination dates. In the absence of vaccination cards, such mothers were asked to recall the vaccination received by their children. Details of the questionnaires and data collection procedure have been published elsewhere.19 A dataset was created from information obtained from these questionnaires. From the dataset, we included all children aged 12–23 months, excluding those younger than 12 months and those older than 23 months. A total of 708 children aged 12–23 months, living in 416 communities, within 9 provinces were included in the study. To address our objective, we extracted data on vaccination status; child characteristics: sex, birth order, and birth weight; maternal characteristics: age, education, employment status, marital status, media exposure, antenatal care attendance, and household wealth index; and demographic characteristics: district and province of residence. We limited our study to children aged 12–23 months as children at this age are expected to have received all the basic doses of vaccines. This age group represents the birth cohort of the previous year and should have received all scheduled vaccines by 1 y of age. Vaccination completeness was assessed using a composite outcome of nine doses of four vaccines for which SADHS data were collected. The vaccines include; bacillus Calmette–Guérin (BCG) (one dose), polio vaccine (four doses; including two doses of oral polio vaccine (OPV) and two doses of inactivated polio vaccine (IPV)), diphtheria-tetanus-pertussis-containing vaccines (DTP) (three doses) and measles-containing vaccines (MCV) (one dose). The DTP-containing vaccine currently used in South Africa is a hexavalent vaccine that also includes IPV, Haemophilus influenzae type b and Hepatitis B virus antigens (See Table 1).14 According to the WHO Guidelines, complete vaccination is defined as receipt of one dose of BCG, three doses of DTP-containing vaccine, at least three doses of vaccine against polio and one dose of measles vaccine within the first year of life.20 Although there are now newer vaccines like the pneumococcal conjugate vaccine (PCV) and rotavirus vaccine (RV) in the country’s child vaccination schedule (See Table 1), this study’s definition of complete vaccination was limited to the so-called ‘traditional vaccines’ with which complete vaccination is defined by WHO.20 Children who received all the nine vaccine doses were categorized as completely vaccinated and those who received less than nine doses were defined as incompletely vaccinated. The following determinant variables were considered in the study; child characteristics: sex, birth order, and birth size; maternal characteristics: age, education, household wealth index, marital status, occupation, media exposure, and antenatal care attendance; demographic characteristics: area and province of residence. Sex of child was defined as male and female. Birth order of children was grouped into 1st – 3rd order and 4th+ order. The weight of the child at birth was grouped into three categories of size; large, average, and small. The mother’s age was grouped into 15–24, 25–34, and 35+ y. Educational levels were defined as less than secondary and secondary or higher. Wealth index was originally presented in five quintiles by the SADHS 2016 which were derived from the measurements of ownership of household items such as car, radio, television, and dwelling features like toilet facilities, water source, and type of roofing/floor. This mode of measurement has been used by the World Bank to categorize households into poverty levels based on principal components analysis.21 For easy interpretation, we reclassified the wealth index into three categories (poor, middle, and rich). Marital status was grouped into unmarried and married. Maternal occupation was classified as not working and working. Media exposure refers to the frequency of exposure to a newspaper, radio, and television. Those who were exposed to any of the three outlets (for any number of times in a week) were defined as having media exposure and others were considered as not having media exposure. Antenatal care attendance was defined as making at least one antenatal clinic visit during the pregnancy of the index child and categorized as attended and never attended. Descriptive statistics were used to summarize the data, by presenting the distributions of independent variables by the outcome variable. The distributions were expressed as frequencies and percentages. Bivariate analyses and a multivariable logistic regression model were used to examine the association between various independent variables and incomplete childhood vaccination. Bivariate analyses were used to examine the crude association between each independent variable and incomplete vaccination, while the multivariable logistic regression was used to examine the adjusted association between each independent variable and incomplete childhood vaccination. We have applied binary logistic regression because of the dichotomous nature of the outcome variable. Variables used to build the model were based on previous literature. Measures of association were presented as odds ratios (OR) with their corresponding 95% confidence intervals (CI), with statistical significance considered at p-value <0.05. We used the Bayesian Information Criterion (BIC) to assess the goodness of fit of the model. Variance Inflation Factor (VIF) was applied to test for multicollinearity. To ensure the representativeness of the sample, sampling weight was assigned at various levels of analysis to account for over and under-sampling within sampling units. This was done by taking into account the probabilities of selection at the various strata in the weighted sample. Given the stratified nature of the data, a multilevel analysis would have been an ideal analytical approach. However, due to a small sample size and insufficient effective sample size of children per PSU, we resorted to a multilevel logistic regression approach. To account for the clustering effect, we used a mixed effect logistic regression approach in fitting the model. Data analysis was done using Stata statistical software (Version 14).22

N/A

• Dima Health