BACKGROUND: Previous studies have shown that population susceptibility to non-optimum temperatures has changed over time, but little is known about the related time-varying factors that underlie the changes. OBJECTIVE: Our objective was to investigate the changing population susceptibility to non-optimum temperatures in 47 prefectures of Japan over four decades from 1972 to 2012, addressing three aspects: minimum mortality temperature (MMT) and heat and cold-related mortality risks. In addition, we aimed to examine how these aspects of susceptibility were associated with climate, demographic, and socioeconomic variables. METHODS: We first used a two-stage time-series design with a time-varying distributed lag nonlinear model and multivariate meta-analysis to estimate the time-varying MMT, heat and cold-related mortality risks. We then applied linear mixed effects models to investigate the association between each of the three time-varying aspects of susceptibility and various time-varying factors. RESULTS: MMT increased from 23.2 [95% confidence interval (CI): 23, 23.6] to 28.7 (27.0, 29.7) °C. Heat-related mortality risk [relative risk (RR) for the 99th percentile of temperature vs. the MMT] decreased from 1.18 (1.15, 1.21) to 1.01 (0.98, 1.04). Cold-related mortality risk (RR for the first percentile vs. the MMT) generally decreased from 1.48 (1.41, 1.54) to 1.35 (1.32, 1.40), with the exception of a few eastern prefectures that showed increased risk. The changing patterns in all three aspects differed by region, sex, and causes of death. Higher mean temperature was associated (p &lt
0: 01) with lower heat risk, whereas higher humidity was associated with higher cold risk. A higher percentage of elderly people was associated with a higher cold risk, whereas higher economic strength of the prefecture was related to lower cold risk. CONCLUSIONS: Population susceptibility to heat has decreased over the last four decades in Japan. Susceptibility to cold has decreased overall except for several eastern prefectures where it has either increased or remained unchanged. Certain climate, demographic, and socioeconomic factors explored in the current study might underlie this changing susceptibility.