Machine Learning Analysis of Maternal Mortality Determinants in Nigeria: An Exploratory Study

BACKGROUND: Nigeria confronts an ongoing maternal mortality crisis, accounting for 28.7% of global maternal deaths from 2.9% of the world's population. While proximate clinical causes are well characterised, the relative importance of distal socioeconomic and health system determinants remains poorly quantified, limiting evidence-informed prioritisation of scarce resources. OBJECTIVES: To explore the predictive influence of individual, population, and health system factors on maternal mortality in Nigeria using machine learning approaches. METHODS: This exploratory ecological study used annual national-level data spanning Nigeria's independence to the most recently available dataset (1960-2024). XGBoost and Random Forest ensemble algorithms were applied to examine associations between multidimensional determinants and the maternal mortality ratio. Missing data were addressed using exponential triple smoothing, and model stability was assessed via 5-fold cross-validation. RESULTS: Female tertiary education and service-sector employment showed the strongest and most stable associations with lower maternal mortality ratios. Early marriage, vulnerable employment, rural residence, and relative poverty were associated with a higher mortality ratio. Determinants of health-system financing were largely unstable; government per capita expenditure was associated with a lower maternal mortality ratio, whereas external funding dependency was associated with higher mortality ratio. CONCLUSION: This study challenges traditional perspectives on determinants of maternal mortality while highlighting key areas for policy intervention. Future research should examine the relationships between education, employment, and maternal mortality ratio through a longitudinal study to inform strategic investments.