Assessing and managing heat vulnerability in Henan Province, China: An entropy-weighted and random forest framework

Background: Extreme heat poses growing public health challenges, particularly in rapidly urbanizing regions. Assessing heat vulnerability at the county level is crucial for guiding adaptation strategies. Methods: A county-level Heat Vulnerability Index (HVI) was developed for Henan Province, China, within the Exposure-Sensitivity-Adaptation (ESA) framework. In China, counties represent the smallest administrative units with consistent socioeconomic and health data, typically covering populations of several hundred thousand residents. Indicator weights were objectively derived using the entropy weight method, and both additive and multiplicative aggregations were compared to evaluate robustness and external validity. Random Forest (RF) models identified dominant drivers, with interpretability provided by aggregated SHAP importance scores. To externally validate the HVI, heatwave-related mortality risks were modeled with distributed lag nonlinear models and meta-analyzed across HVI quartiles to assess trend significance. Results: The HVI revealed pronounced spatial heterogeneity, with high-vulnerability belts concentrated in central and northern Henan. Socioeconomic indicators—including educational attainment, social insurance, and healthcare accessibility—consistently outweighed environmental and demographic factors. Trend analyses of heatwave-related mortality further indicated that the additive HVI demonstrated stronger and more stable associations with heatwave-related mortality. At the dimension level, exposure was driven by heatwave days and built-up intensity, sensitivity by age structure and student ratios, and adaptation by education, economic resources, and social protection. Conclusion: Heat vulnerability in Henan Province is shaped more by socioeconomic inequalities than by environmental exposure alone. The additive HVI better captures cumulative vulnerability effects, offering a robust, interpretable decision-support framework to support regional climate adaptation planning and adaptive environmental management.