Cooling related household expenditure proxies and heatstroke mortality in an ecological panel study of Japanese prefectures from 2013 to 2022

Heatstroke (ICD-10 T67) mortality in Japan concentrates among adults aged ≥ 65 years. Whether cooling-related household electricity expenditure modifies the temperature – mortality relationship remains uncertain across proxy operationalisations. Ecological panel of 47 Japanese prefectures, 2013–2022 ( n = 470). A summer–shoulder electricity expenditure ratio (SSR’) was tested alongside three positive (carbonated drink, ice cream, mineral water) and two negative (gas, kerosene) expenditure proxies from the Family Income and Expenditure Survey, using Poisson regression with two-way fixed effects, prefecture-clustered robust standard errors, and offset on population aged ≥ 65 years. Five alternative thermal indicators provided sensitivity analyses. Temperature anomaly had a large main effect (IRR = 1.325, p < 0.001), robust across all thermal indicators. The primary temperature × SSR’ interaction was non-significant (IRR = 1.031, p = 0.252); carbonated drink reached nominal significance (IRR = 1.062, p = 0.019), with five of six proxies directionally concordant. The 2020 year fixed-effect was non-significant (IRR = 1.10, p = 0.39) despite the period’s highest temperature anomaly (+ 1.57 °C), and the 2020-excluded estimate matched the full sample—indicating that extreme heat, not COVID-19, drove the 2020 mortality peak. Substituting tropical-night counts (days with minimum ≥ 25 °C) for daytime peak temperature yielded a significant negative interaction with SSR’-electricity (IRR = 0.917, p = 0.017), suggesting that cooling-related expenditure specifically attenuates the effect of failed nocturnal cooling on heatstroke mortality. Behavioural and nocturnal heat dimensions, captured respectively through carbonated-drink expenditure and tropical-night exposure, warrant individual-level investigation as targets for heatstroke prevention.