Differential Associations of Heat Metrics with Health Statistics: A Spatiotemporal Analysis of Temperature Indicators and Mortality Across Thailand Provinces

Background: Assessments of climate change health impacts often rely on ambient temperature, often neglecting the role of humidity and wind speed. This can be problematic in countries with high climate diversity and pronounced climate warming as it is the case in many South-East Asian nations. This study compared trends in multiple heat metrics (i.e., actual heat, Heat Index, Humidex and Apparent Temperature) across Thailand provinces and evaluated their respective associations with province-level mortality. Methods: We analysed daily meteorological data (2006-2024) to calculate trends across heat metrics for 71 provinces (6 missing). Then, using negative binomial mixed-effects regression models, we assessed the 1-year lagged associations between each annual heat indicator and all-cause mortality from 2008 to 2023, adjusting for year and province. Results: Composite heat indices increased at a faster rate than ambient temperature in the majority of provinces. All composite heat metrics surpassed actual heat in predicting mortality. Specifically, our analyses revealed that a 1dC increase in one-year lagged Apparent Temperature was associated with 47,196 Excess Deaths (ED) (95% CI: 13,287 to 82,704) over the study period, representing the strongest association among the metrics tested. However, the strength of the association varied across regions with different climate trajectories. Stratified analyses by death in- and out-of-hospital settings show strong association of Humidex and Apparent Temperature with hospital deaths at 1-year lag but out-of-hospital deaths show no associations and broad confidence intervals. Conclusion: Heat stress is increasing faster than Apparent Temperature across Thailand. While composite measures like the Apparent Temperature and Humidex are stronger predictors of national-level mortality, regional variations in these relationships underscores the need to develop local heat metric that account for both local characteristics and climate change patterns to accurately tailor public health responses.