RE-EVALUATION OF TEMPERATURE-MORTALITY MODEL USING AUTOMATED METEOROLOGICAL DATA AQUISITION SYSTEM

点击翻译按钮获取中文摘要

(1) Correspondence between the meteorological data and city (The word city referred here represents village or town as well as city) : We decided to take the points of Automated Meteorological Data Acquisition System (AMeDAS) within 15km of the barycenter of the city, and to set the calculated average of daily maximum temperature measurements of these points as the city measurement. We used 15km because 10km or 5km left some cities without AMeDAS points.(2) Preparation of data for the analysis : Prime Minister's Office provided the computerized population data by age and sex for each city, Ministry of Health and Welfare provided the computerized death certificate file, and Meteorological Agency provided AMeDAS data. We used city number assigned by Japanese government to combine the above three types of data, and calculated number of deaths, number of person-days by daily maximum temperature category for each city.(3) Comparison between city-based analysis and prefecture-based analysis : In the city-based analysis, we construct prefectural number of deaths and person-days by combining all the cities in the prefecture. On the other hand, in the prefecture-based analysis, we used only one measurement point for each prefecture, which is located in the capital cities. Thus, if there is a difference between the temperature in the capital city and that in other cities, misclassification of deaths and person-days would occur. In this analysis, we used the data in 1995 and restricted the subjects to 65+years old.(4) Result : The misclassification pattern of deaths was so similar to that of person-days that the mortality rates were not biased. This suggests that the prefecture-based analyses studied previously were valid. However, we need to use city-based data in projecting the number of deaths by multiplying mortality rate with corresponding person-days, because the temperature-specific number of person-days is biased.