Associations Between Source-Resolved Particulate Matter and Cardiorespiratory Emergency Department Visits

P-059 Background: Recent particulate matter (PM) health effects research has focused on identifying the specific components and sources of aerosol that pose the greatest health risks. Few epidemiologic studies, however, have included source apportionment estimates in their examinations of PM health effects. This is due, in part, to uncertainties regarding optimal methods for conducting PM source apportionment, as well as the lack of suitable air quality and health effects data for analysis. To address these issues, we analyzed an extensive time-series of chemically-resolved PM measurements and conducted a source-specific epidemiologic analysis using data from three source apportionment methods. Methods: Data for this analysis included daily speciated PM concentrations and emergency department (ED) visits collected between November 1998 and December 2002. PM source apportionment was conducted using: a) an extended chemical mass balance (CMB) approach; b) positive matrix factorization (PMF) with temperature-resolved carbon fractions; and c) single-specie tracers. CMB and PMF analyses quantified impacts from 7 and 11 sources/factors, respectively. Visits for 8 cardiovascular and 5 respiratory outcomes were assessed as endpoints. Relative risks (RRs) associated with 3-day moving averages of PM were estimated using Poisson generalized linear models. Results: In total, 544 source-specific models for ED visits were examined. Initial findings showed that slope parameter estimates from CMB and PMF were similar for the 78 individual models that shared source categories identified by both methods (Pearson's correlation coefficient, R = 0.80). Slightly weaker associations existed between the slope parameter estimates generated using the inorganic tracers and those from both CMB (R = 0.69) and PMF (R = 0.74). Results showed specific sources/factors to be related to differing outcome groups. Cardiovascular ED visits were most strongly associated with PM from diesel sources; respiratory visits with woodsmoke; and several cardiorespiratory outcomes with secondary organic aerosol. These results were typically robust to the source apportionment method used. Conclusions: Despite methodological differences among the source apportionment techniques, there was generally good inter-method agreement among the method-specific relative risk estimates. These findings suggest that it is possible in this location to obtain robust, source-resolved epidemiologic results using several apportionment methods. Correspondingly, agreement among the methods provides support for the interpretation that the health effects may indeed be associated with the identified sources/factors. Finally, the findings showing specific PM sources associated with different outcomes is consistent with the emerging evidence concerning the multiple etiologies of PM health effects and the varying role of specific chemical components.