Predicting heat-related morbidity in Japan through integrated meteorological and behavioral factors

Abstract

Heat-related illness has become a critical public health concern due to the increasing frequency of extreme heat events, intensified by global climate change. Accurate prediction of these illness is crucial to managing public health risks. Existing prediction models often rely solely on meteorological factors, limiting their effectiveness. This study aimed to improve prediction models for heat-related illness by integrating meteorological variables (ambient temperature, humidity, and solar radiation) and behavioral factors (holiday periods and rainy days) that influence a person's exposure to extreme heat. Data from all 47 prefectures in Japan were analyzed from 2014 to 2019 and in 2023. Our refined model accounted for regional variations in heat acclimatization and behavioral patterns, and was validated using the leave-one-out cross-validation method. The accuracy of the proposed approach was reflected by mean absolute error of 1.30 for outdoor cases and 0.95 for indoor cases, representing improvements of 11.79 % and 3.72 %, respectively, compared to the previous models that considered solely on temperature. The model was implemented as a web-based platform for real-time risk assessments and to help emergency services and local governments manage medical resources effectively during heatwaves. These findings underscore the critical role of improved predictive models in mitigating the public health impact of global warming. The proposed model is scalable and can be adapted to other regions facing the negative effects of extreme heat, ultimately enhancing public health preparedness and preventive measures.