Japan's national heat warning threshold covers 2-3% of preventable heatstroke deaths

Japan runs one of the world's more developed heat-health warning systems, built around a single national trigger: a wet bulb globe temperature of 33°C. A study published on 16 October puts a number on how much of the country's preventable heatstroke mortality that threshold actually reaches [s1].

The number is 2% to 3% [s1].

How the estimate was built

The authors used daily heatstroke mortality data coded ICD-10 X30 — the code for exposure to excessive natural heat — for all 47 prefectures from 2010 to 2019 [s1]. Over that period 9,702 heatstroke deaths were recorded, concentrated in late summer and among older individuals [s1].

The analysis used a time-stratified case-crossover design based on conditional quasi-Poisson regression combined with a distributed lag non-linear model [s1]. Case-crossover designs compare each decedent against themselves on other days, which removes confounding by anything that does not change day to day — age, sex, chronic illness, housing.

The authors then asked a policy question rather than an epidemiological one. Assuming heatstroke deaths are preventable when a warning prompts intervention, what WBGT threshold would be needed to reduce preventable heatstroke mortality by half? They call that target "target50" [s1].

What they found

The current national threshold — WBGT maximum of 33°C — accounted for only 2% to 3% of preventable heat-related deaths during summer months [s1].

A threshold of approximately 31°C achieved target50 in most prefectures [s1]. Northern regions required lower thresholds still [s1].

Two degrees is a small change on a thermometer and a very large change in how often an alert fires. That is the point: a threshold set high enough to be rare is, almost by definition, set above where most of the deaths occur. Heatstroke mortality is not concentrated in the extreme tail. It accumulates across many moderately hot days, in the same way that most alcohol-related harm comes from moderate drinkers rather than the heaviest ones.

The study also found systematic variation the single national number cannot express. Thresholds differed significantly between summer phases, with lower thresholds needed in early summer than in late summer, and there were regional and demographic differences in heat sensitivity [s1].

Early-summer sensitivity is a well-described physiological phenomenon — acclimatisation takes roughly one to two weeks of repeated exposure, so the first hot spell of a season kills at temperatures that would be tolerated in August. A fixed threshold cannot track that.

Why the second paper matters here

A projection study published on 3 October covers the other half of the question: what happens to Japanese heatstroke mortality as the climate and the population both change [s2].

The authors projected heatstroke mortality using ICD-10 X30 for all 47 prefectures across two age groups, under five climate models and three emissions scenarios aligned with shared socio-economic pathways, comparing a base period of 1995–2014 against 2031–2050 and 2081–2100 [s2].

Without long-term heat adaptation, the average heatstroke mortality rate by the end of the century was projected to rise 4.26-fold for the under-65 group and 3.63-fold for the 65-and-over group relative to the base year, across all climate models and emissions scenarios [s2]. With adaptation factored in, those increases fell to 1.76-fold and 1.49-fold [s2].

The projected case counts tell a similar story: without adaptation, 2.07-fold for under-65s and 3.86-fold for 65-plus; with adaptation, 0.88-fold and 1.58-fold [s2]. A figure below 1.0 means fewer deaths than the base period despite a hotter climate.

What the two papers say together

The projection study establishes that adaptation is the dominant term. The difference between adapting and not adapting is larger than the difference between emissions scenarios over the periods studied [s2]. The threshold study is a concrete specification of what one piece of that adaptation could look like, and finds the current setting is capturing almost none of the available benefit [s1].

That is an unusually actionable pairing. Most climate-health research identifies a burden. These two identify a burden and a specific, cheap, already-built lever — the number in an existing warning system — that moves it.

The assumptions doing the work

Both studies rest on modelling assumptions that deserve to travel with their numbers.

The threshold analysis assumes heatstroke mortality is preventable when interventions are prompted by a warning [s1]. That assumption is the hinge of the entire result. If people do not change behaviour when an alert fires — and alert fatigue from more frequent warnings is a real risk of lowering the threshold — then the modelled preventable fraction overstates what a lower trigger would achieve. The study models the threshold; it does not measure the behavioural response to one.

The projection study depends on how "long-term heat adaptation" is parameterised, which is an extrapolation from observed historical declines in temperature sensitivity. Whether that historical trend continues as temperatures move outside the observed range is unknown, and the study cannot test it [s2].

Both restrict themselves to ICD-10 X30 — deaths coded directly to heat [s1][s2]. That is a deliberately narrow definition. The much larger indirect burden of heat, expressed as cardiovascular, renal and respiratory deaths during hot periods, is not in either count.

What to watch

Japan's system is reviewed against evidence, and the paper explicitly offers a scientific basis for revisiting it [s1]. Whether the threshold moves, and whether it moves to a single lower national number or to prefecture-specific and season-specific values, is the decision these two studies were written to inform.

Sources

• [s1] Determining Location-Specific Thresholds for Heat Warning Systems to Mitigate Heatstroke Mortality in Japan — Environment & Health (ACS), published online 16 October 2025. https://doi.org/10.1021/envhealth.5c00113
• [s2] Future heatstroke mortality in Japan: Impacts of climate, demographic changes, and long-term heat adaptation — Environmental Research, published online 3 October 2025. https://doi.org/10.1016/j.envres.2025.123012