Dirtier air days bring more deaths, and weather cannot explain it away

The London air pollution disaster of 1952 settled one question: very high levels of particulate smog can cause dramatic spikes in daily deaths. The harder question was whether the far lower particle levels of ordinary city air still carry a mortality cost. More than a dozen studies at much lower concentrations had reported small but nonzero increases in daily deaths, yet sceptics could argue that weather, season or other confounders were driving the pattern. This analysis pooled those studies to test how consistent the association really was, and whether confounding could plausibly explain it.

The author conducted a pooled analysis of the studies linking airborne particle concentrations to daily mortality at low pollution levels. A primary focus was comparing effect sizes across locations with large differences in potential confounding factors, and in how strongly those factors correlated with particle concentrations. The analysis examined cities with different pollution seasonality, temperatures and humidity. Data from Philadelphia were examined in detail, including a very flexible nonlinear adjustment for weather, to test whether the standard controls for season and weather were adequate.

In the primary analysis, airborne particle concentration was a significant risk factor for elevated mortality, with a relative risk of 1.06 per 100 micrograms per cubic metre rise in total suspended particles, and a tight range of 1.05 to 1.07. The relative risk was 1.06 even when restricted to cities where pollution peaked in summer, and it was identical in warmer versus colder cities and in drier versus more humid climates. Mortality peaked in cold months everywhere while particles usually peaked in warm months, arguing against seasonal patterns as the explanation. The detailed Philadelphia analysis showed weather control was adequate and that flexible weather modelling did not disturb the association.

The striking feature is consistency: the same effect size appeared across climates, seasons and confounder structures, which is hard to reconcile with weather or season driving the result. The author concludes that a causal association is the most reasonable interpretation, a reading supported by other studies linking particulate pollution to lung function deficits, increased symptoms and hospitalisations. This analysis dates from 1994 and uses total suspended particles as its measure, so how the exact figures map onto modern particle metrics is an open question, though the broader field has since moved in the same direction.

The practical knowledge here is that air pollution's toll is not confined to historic disasters: ordinary day-to-day swings in particle levels tracked with daily death counts across many cities. That makes local air quality worth thinking of as a real health exposure alongside diet, activity and sleep, rather than background noise. The study cannot tell any individual what a given pollution day does to their personal risk, but it establishes that at a population level, cleaner air days are quieter days for mortality.