Differential distributed lag patterns of source-specific particulate matter on respiratory emergency hospitalizations.

While different emission sources and formation processes generate mixtures of particulate matter (PM) with different physicochemical compositions that may differentially affect PM toxicity, evidence of associations between PM sources and respiratory events is scarce. We estimated PM10 sources contributed from 19 chemical constituents by positive matrix factorization, and examined association of short-term sources exposure with emergency respiratory hospitalizations using generalized additive models for single- and distributed lag periods. PM10 contributions from eight sources were identified. Respiratory risks over a consecutive 6-day exposure period were the highest for vehicle exhaust [2.01%; 95% confidence interval (CI): 1.04, 2.99], followed by secondary sulfate (1.59%; 95% CI: 0.82, 2.37). Vehicle exhaust, regional combustion, and secondary nitrate were significantly associated with 0.93%-2.04% increase in respiratory hospitalizations at cumulative lag2-5; significant associations of aged sea salt (1.2%; 95% CI: 0.63, 1.78) and soil/road dust (0.42%; 95% CI: 0.03, 0.82) were at lag0-1. Some effect estimates were no longer significant in two-pollutant models adjusting for PM10; however, a similar temporal pattern of associations remains. Differential lag associations of respiratory hospitalizations with PM10 sources were indicated, which may reflect the different particle size fractions that sources tend to emit. Findings may have potential biological and policy implications.