Wildfire smoke keeps killing after the sky clears, a 1.25-million-person study finds

Wildfire smoke is reported as an event. The air goes orange, the advisories go out, people stay indoors, and then the wind changes and the story ends.

A study published in the Proceedings of the National Academy of Sciences on 15 December suggests the accounting should not close there. Researchers at the UC Berkeley School of Public Health, with colleagues at Kaiser Permanente, UCLA, UC San Diego and the University of Washington, followed 1.25 million Kaiser Permanente Southern California members aged 60 and over from 2009 to 2019 [s1].

Over a three-year exposure window, people in the most heavily smoke-exposed group had a 7% higher chance of dying than those minimally exposed [s1].

The finding that changes how you would measure it

The headline percentage is not the most useful part. The measurement choice is.

The team found that cumulative smoke exposure was a stronger predictor of mortality than the frequency of smoke events [s1]. In other words: total dose over years, not how many bad-air days a person lived through.

That distinction cuts against how wildfire smoke is normally tracked. Public communication is built around episodes — days above a threshold, advisories issued, schools closed. If the mortality signal is carried by accumulated exposure instead, then a region with many moderate smoke days may be carrying a burden that an event-counting system never registers.

The risk was not evenly distributed. It was greatest among people aged 60 to 75, and among Black participants [s1].

What the design supports

This is an observational cohort study, and it establishes association, not causation. Smoke exposure is not randomly assigned; people who live in heavily smoke-affected areas differ from those who do not in income, occupation, housing quality, baseline health and access to filtration. Cohort studies adjust for what they can measure and remain vulnerable to what they cannot.

The cohort is also specific: insured members of one health system, in Southern California, aged 60 and over [s1]. Older adults are the group in which smoke effects are most detectable, which is why they were chosen; it also means the estimate should not be transferred to a general population.

Ten years of follow-up in 1.25 million people is nonetheless a substantial base, and the subgroup pattern — larger effects in the younger-old and in Black participants [s1] — is the kind of finding that is hard to produce by chance in a sample that size, though it remains subject to the same confounding as the main result.

The larger picture the number sits in

Wildfire smoke has moved from a regional nuisance to a global mortality contributor. The 2025 Lancet Countdown report, published in October, attributed 154,000 deaths in 2024 to air pollution from wildfire smoke [s2]. For comparison, the same report put deaths from fossil fuel air pollution at 2.5 million a year [s2].

The mechanism is fine particulate matter, and its health effects are among the best-characterised in environmental epidemiology. The World Health Organization estimates ambient air pollution caused 4.2 million premature deaths worldwide in 2019, of which 68% were from ischaemic heart disease and stroke, 14% from chronic obstructive pulmonary disease, 14% from acute lower respiratory infections and 4% from lung cancer [s3].

That distribution is the reason wildfire smoke deaths are systematically undercounted. They present as heart attacks and strokes.

Why it matters

If the operative exposure metric is cumulative rather than episodic, then two policy consequences follow, neither of which is currently standard.

Long-term smoke exposure would need to be tracked at the individual or small-area level over years, not summarised as annual counts of bad-air days. And interventions aimed at the top of the distribution — filtration in homes, clean-air shelters, targeting of older adults with existing cardiopulmonary disease — would be judged on total dose averted rather than on how many advisory days they covered.

Neither of those follows from a single cohort study. Both are testable.

What to watch

Whether the cumulative-versus-episodic finding replicates in cohorts outside California and in younger populations, and whether air quality agencies begin publishing cumulative exposure metrics alongside daily indices. The second would be the visible sign that this line of research has landed.