BackgroundAmbient temperature affects mortality in susceptible populations, but regional differences in this association remain unclear in Japan. We conducted a time-series study to examine the variation in the effects of ambient temperature on daily mortality across Japan.MethodsA total of 731 558 all-age non-accidental deaths in 6 cities during 2002–2007 were analyzed. The association between daily mortality and ambient temperature was examined using distributed lag nonlinear models with Poisson distribution. City-specific estimates were combined using random-effects meta-analysis. Bivariate random-effects meta-regressions were used to examine the moderating effect of city characteristics.ResultsThe effect of heat generally persisted for 1 to 2 days. In warmer communities, the effect of cold weather lasted for approximately 1 week. The combined increases in mortality risk due to heat (99th vs 90th percentile of city-specific temperature) and cold (first vs 10th percentile) were 2.21% (95% CI, 1.38%–3.04%) and 3.47% (1.75%–5.21%), respectively. City-specific effects based on absolute temperature changes were more heterogeneous than estimates based on relative changes, which suggests some degree of acclimatization. Northern populations with a cool climate appeared acclimatized to low temperature but were still vulnerable to extreme cold weather. Population density, average income, cost of property rental, and number of nurses appeared to influence variation in heat effect across cities.ConclusionsWe noted clear regional variation in temperature-related increases in mortality risk, which should be considered when planning preventive measures.