A Machine Learning Analysis of Maternal Mortality Determinants in Nigeria

Background: Nigeria confronts an ongoing maternal mortality crisis that contributes disproportionately to global maternal deaths. While the clinical causes are well-documented, the underlying socioeconomic determinants and entrenched health inequities remain underexplored, creating a critical knowledge gap that hinders effective resource allocation by stakeholders. Objectives : To determine the relative predictive influence of individual, population, and health system factors on maternal mortality in Nigeria using machine learning approaches. Methods: This ecological time series study examines associations between individual, population, and health system factors and maternal mortality in Nigeria (1960-2024). Using data from reputable institutions, we employed XGBoost and Random Forest algorithms to analyse the complex relationships between these multidimensional factors and maternal mortality outcomes. Results: Analysis revealed that educational attainment and service sector employment consistently predict reduced maternal mortality in Nigeria. Early marriage, vulnerable employment, rural residence, and poverty were associated with increased mortality but showed less predictive reliability. Health financing demonstrated lower reliability overall, with government expenditure decreasing maternal mortality while external funding dependence tended to increase mortality. Conclusion: This study challenges traditional perspectives on maternal mortality predictors while identifying key policy intervention areas. Future research should examine the complex relationships between education, employment, and maternal mortality to inform strategic investments.