Comparing Integrated Heat Stress Indicators With Raw Meteorological Variables in Predicting Heat Stroke-Related Ambulance Transportations in Japan

Abstract

The increasing threat of heat stress poses significant risks to human health globally. To quantify heat exposure more effectively, integrated heat stress indicators (HSIs) have been developed to simplify the classification of heat stress severity and assist in public heat warnings. However, their ability to accurately predict daily heat stroke cases has not been fully assessed. In this study, we evaluated the performance of multiple HSIs in forecasting the number of heat stroke-related emergency ambulance dispatches (HT-EADs) across 47 prefectures in Japan and compared their accuracy to models using raw meteorological variables. Our results indicate that, while HSIs simplify the process of assessing heat stress, they generally show lower performances than models based on raw meteorological data. Among the eight HSIs tested, the Wet Bulb Globe Temperature (T_WBG) showed the strongest predictive power, with median R² values of 0.77 and 0.70 for the calibration and validation periods, respectively. However, models incorporating air temperature, relative humidity, wind speed, and solar radiation outperformed T_WBG, achieving R² values of 0.85 and 0.74. We also observed spatial variability in HSI performance, particularly in cooler regions like Hokkaido, where HSIs provided no improvement over temperature alone. Given these findings, we recommend that HSIs be rigorously evaluated with local health data before being used in heat warning systems for specific locations. For predictions requiring high accuracy, raw meteorological variables could be prioritized to ensure greater precision.