Background Air pollution may disproportionately affect low SES communities; however, assessments often occur without high spatiotemporal resolution. We developed highly resolved pollution surfaces to assess: 1) 2012-2019 trends, 2) disparities in exposure, and 3) associations with respiratory symptoms using digital short-acting beta-agonist (SABA) data.

Methods We developed daily land use regressions (LUR), generating 100m surfaces of nitrogen dioxide (NO2), fine particulate matter (PM2.5) and ozone (O3) in California (CA, US). Surfaces were overlaid with SABA in space-time (Propeller Health). Using a generalized linear mixed model (glmmTMB), we evaluated associations of 3 pollutants. We evaluated differences in exposure by census tract (CT) SES (CalEnviroScreen).

Results From 2012-2019, mean annual NO2 and PM2.5 decreased by 18.0% and 7.0%, respectively, but O3 increased by 2.1%. In 2012, the lowest SES CT demonstrated more exposure vs. highest SES, respectively: 13.1 vs 7.0 ppb for NO2, 12.0 vs 8.7 µg/m3 for PM2.5, and 34.3 vs 33.0 ppb for O3, and the trend maintained to 2019 (all p < 0.05). The glmmTMB (n=3,386 (mean age 37.8y, 69.6% female) identified positive associations of SABA with 3 pollutants (all p<0.001). Per 10 ppb, 10 ug m-3 and 30 ppb increase in NO2, PM2.5 and O3 was associated with a 2.5% (95% CI: 1.3-3.8), 9.2% (7.4-10.8) and 17.2% (12.7-21.9) increase in daily SABA puffs/person.

Conclusions Air pollution disparities persist despite improvements in CA?s air. NO2, PM2.5 and O3 were significantly associated with SABA, and a greater burden may exist in low SES areas. The resolution of the LUR and SABA data may support advances in epidemiological research.