Beyond Mean Exposure: How Exposure Variance Shapes Air Pollution Mortality Risk

Ambient air pollution contributes substantially to premature mortality globally, and the effectiveness of policies addressing it depends critically on the shape of the dose-response relationship, which remains poorly understood.Linking hourly air pollution data with daily vital statistics from 251 Chinese cities over four years, we show this relationship depends not only on the mean, but also on the variance of within-period exposure.At a 7-day mean of 150 μg/m³, predicted mortality varies more across observed hourly PM2.5 distributions than across mean exposure differences of 100 μg/m³.The sign of this variance effect also depends on the mean: at low mean concentrations (a convex region), higher variance is associated with modestly higher mortality.At higher concentrations (a concave region), greater variance is associated with lower mortality -implying constant exposure is more harmful than fluctuating exposure for a given mean.These findings reveal important relationships masked using temporally aggregated air pollution data.