We developed a hybrid chemical transport model and receptor model (CTM-RM) to conduct source apportionment of both primary and secondary PM2.5 (
particulate matter ≤2.5 μm in diameter) at 36 km resolution throughout the U.S. State of Georgia for the years 2005 and 2007. This novel source apportionment model enabled us to estimate and compare associations of short-term changes in 12 PM2.5 source concentrations (agriculture, biogenic,
coal, dust,
fuel oil, metals,
natural gas, non-road mobile diesel, non-road mobile
gasoline, on-road mobile diesel, on-road mobile
gasoline, and all other sources) with emergency department (ED) visits for pediatric
respiratory diseases. ED visits for
asthma (N = 49,651),
pneumonia (N = 25,558), and acute
upper respiratory infections (acute URI, N = 235,343) among patients aged ≤18 years were obtained from patient claims records. Using a case-crossover study, we estimated odds ratios per interquartile range (IQR) increase for 3-day moving average PM2.5 source concentrations using conditional logistic regression, matching on day-of-week, month, and year, and adjusting for average temperature, humidity, and holidays. We fit both single-source and multi-source models. We observed positive associations between several PM2.5 sources and ED visits for
asthma,
pneumonia, and acute URI. For example, for
asthma, per IQR increase in the source contribution in the single-source model, odds ratios were 1.022 (95% CI: 1.013, 1.031) for dust; 1.050 (95% CI: 1.036, 1.063) for metals, and 1.091 (95% CI: 1.064, 1.119) for
natural gas. These sources comprised 5.7%, 2.2%, and 6.3% of total PM2.5 mass, respectively. PM2.5 from metals and
natural gas were positively associated with all three respiratory outcomes. In addition, non-road mobile diesel was positively associated with
pneumonia and acute URI.