Study design and population

This study was based on a secondary analysis of the latest Bangladesh Demographic and Health Survey (BDHS) 2014 dataset. The DHS is a nationally representative cross-sectional survey, which uses a two-stage stratified sampling design to cover the target population of the entire country. This survey used the sampling frame provided by the Bangladesh Bureau of Statistics. This survey was carried out for 6 months, from June 2014 to November 2014. The BDHS 2014 used three types of questionnaires: household questionnaire, woman’s questionnaire and community questionnaire. Through the women’s questionnaire, up-to-date information on sociodemographic, maternal and child health indications including individual-level vaccination coverage was collected. Participants of this study were children aged between 1 and 5 years. Childhood immunisation history was collected for all surviving children over the last 5 years. Immunisation data were collected mainly from the records on vaccine cards; if vaccine card was not available, mothers were asked to recall about the vaccination history of the respective child. Face-to-face interviews that were conducted with reproductive-aged women (15–49 years) were conducted for the collection of data using the structured questionnaire.22 Details about a sampling technique, survey design and quality control have been described elsewhere.23 All the DHS data were publicly accessible and were made available on request by the MEASURE DHS. Furthermore, approval was sought from and given by the MEASURE DHS programme office in the use of this dataset for this specific study. According to the DHS, written informed consent was obtained from all participants prior to the interviews.

Outcome variable

The outcome variable of the analysis was the children’s immunisation status and categorised as fully immunised and partially/unimmunised. Vaccination status was categorised as ‘fully immunised’ if the children had received doses of all the eight recommended vaccines: one dose of the vaccine against tuberculosis (BCG), three doses of pentavalent (DPT, Haemophilus influenzae type B and HepB), three doses of oral polio vaccine (OPV) and one dose of measles vaccines.23 Children aged 12–59 months of age were included in this analysis to capture the vaccination status of the children. Children younger than 12 months of age were excluded as they were not old enough to receive the full schedule of the routine vaccines. WHO recommended vaccination schedule for Bangladesh24 is presented in table 1. Immunisation histories for all vaccines were coded as dummy responses, with 1 for fully immunised and 0 otherwise. In a small number of cases, the immunisation cards were not available, and mothers responded with ‘don’t know’ while asked about their children’s vaccination status on certain vaccines. In that cases, the vaccination status of children was considered as ‘not fully immunised’ since such responses reflect a negative response regarding immunisation.14

Explanatory variables

Explanatory variables were selected based on the published literature, prior knowledge and availability of variables in the BDHS 2014 dataset. Area of residence, administrative division, sex and age of children, birth order, family size, parental education, mass media exposure, household wealth index and the mother’s decision-making ability for both the children and their own healthcare matters were included as explanatory variables. In this analysis, the categorisation of continuous variables was done in light of previous literature where the child’s age was categorised into four groups at 12-month intervals, and maternal age was categorised into three groups (less than 20 years, 20–34 years and more than 34 years). Self-reported parental (both mother and father) educational attainment was used and categorised as ‘no education’, ‘primary’, ‘secondary’ and ‘higher’. No education refers to there not having any formal education, with primary was defined as completing grade 5, secondary as completing grade 10 and higher was defined as attaining more than grade 10. Family size was determined by the number of family members who lived together in a household, and defined by the BDHS. Family size was categorised on the basis of other published literature and categorised as ‘small’, ‘medium’ and ‘large’. A ‘small’ family size refers to the family consisting of less than four members, ‘medium’ as four to six members and ‘large’ if the family member exceeds six members. Mass media access was categorised as ‘yes’ if the family had access to televisions and radios and ‘no’ otherwise. We used the predetermined wealth index category provided in the dataset generated from selected household assets using principal component analysis and classified it into five groups, namely: ‘poorest’, ‘poorer’, ‘middle’, ‘rich’ and ‘richest’. Moreover, both the decision-making ability of the mothers for both their children and personal healthcare were categorised into four groups as ‘herself’, ‘jointly with husband’, ‘husband alone’ and ‘other’.

Statistical analysis

The original dataset comprised 8092 children, aged 0–59 months of age. In the case of our analysis, 1555 children were excluded as they were under the age of 12 months. Moreover, 307 cases of missing data were present, and these were cases where no information was provided (no responses, neither ‘yes’, ‘no’ nor ‘don’t know’) on vaccination information, and thus also excluded from the analysis. Finally, a total of 6230 children were eligible and included in the analysis. A proper sampling weight was used in this analysis to make the sample more representative of the population at the national level. Descriptive statistics, such as frequency with percentages, were executed to represent the background characteristics of study participants, and a proportion with 95% CI was used for presenting the coverage rate of fully vaccinated children. The association between each independent variable and the full vaccination uptake was investigated using a univariate analysis. Multivariable logistic regression that adjusted for all the selected independent variables was also constructed to determine the significant influencing factors for fully vaccination and results were presented in terms of adjusted OR(s) (AOR) with a 95% CI. Before the execution of a multivariate regression model, a bivariate analysis was conducted to trace out the significant factors for full immunisation. Variables significant at p ≤0.05 were included in the multivariate logistic regression analysis. Variance inflation factor (VIF) was calculated to detect any multicollinearity in the model. The low value of average VIF (3.07) confirms no notable multicollinearity among variables. Linktest command through Stata/SE V.13.0 indicated that the constructed model was well specified where the Hosmer-Lemeshow test statistics of goodness of fit indicates the acceptance of the model. Data cleaning, validation and all statistical analyses were performed using Stata/SE V.13.0 (Stata Corporation).

Patient and public involvement

No patients were involved in developing the research question, outcome measure and design of the study. We are unable to disseminate the results of the research directly to study participants.

Descriptive statistics

Distribution of study participants across the sociodemographic variables is presented in table 2. The proportion of full vaccination was 86% (5356 out of 6230 children). Among the participants, 74.7% were from rural areas and 35.1% were from the Dhaka division. Participants were almost equally distributed by sex and age categories. Among the participants, 37.7% was the first-born child, and most of the children’s families (57.9%) consisted of four to six members. Majority of the mothers (80.4%) were aged between 20 and 34 years. 46.1% of the mothers and 29.8% of fathers had secondary-level education. Less than half of the mothers (40.9%) had access to mass media, for example, radio and television. In addition, a majority of the mothers reported that they had the capacity for decision-making with regards to both their own as well as their children’s healthcare.

Vaccination coverage rate among study participants

Full immunisation coverage rate of study participants is also presented in table 2. Findings revealed that full immunisation coverage was slightly higher in urban areas (88.5%) as compared with rural areas (85.1%) and was found to be the highest in the Rangpur division (91.5%) followed by Khulna (89%), Dhaka (88.8%) and the lowest in Sylhet division (69.8%). The immunisation coverage across individual vaccines is shown in figure 1. Full immunisation coverage was similar across sex and age categories of the children. However, first-born child had slightly higher vaccine coverage rates (88.4%) than the other (87.7% for second and 81.6% for third or more birth order).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Coverage rate by individual vaccines (BCG, pentavalent, oral polio vaccine (OPV) and measles) across administrative divisions in Bangladesh (map obtained from open access source and edited according to study findings).

Full immunisation coverage was highest among the children whose parents had higher educational attainment (94.2% and 93.7% for higher educated mother and father, respectively). Vaccine coverage rate was also higher among the children of mothers who have access to mass media. Full vaccination coverage was found to increase in accordance with a higher economic status of the children’s family (73.9% for the poorest to 93.2% for the richest quintile). Full immunisation coverage rate was higher among the children whose mother had her own healthcare decision-making ability, including that of her children’s healthcare. The overall coverage rate of BCG vaccine was the highest (97.1%) followed by OPV-1 (97%), pentavalent 1 (96.6%), OPV-2 (95.8%) and pentavalent 2 (95.4%) where the coverage of measles was the lowest (88%) (figure 2). Furthermore, the coverage of each vaccine by urban and rural areas is also presented in figure 2.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Coverage rate by individual vaccines across place of residence of respondents (urban vs rural) in Bangladesh. OPV, oral polio vaccine.

Factors associated with full immunisation coverage

Table 3 shows factors associated with the full-immunisation coverage among children aged 12–59 months in terms of both unadjusted and AORs. The analysis shows that several sociodemographic factors, like administrative divisions, age of children, family size, parental education, wealth index and child healthcare decision-makers were associated with the full immunisation status in both models. From the univariate analysis, we found that the area of residence, childbirth order, access to mass media and mothers’ healthcare decision-maker were significant factors for full immunisation. Urban children were significantly 1.35 times more likely to be fully vaccinated than rural ones (95% CI 1.13 to 1.60, p=0.001), and the first-born child was 1.71 times more likely to be fully vaccinated than the child of third or more birth order (95% CI 1.45 to 2.03, p<0.001). We also observed that mothers of the children who had access to the mass media had higher odds than their counterparts (OR=2.01; 95% CI 1.72 to 2.36, p<0.001). Similarly, parental education and wealth status played a significant role for full immunisation coverage among children. Findings revealed that children of higher educated mother were significantly 4.93 times (95% CI 3.37 to 7.22, p<0.001) and children of higher educated father were 4.03 times (95% CI 2.98 to 5.44, p<0.001) more likely to be fully immunised than children of uneducated mothers and fathers. Findings also revealed that children from the richest households were significantly 4.27 times (95% CI 3.31 to 5.52, p<0.001) more likely to be fully immunised than the poorest.

In the adjusted model, the likelihood of being fully vaccinated was significantly lower in the Sylhet division than all other regions. We found that the number of full immunised children was significantly higher in Rangpur division (AOR = 3.46; 95% CI 2.45 to 4.88, p<0.001) followed by the children from Dhaka (AOR = 2.59; 95% CI 2.05 to 3.28, p<0.001), Khulna (AOR = 2.33; 95% CI 1.62 to 3.33, p<0.001) and Rajshahi (AOR = 1.96; 95% CI 1.46 to 2.64, p<0.001) division. The odds of full immunisation differed across age groups of the children. For instance, children aged 48–59 months had the highest odds of being fully vaccinated (AOR=1.32; 95% CI 1.06 to 1.64, p=0.013) compared with children aged 12–23 months. Family size was also appeared to be a significant factor for full immunisation as medium households (four to six members) often fully immunised more than their counterpart (AOR=1.56; 95% CI 1.32 to 1.86, p<0.001). The findings also revealed that parental education was significantly associated with full immunisation. It was observed that the odds of being full immunisation were increased as the educational attainment of parents increased, and that children from higher educated mother (AOR=1.96; 95% CI 1.21 to 3.17, p=0.006) and higher educated father (AOR=1.55; 95% CI 1.05 to 2.29, p=0.026) were more likely to be full vaccinated compared with those with uneducated parents. Children from the richest households had the highest odds of being fully immunised (AOR=2.20; 95% CI 1.50 to 3.21, p<0.001) followed by the middle wealth quintile (AOR=1.78; 95% CI 1.37 to 2.30, p<0.001). The lowest odds were found among the children whose healthcare decision was made solely by their father (AOR=0.69; 95% CI 0.51 to 0.92, p=0.012) compared with children with healthcare decisions that were made by the mothers.

Strengths and limitations

The study has several limitations. The study was based on secondary data where immunisation cards and mother’s report were considered as the source of information for their child’s immunisation status. Therefore, the potential of recall bias in our results cannot be ignored. Nonetheless, the mother’s report is considered to be a valid measure of coverage in the absence of a health card, especially in developing countries.41 The explanatory variables were selected based on previous studies and relied on the information available from the dataset. Therefore, there might be some other potential predictors that might be influenced by full immunisation which were not captured in this study. Despite the limitations, the study results can be generalised at a country level because the study used data from the latest nationally representative household survey. Thus, our findings are still significant and relevant in drawing attention to the health policy-makers in ensuring the benefit of vaccination for the betterment of child health.