Regional economy, physical environment and self-rated health among older adults in China: Evidence from XGBoost and SHAP.

Although the effects of the regional economy and physical environment on self-rated health among older adults in China have been widely recognized, their confounding relationships and non-linear impacts remain insufficiently understood. This study draws on the 2015 1% Population Sample Survey of China and employs an integrated approach combining the XGBoost machine learning model with SHAP (SHapley Additive exPlanations) to systematically evaluate the relative importance of individual socioeconomic status (SES), the regional economy, housing conditions, and environmental factors on older adults' self-rated health (SRH). A multi-scale Geographically Weighted Regression (MGWR) model assessed how the regional economy moderates the association between environmental thresholds and SRH. The findings indicate that older adults' self-rated health in China is shaped by multiple factors, with individual socioeconomic status playing a dominant role, followed by environmental conditions, while regional economy (measured by per capita GDP) has a stronger contribution than regional inequality (measured by the Theil Index). The relationship between regional economy and health is complex and non-linear. Higher per capita GDP is consistently associated with better health outcomes; while, the non-linear relationship between the Theil Index and health implies that merely reducing disparities may not lead to uniform health improvements. XGBoost and SHAP analyses reveal clear environmental threshold effects: PM2.5 above 40 μg/m3 sharply increases health risks, per capita housing area above 98 m2 improves health, and optimal outcomes occur at NDVI values of 0.50-0.60, with diminishing returns beyond 0.75. In contrast, quantile regression detects no significant threshold breakpoints for PM2.5 and housing, although per capita GDP and the Theil Index substantially moderate their impacts. NDVI shows the most robust and consistent threshold effect, with no significant moderation by regional economic factors. MGWR analysis shows that the influence of environmental and housing thresholds on health is moderated by regional economy and exhibits strong spatial heterogeneity. These findings highlight the joint roles of environmental exposure, housing, and regional inequality, providing evidence for targeted public health interventions and equitable, age-friendly regional development strategies.