Applying machine learning to predict quality ANC determinants in Bangladesh: a BDHS-2022 cross-sectional study

Quality antenatal care (ANC) is critical for maternal and neonatal health. Despite improvements in healthcare, disparities in ANC access and quality persist, particularly in underserved areas of Bangladesh. This study aimed to identify the key determinants of quality ANC in Bangladesh and provide evidence to support data-driven maternal health strategies aligned with Sustainable Development Goal (SDG) 3. This study analyzed data from 3,549 women aged 15-49 years from the BDHS 2022. Machine learning models, including Random Forest, XGBM, Neural Networks, and Logistic Regression, were applied to predict quality ANC. Feature importance was assessed using SHapley Additive exPlanations (SHAP) and Gini-based rankings to identify the most influential predictors. Only 21.9% of women received quality ANC. Wealth index, maternal and partner education, maternal age, media exposure, and urban residence emerged as the strongest determinants. Random Forest demonstrated the highest predictive performance (accuracy: 74.1%, precision: 79.8%, F1-score: 0.83). SHAP and feature importance analyses confirmed that wealth index was the most influential predictor. Significant inequalities in ANC quality and coverage exist in Bangladesh. Targeted interventions addressing socioeconomic and educational disparities, along with improved media outreach and urban-rural healthcare access, are essential to enhance maternal healthcare. These findings provide actionable insights for policy and programs aiming to achieve SDG 3 and reduce maternal mortality.