City air pollution spikes track with spikes in daily deaths

Air pollution is usually discussed as a slow, cumulative threat, something that shaves years off life over decades of exposure. A different and in some ways more unsettling question is whether day-to-day swings in pollution show up in day-to-day death counts. Time-series studies, which compare a city's daily mortality against its daily pollutant levels, had accumulated around the world, but individual studies varied in which pollutants they examined and what they found. This synthesis set out to pool that evidence systematically, separating the effects of particles from gases and examining how cause of death, age and season shaped the picture.

The authors conducted a comprehensive, systematic synthesis of daily time-series studies of air pollution and mortality from around the world, extracting effect estimates from 109 studies. Estimates were taken from both single-pollutant and multipollutant models, and broken down by cause of death, age and season. Pooled estimates were produced using random effects methods, with the increment for each pollutant set to the mean concentration among a representative group of cities. The team also examined how much the variability of pollutant concentrations explained differences between studies, and tested whether the results held under alternative approaches to selecting estimates.

In single-pollutant models, excess all-cause mortality was 2.0% per 31.3 micrograms per cubic metre of PM10 particles, 1.7% per 1.1 ppm of carbon monoxide, 2.8% per 24.0 ppb of nitrogen dioxide, 1.6% per 31.2 ppb of ozone and 0.9% per 9.4 ppb of sulphur dioxide. When several pollutants were modelled together the effect sizes generally fell, but remained significantly above zero for PM10 and sulphur dioxide. Respiratory mortality showed larger effects than all-cause mortality for every pollutant except ozone. Differences between studies were partly explained by how variable pollutant concentrations were, and the findings were robust to alternative analytic choices.

The authors' own conclusion is blunt: the synthesis leaves little doubt that acute air pollution exposure is a significant contributor to mortality. The persistence of the particulate matter signal in multipollutant models is particularly notable, since it suggests particles are not merely standing in for some other pollutant. The work also sharpened the picture of who and what is most affected, with respiratory deaths responding more strongly than deaths overall. As a pooled analysis of city-level patterns, it cannot assign cause for any individual death, and the mixture and levels of pollutants vary between places and eras.

The practical knowledge here is that air pollution operates on short timescales as well as long ones: high-pollution days were followed by measurably more deaths in these cities, especially from respiratory causes. That reframes air quality from an abstract lifetime statistic into something with a daily rhythm. For a general reader it is a reason to treat local air quality information as meaningful health information, in the same category as weather, rather than as background noise.