Abstract / Summary
Objectives: To describe the prevalence of having had a live birth or being currently pregnant among ever-married adolescents in Bangladesh and examine its associations with cohabitation timing and partner characteristics among currently married adolescents. Design: Cross-sectional secondary analysis of the 2022 Bangladesh Demographic and Health Survey (BDHS). Methods: Descriptive estimates used 2,449 ever-married respondents aged 15-19. The long questionnaire was administered to 1,635; 1,603 were currently married, of whom 1,601 had complete information for regression. Weighted logistic regression included current age, respondent education, five wealth quintiles, cohabitation age, husband education, signed spousal age gap, residence and division. Sampling weights, strata and primary sampling units (PSUs) were incorporated using a stratified cluster-based Taylor variance estimator. Current working status and an asset-based principal component were examined in separate sensitivity analyses. Results: Among ever-married adolescents, weighted outcome prevalence was 59.7% (95% confidence interval [CI] 57.6-61.7). In the expanded model, cohabitation at ages 15-17 and 18-19, compared with below 15, was associated with lower adjusted odds (adjusted odds ratios [AORs] 0.170, 95% CI 0.113-0.256; and 0.026, 0.015-0.045, respectively). A signed spousal age gap of at least 11 years, compared with at most five years, was associated with higher odds (AOR 1.752, 95% CI 1.241-2.474). Higher husband education was inconclusive (AOR 0.612, 95% CI 0.338-1.109). Cohabitation estimates changed substantially after current-age adjustment. Mean held-out-PSU area under the receiver operating characteristic curve was 0.793, and mean calibration slope was 0.885. An exploratory specification diagnostic indicated possible misspecification (p approximately 0.0205). Conclusions: Cohabitation timing and spousal age gap were associated with childbearing or current pregnancy in the analysed population. Differences in time available for childbearing, selection into marriage and interview-time measurement of covariates constrain interpretation. These findings do not establish causal effects or estimate the impact of delaying marriage or implementing a policy. Regression estimates remain provisional because long-questionnaire weighting requires confirmation.