Temporal changes in mortality impacts of heat wave and cold spell in Korea and Japan

Heatwave-Related Mortality Risk and the Risk-Based Definition of Heat Wave in South Korea: A Nationwide Time-Series Study for 2011–2017

“…The threshold points and the maximum risk point for heatwaves were observed around the 95th percentile and the 99th percentile of the mean temperature, respectively. These findings are consistent with those of previous studies [ 22 , 25 , 27 ]. Guo et al evaluated 400 communities in 18 countries and reported that higher temperature cut-offs had higher effect estimates for the heatwave in almost every country [ 22 ].…”

“…Guo et al evaluated 400 communities in 18 countries and reported that higher temperature cut-offs had higher effect estimates for the heatwave in almost every country [ 22 ]. A study in Korea and Japan also reported that the heatwave-related mortality risk sharply increased at the 95th percentile of daily mean temperature and showed the maximum heatwave-mortality risk at the 99th percentile [ 25 ]. Another study reported that heatwave-related mortality relative risks started to increase around the 95th percentile of the temperature, and the highest heatwave-related risk was identified at the 99th percentile in Melbourne, Australia [ 27 ].…”

“…This study has some limitations. Previous studies have reported that the vulnerability to heat changed during the last decades [ 25 , 35 ] and accordingly suggested that the definition of heatwave should be changed over time [ 25 ]. However, we could not consider the temporal variations in heatwave-related mortality risk, due to our short study period (2011–2017).…”

“…Particularly, this study included rural districts that have been rarely studied previously. In addition, we used the district-level time-series data that has higher spatial resolution than metropolitan city-level data used in previous Korean studies [ 25 , 26 ].…”

“…The day of week on day t (DOW t ) was also controlled as a categorical variable. These modeling specifications were based on previous studies [ 24 , 25 ]. In the second stage, we pooled the estimates in the total population, for urban and rural areas, and for each sub-population using meta-analysis with potential confounder variables: population, local income tax, number of beds in the hospitals, proportion of total urban forest area, latitude, longitude, mean temperature, and range of daily mean temperature ( Table S2 ).…”