AAAR 32nd Annual Conference
September 30 - October 4, 2013
Oregon Convention Center
Portland, Oregon, USA
Abstract View
Association of Respiratory and Circulatory Hospitalizations with PM$_(2.5) Elemental Carbon (EC), Organic Carbon (OC), and Gaseous Co-Pollutants in Pittsburgh, Pennsylvania, during 2001-2002
RICHARD BILONICK, Daniel Connell, Evelyn Talbott, Judith Rager, University of Pittsburgh
Abstract Number: 531 Working Group: Health Related Aerosols
Abstract Time series structural equation models (SEMs) were fitted separately to daily respiratory and circulatory hospitalization counts from Allegheny County, Pennsylvania, from August 2001 to August 2002. Models included two independent daily measurements of fine particulate EC and OC (made using a PM$_(2.5) speciation sampler and a collocated semi-continuous carbon analyzer), as well as daily measurements of gaseous co-pollutant concentrations (SO$_2, NO, NO$_2, CO, and O$_3) and mean daily temperature. All air quality measurements were made at a monitoring site located in Schenley Park in the City of Pittsburgh. Respiratory and circulatory hospitalization counts were adjusted for day-of-week effects and holidays. A calibration model encompassing the four carbon measurements was included in each SEM. Ordinarily, calibration with only two methods is not possible without making strong assumptions that cannot be verified. However, given the strong correlation between EC and OC, the use of two correlated congeneric measurement error models (two correlated latent factors - one for EC and one for OC, each with two indicators) allowed estimation of separate scale bias and method imprecision (random error) components for EC and OC. Results indicated that EC and OC had similar and statistically significant scale biases between the two measurement methods, but with substantially different intercepts. Hospitalizations were allowed to depend on latent EC and OC (0-4 day lags), gaseous co-pollutants (0-1 day lags), and temperature (both directly and indirectly through the dependency of latent EC and OC on temperature). In addition, hospitalizations, EC, OC, gaseous co-pollutants, and temperature were each allowed to depend on its previous day's value. In every case, autocorrelation with previous day's values was statistically significant as were temperature dependencies. Preliminary results indicated a differential effect between EC and OC on hospitalizations, although not statistically significant. Among air pollutants, only NO$_2 had a statistically significant association with respiratory hospitalization.