MaternaAI: Enhancing Equitable Maternal Healthcare in Kerala with Fairness-Aware and Explainable Learning Models

Maternal healthcare prediction systems often inherit biases from imbalanced datasets and socio-economic disparities, undermining their value in equitable healthcare policymaking. We present MaternaAI , a fairness-aware and explainable learning framework tailored to enhance maternal healthcare predictions in Kerala, India. The framework focuses on three key indicators: (1) Tetanus Toxoid (TT) booster uptake, (2) immunization coverage, and (3) the percentage of pregnant women completing four or more Antenatal Care (ANC) visits. To address fairness, we propose Adaptive Equity Score Optimization (AESO), a novel, model-agnostic optimization algorithm that dynamically adjusts group equity weights based on real-time disparities. For transparency, MaternaAI integrates explainable AI (XAI) techniques, including SHAP, LIME, and feature permutation methods to enable both global and local interpretability. Empirical evaluation using real-world data from Kerala’s Health Management Information System (HMIS) shows that MaternaAI improves fairness and predictive accuracy across machine learning and deep learning models, offering actionable, interpretable, and equitable decision support for public health stakeholders.