Development of prediction formula for the number of heat stroke patients considering regionality and seasonality

  • Asami SHIMIZU, 2016: Development of prediction formula for the number of heat stroke patients considering regionality and seasonality .


In this study, we examine the relationship between meteorological factors and the number of hospital transport due to heat stroke for people aged 65 years and older who have the highest number of heat stroke occurrence for each area, and clarify regionality and seasonality. Also, based on the characteristics of those, we aimed to develop a practical prediction formula that predicts the number of heat carrier patients. Firstly, correlation analysis was performed for each prefecture using the data of heat stroke occurrence andmeteorological condition, and meteorological factors which most affected the number of heat stroke occurrence were selected. As a result, the correlations between heat stroke occurrence and daily average temperature as well as daytime average WBGT were high nationwide. Comparing these two factors, correlation with daily average temperature was correlated nationwide, but the regional nature that the daytime average WBGT was higher in the Kyushu region. Secondary, using data of the Kanagawa prefecture as a model case, Poisson regression was performed using a generalized linear model to develop a prediction formula to find the number of hospital transport due to heat stroke. As a result, it was shown that heat stroke was prone to occur in early summer and it was hard to occur in late summer on the Kanagawa prefecture. Thirdly, in order to examine the influence of meteorological factors other than daily average temperature and daytime average WBGT, we performed multiple polynomial Poisson regression by adding other meteorological factors as a second explanatory variable for each prefecture. Among the added explanatory variables, meteotological factors whose correlation coefficients between the predicted value and the measured value increased at a relatively large number of prefecture were black-bulb temperature and solar radiation. In particular, it was remarkable in the Kansai, Tohoku, and Hokuriku area, suggesting that the effect of solar radiation is more effective than the other areas. Finally, we divided prefectures according to characteristics of occurrence of heat stroke into several groups to develop more useful prediction formula. A prediction formula was created for each group with convertional meteorological factors, daily averaged temperature. As a result, prefectures could be classified into six groups according to the geographical position and urbanization of prefecture. Comparing between groups, it was found that heat stroke was likely to occur in groups with low daily average temperature, and heat stroke was difficult to occur in relatively more urbanised prefectures.

Keywords : Heat stroke, Prediction formula, GLM, Poisson regression