Health beliefs, medication adherence, and blood-pressure control in rural Bangladeshi hypertension: an exploratory sequential mixed-methods study

Abstract Background Hypertension control remains poor in low and middle income countries, including Bangladesh, where medication adherence is often suboptimal. How patients' health beliefs translate into adherence and blood pressure control in rural settings is poorly understood. We examined how the six Health Belief Model (HBM) constructs relate to medication adherence and, in turn, systolic blood pressure among rural Bangladeshi adults with hypertension. Methods We conducted an exploratory sequential mixed-methods study in Kamalganj, north-east Bangladesh (October 2025–February 2026). The qualitative phase comprised 28 sessions (16 in-depth interviews, 10 key-informant interviews, 2 focus group discussions) analysed by directed content analysis using the HBM as a deductive framework. A content-validated 38-item HBM questionnaire was administered alongside the Bangladesh Medication Adherence Scale (BMAS) to 461 hypertensive adults in a community-based cross-sectional survey. Six covariate-adjusted PROCESS Model 4 analyses tested HBM construct → adherence → systolic blood pressure pathways. Qualitative and quantitative components were integrated through joint displays and meta-inferences. Results Three qualitative themes emerged: Health Risk Appraisal, Treatment Decision Appraisal, and Activity and Agency. Blood pressure was controlled in 41.4% of participants and adherence was optimal in 44.2%. Perceived Barriers (β = 0.375) and Self-Efficacy (β = 0.304) were the strongest independent predictors of adherence (both p < 0.001). Cross-sectional decomposition revealed three distinct patterns: statistical mediation through adherence (Perceived Barriers, Cues to Action); an adherence-independent association for Self-Efficacy (c′=−6.41 mmHg, 95% CI − 12.68 to − 0.14; hypothesis-generating); and non-significant pathways for Susceptibility, Severity, and Benefits. Treatment-seeking was symptom-driven and socioeconomic status was a strong moderator. Belief endorsement alone did not predict outcome. Conclusions Hypertension control in rural Bangladesh is driven less by patient knowledge than by structural conditions. Reducing structural barriers and cultivating domain-specific self-efficacy are the principal modifiable levers for improving control.