Modeling Individual Daily Exposure Estimates for PAHs for an Asthma Epidemiology Study

ELIZABETH M. NOTH (1), S. Katharine Hammond (1), Gregory S. Biging (2), Ira B. Tager (1)

(1) Environmental Health Sciences, School of Public Health, University of California, Berkeley (2) Environmental Science, Policy and Management, College of Natural Resources, University of California, Berkeley

     Abstract Number: 357
     Last modified: November 9, 2009

Abstract
Exposure assessment for large-scale, panel-based epidemiology studies can be challenging even in a data-rich study. This research highlights the importance of both spatial and temporal aspects of exposure assessment for two pollutants from the class of polycyclic aromatic hydrocarbons (PAHs), which are often used as a surrogate for diesel exhaust but have other sources as well. Using data collected in the Fresno Asthmatic Children’s Environment Study (FACES), land use regression models for phenanthrene (PHE) and the sum of PAHs with 4-, 5- or 6-rings (PAH456) were built using mixed modeling regression to incorporate both temporal and spatial co-variates.

PAH data were collected daily by the PAS2000 monitor at the US EPA Supersite in Fresno, CA, from November 2000 through February 2007. The PAS2000 provides realtime measurements of total particle-bound PAH. From 2/2002-2/2003, intensive air pollution sampling took place at 83 homes including five to ten 24-hr PAH filter measurements per home. However, since the measurements were not made contemporaneously, the between- and within-in home variability had to be accounted for using mixed modeling. The filter concentrations at participant homes were the dependent variable for PHE and PAH456; the candidate covariates in the model included the particle-bound PAH concentrations at the US EPA Supersite, meteorological data, source data (traffic and land use), and other temporal and spatial variables (agricultural burning, season, etc).

The PHE model calculated 529,884 individual daily estimates of outdoor PHE concentrations for the residential locations for all FACES participants from individual entry into the study through 2/28/2007. The model contains 10 covariates – 24-hour average PAH at US EPA Supersite, total number of cars per capita within census blockgroup, proximity to major collector roads, location of home in relation to US EPA Supersite, % of homes in blockgroup using gas heat, three wind direction variables, number of agricultural burns within 5 miles on day of sampling, and a seasonal variable. The ratio of the 90th :10th percentiles for average individual temporal variability from entry to study through 2/2007 for estimated PHE was 2.5, and the ratio of the 90th :10th percentiles of the average daily spatial range was 2.

The model for daily, outdoor PAH456 concentrations calculated 532,135 individual estimates. The model had seven covariates - 24-hour average PAH at US EPA Supersite, total number of cars per capita within census blockgroup, % of homes in blockgroup using gas heat, wind direction, 24-hour relative humidity (%) at a fixed site in Clovis/Fresno, and total length of minor collector roads within 400m of the home. The ratio of the 90th :10th percentiles for average individual temporal variability from entry to study through 2/2007 for estimated PAH456 was 3.4, and the ratio of the 90th :10th percentiles of the average daily spatial range was 1.5.

We found that PAH concentrations within Fresno fluctuate significantly both over time and space. This indicates that it is necessary to account for the variability in both time and space when modeling the class of PAHs.