Relationships between exposure to air pollution and demographic attributes exhibit between-city heterogeneities
Abstract
Differentials in exposure to air pollution across population subgroups is a source of environmental health inequality. An important question is the generalisability of the trends between air pollution exposure and demographic attributes across different populations: comparisons are often confounded by different study methodologies. A key challenge is that individual-level demographic data are not usually available. In this study, we apply identical methods to a single cohort of 56 734 individuals to compare trends between exposures to ambient NO2, PM2.5 and O3 and sex, age, ethnicity, and socioeconomic status (SES) in the cities of Glasgow and Edinburgh, UK. We also use individual-level information on place of work or study (where that applies) to derive more realistic estimates of exposures than studies using exposures at place of residence only. We find that whilst some broad trends in exposure-demographic relationships are similar between Glasgow and Edinburgh—for example that younger adults, those of lower SES and of non-White ethnicity are, on average, exposed to higher concentrations of NO2 and PM2.5 (and to lower concentrations of O3)—our results also confirm heterogeneities in these relationships between the two cities. Importantly, we show that this geographical specificity extends to between-city geographical distributions of places of work and study as well; for example, males and females have equal exposures to air pollutants in each city based on residence location only, but have different patterns of exposure in the two cities when place of work is included. The heterogeneities we observe presumably arise from the particular way in which the urban fabric of a given city develops over time. The policy message is that it is not appropriate to assume geographically transferable relationships between air pollution exposure and SES, ethnicity, age and sex, even when the population demographic profiles are broadly similar.