Dataset and Code for: Explainable AI for Multifactorial Mortality Risk Assessment in the Northeast Corridor

This study develops and validates an integrated socioeconomic-environmental risk modeling framework for mortality risk prediction across the Northeast Corridor of the United States. Combining data from the U.S. Census Bureau, Environmental Protection Agency, National Oceanic and Atmospheric Administration, and Centers for Disease Control and Prevention, we analyze 72 counties across five states using a machine learning approach (Random Forest). A Random Forest model incorporating 20 socioeconomic, environmental, demographic, and healthcare variables achieves exceptional predictive performance at the regional level (R² = 0.935), with state-level models demonstrating varied but substantial explanatory power for four of the five states (R² = 0.510-0.758), while one state model showed limited fit. SHAP (SHapley Additive exPlanations) analysis reveals median income as the dominant predictor across all jurisdictions (94.5% regional importance), while environmental factors—particularly heat wave days and PM2.5 concentrations—emerge as significant secondary predictors with demonstrated threshold effects. Our analysis identifies environmental justice patterns, with counties exceeding EPA PM2.5 standards showing 1.7% higher mortality rates than low-pollution counties. The research contributes both methodological innovations through explainable AI implementation and practical insurance applications through a tripartite risk classification system enabling actuarially sound premium differentiation. We further propose a regulatory compliance framework addressing transparency requirements through SHAP explainability and demonstrate how environmental factors can serve as objective, non-discriminatory predictors in insurance risk classification. This integrated approach advances risk modeling methodology while providing actionable insights for insurers, regulators, and public health authorities addressing complex multifactorial mortality determinants.