Development of Air Quality Indicators for use in Epidemiologic Studies
JORGE PACHON (1), Ted Russell (1), Jim Mulholland (1), Siv Balachandran (1), Stefanie Sarnat (2), Jeremy A. Sarnat (2), Mitch Klein (2), Paige Tolbert (2)
(1) Georgia Institute of Technology (2) Emory University
Abstract Number: 304
Preference: Platform Presentation
Last modified: November 9, 2009
Working Group: sq1
Respiratory (RD) and cardiovascular disease (CVD) have been associated with pollutants from diverse emission sources [1,2,3]. In an effort to assist in epidemiologic studies linking health end points to sources, we are developing and assessing air quality indicators of varying complexity. The primary focus of this work is on indicators for mobile source impacts.
In order to identify the indicators that are most associated with health endpoints, we use factor analysis techniques, such as principal component analysis (PCA) and positive matrix factorization (PMF), to help identify groups of pollutants that are associated with health endpoints. In this case, we use not only particulate matter species concentrations, but also gaseous species and residuals from epidemiologic models. Such residuals are the difference between the predicted and observed emergency department (ED) counts for selected outcomes in epidemiologic models in which air pollution has not been included as a variable.
Air quality data for this project are collected from the Jefferson Street (JST) site, a mixed industrial-residential area in Atlanta, GA during the period 1999-2004. JST is part of the SEARCH network . After a data screening process, a total of 1,888 days within the six year period were considered valid for the study. For the epidemiologic models, data from 31 hospitals in the metro-Atlanta area have been collected from January 1993 to August 2000. A total of 4,407,535 ED visits from residents were recorded during this period and classified as RD or CVD.
Indicators for mobile sources that we are testing include single species, such as CO, NOx and elemental carbon, and combinations of species that are commonly associated with mobile sources, derived from the application of receptor models. We are also looking at the fraction of ozone associated with automobile emissions, developed from chemical transport modeling and sensitivity analysis . Additionally, we have also used regression modeling to help quantify the amount of primary organic and secondary organic aerosol, and the method provides information on the amount that is due to mobile sources . In addition to use in health studies, the indicators developed can also be used to help assess how source impacts on air quality have varied over time.
1.Peel, J. L. et al. Epidemiology 2005, 16, (2), 164-174.
2.Metzger, K. B. et al. Epidemiology 2004, 15, (1), 46-56.
3.Sarnat, J. A. et al. Environ. Health Perspect. 2008, 116, (4), 459-466.
4.Hansen, A. et al. Journal Air & Waste Manag. Assoc. 2006, 56: 1445-1458.
5.Cohan, D. S., et al. Environ. Sci. Technol. 2005, 39(17): 6739-6748.
6.Pachon, J., et al. AAAR 28th Annual Conference, Minneapolis, MN. 2009