Explainable Machine Learning for Predicting Health Insurance Enrolment among Women in Kenya.

Health insurance coverage remains a critical challenge in Kenya, particularly among women in the informal sector who are disproportionately uninsured despite government reforms. This study investigates the socio‑economic determinants of enrolment in the National Hospital Insurance Fund (NHIF), now renamed the Social Health Authority (SHA), to understand barriers and opportunities for expanding coverage. The problem addressed is the persistent inequity in access, where policies such as free maternity care have improved service utilization but have not fully resolved disparities in insurance uptake. Using data from the Kenya Integrated Household Budget Survey (KIHBS 2015/2016), the analysis focused on women aged 18 years and above. Machine learning models including Linear Regression, Random Forest, and XGBoost were applied to predict enrolment outcomes. SHAP (SHapley Additive Explanations) was employed to interpret feature importance, highlighting the role of income, education, marital status, and county‑level demographics. Results showed that Random Forest achieved the most reliable performance, capturing meaningful socio‑economic relationships, while XGBoost emphasized demographic patterns with less stability. The findings contribute to the literature by demonstrating the utility of explainable AI in health economics and by identifying key predictors of insurance coverage. Policy recommendations include targeted subsidies for low‑income households, literacy campaigns to improve awareness, and tailored outreach to informal sector workers. This research underscores the relevance of AI adoption in Kenyan insurance systems, aligning with 2025 initiatives to modernize health financing. By combining methodological rigor with policy insight, the study advances understanding of how data‑driven approaches can support equitable health coverage.