Associations between model-derived annual temperature metrics and physical health across 22 countries

Environmental conditions are relevant to physical health. Associations between annual temperature metrics and subjective physical health outcomes remain insufficiently characterized across diverse populations. This study uses 273,031 pooled observations from 22 countries in the Global Flourishing Study (GFS, 2023–2024), integrated with climate data from Coupled Model Intercomparison Project Phase 6 (CMIP6). We analyze the relationship between Annual Mean Temperature, Annual Temperature Standard Deviation, and Overall Physical Health. We employ a unified two-stage Extreme Gradient Boosting (XGBoost) framework to compare predicted physical-health scores under annual mean temperature increases of 1 °C and annual temperature standard deviation increases of 0.1 °C. Direct contrasts hold generated symptom and behavior probabilities fixed; total contrasts update these probabilities. Uncertainty is summarized using 100 country-stratified native-grid cluster bootstrap draws that refit both stages, with fixed hyperparameters. For annual mean temperature + 1 °C, the pooled direct and total prediction contrasts are 0.00488 (95% bootstrap interval − 0.00219 to 0.00887) and 0.00547 (− 0.00333 to 0.00975) health-score points, respectively. For annual temperature standard deviation + 0.1 °C, the corresponding contrasts are − 0.00155 (− 0.00455 to − 0.00037) and − 0.00054 (− 0.00420 to 0.00157). Country-level contrasts vary in direction. Both pooled indirect-contrast intervals include zero. Across the retained 22-country sample, model-based prediction contrasts indicate small pooled differences in self-rated physical health, alongside country-level variation in direction and magnitude. The magnitude and direction of country-level contrasts depend on the temperature data source and analytical specification. The findings provide a basis for further research on how local environmental and social conditions shape temperature–health associations.