Data, setting, and participants

We used secondary data from the nationally representative cross-sectional Bangladesh Demographic and Health Survey (BDHS) 2017–2018. The BDHS 2017-18 data was collected using two-stage stratified cluster sampling making the data representative at the administrative division level, as well. As we aimed to examine the missed opportunities in ECP awareness and family planning counseling by FWA and ANC visits, we conducted the analyses on all women (5012) aged 15–49 who gave their most recent live birth 3 years preceding the survey. Figure 1 presents the data flow of the analytical sample.

Fig. 1

Data flow of the analytical sample

Outcome measures

Outcome variable 1: ECP awareness was the first outcome variable with two categories: aware of ECP (if a woman ever heard about ECP from any source) and unaware of ECP (if a woman never heard about ECP from any source).

Outcome variable 2: We constructed the second binary outcome variable- “missed opportunity in providing FP counseling during ANC” as follows:

a.

missed opportunity (if women had ANC visits but did not receive FP counseling during ANC),

b.

utilized opportunity (if women had ANC visits and received FP counseling during ANC).

Conceptual framework and covariates

Conceptual framework for outcome 1: We conceptualize that ECP awareness can be influenced by three broad domains: exposure to the FP program, contextual factors, and socio-demographic characteristics (Fig. 2). If a woman is exposed to FP programs like FWA or ANC visits, she may have a higher chance of knowing about ECP than a woman who is unexposed to such programs. Contextual factors shape customs and beliefs of the community which may further influence women’s knowledge about FP methods. For example: A community with conservative customs and beliefs will have less discussion about FP than a community where people are concerned about FP rights and choices. Thus, conservative customs of the community may shape women’s perception and urge of FP which may lessen the chance of knowing about ECP. Contextual factors may further influence ECP awareness through varying intensities of FP programs because FP programs of a region are designed based on the needs of that region. Women from advantageous socio-demographic groups (i.e. better education and wealth, frequent media exposure) may have higher ECP knowledge because of their better exposure to health services and their desire to limit family size.

Fig. 2

Conceptual framework of emergency contraceptive pills awareness

Covariates for modeling outcome 1: Under each domain, we selected covariates based on earlier studies from Bangladesh [19, 20], Nepal [21], India [22, 23] and Sub-Saharan Africa [24]. The details of the covariates are presented in Table 1. The wealth status was constructed using a principal component analysis based on household goods, which ranged from having a television to having a bicycle or a car, and households’ sources of drinking water, sanitation facilities, and flooring materials. Detailed about household wealth can be found in BDHS 2017–18 final report.

Table 1 Covariate construction: Definition of all the categories under each covariate

Conceptual framework for outcome 2: We conceptualized that missed opportunities for FP counseling during ANC can be shaped by two broad domains: providers’ characteristics and women’s characteristics. Providers’ perceptions about FP counseling during ANC can be one of the key drivers of counseling. In regards, medically trained providers (MTP) are likely to hold a more positive attitude towards post-partum FP counseling during ANC than non-MTP, maybe because of receiving more training on ANC services. As we aimed to explore whether the providers are being selective in providing FP counseling during ANC, we included broad sociodemographic groups as covariates. As TFR is comparatively high among women from disadvantageous socioeconomic groups like low education, rural residents, and lower wealth quintiles, they may have a higher chance of not receiving FP counseling during ANC. We also hypothesized that women with first or second parity may be less likely to be counseled because providers may emphasize higher parity cases for FP counseling [25, 26].

Covariates for modeling outcome 2: Under the domain provider’s characteristics, BDHS 2017–18 data only allows to include provider type (MTP or non-MTP). BDHS 2017–18 considered qualified doctors, nurses/midwives/paramedics, family welfare visitors, community skilled birth assistants, and sub-assistant community medical officers as medically trained providers. Under sociodemographic characteristics, we included women’s education, number of children ever born, type of residence, administrative division, and wealth quintile.

Statistical analysisAnalysis under objective 1

Firstly, we estimate the missed opportunities in ANC and FWA visits for generating ECP awareness at the national level and then at the divisional level to explore regional variation. Estimation procedures were as follows:

Missed opportunity in ANC visits in generating ECP awareness: This was measured as the percentage of women who had ANC visits in their most recent live birth but never heard about ECP among women who gave their last live birth in 3 years preceding the survey.

Missed opportunity in FWA visits in generating ECP awareness: This was measured as the percentage of women who had FWA visits in the last 6 months preceding the survey but never heard about ECP among women who gave their last live birth in 3 years preceding the survey.

For measuring missed opportunity we used “all women who gave live birth in 3 years preceding the survey” as the denominator because this will reflect where FP awareness could reach if all contact points could reach their highest mark.

Secondly, we estimated the prevalence of ECP awareness across the factors considered under each domain. As BDHS 2017–18 used a cluster sampling design, we used a mixed-effect multiple logistic regression model considering random intercept at the cluster level to examine the factors associated with ECP awareness. The random intercept model can deal with the hierarchical nature of the clustered data by incorporating both fixed effects and random effects at the cluster level. These random effects capture the variation across clusters by explicitly modeling the clustering structure and fixed effects represent the average relationship between the predictors and the response variable across all clusters.

Analysis under objective 2

Firstly, we estimated the missed opportunities for FP counseling during ANC by providers’ type (medically trained and untrained) at the national level and then at the divisional level to explore regional variation. Missed opportunity in FP counseling during ANC was estimated as the percentage of women who had ANC visits in their most recent live birth but did not receive FP counseling among women who gave their last live birth 3 years preceding the survey. For measuring missed opportunity we used “all women who gave live birth in 3 years preceding the survey” as the denominator because this will reflect where FP awareness could reach if all contact points could reach their highest mark.

Secondly, we measured missed opportunities among “women who had ANC visits for their most recent birth” to understand whether the providers are being selective in providing FP counseling during ANC. We also used a mixed-effect multiple logistic regression considering women nested within clusters to examine the factors associated with missed opportunities in FP counseling during ANC.

To reduce bias, we incorporated appropriate sampling weights that adjusted for the complex survey design characteristics of BDHS. For all the statistical analyses, we used Stata V.14.0 (Stata SE V.14, Stata Corp., College Station, Texas, USA).