MODELLING INTRA-URBAN VARIABILITY IN TRAFFIC-RELATED AIR POLLUTION EXPOSURES: THE ACCESS PROJECT

Abstract

ISEE-306 Introduction: Significant spatial patterning has been observed for childhood asthma, with greater severity in lower-income urban communities where multiple risk factors are clustered. The Asthma Center on Community, Environment, and Social Stress (ACCESS) project is a multifactorial birth cohort study aimed at disentangling this spatial patterning by distinguishing effects of environmental exposures, social stressors, and genetics in the etiology of childhood asthma. Clearly, this effort requires exposure data that captures significant intra-urban variability. Aim: To estimate traffic-related air pollution exposures, we are developing a GIS-based regression model that allows for extrapolation from air pollution measures at 43 homes across the Boston urban area to a cohort of 1000. The method should reduce exposure misclassification for epidemiological analysis by focusing on intra-urban variability, incorporating spatial and temporal variability, and using site characteristics and other measures associated with long-term exposure variability. Methods: We measured indoor and outdoor concentrations of four traffic-related primary pollutants (fine particles (PM 2.5), nitrogen dioxide (NO2), elemental carbon (EC), and polycyclic aromatic hydrocarbons (PAHs)) at 43 homes across Boston, for one week per home during heating and cooling seasons in 2003 and 2005. We measured temperature and humidity indoors and out, and diurnal traffic counts outside the home. Sample homes were selected to represent a range of traffic densities, using Massachusetts Highway Department (MHD) data and the Spatial Analyst Extension of ArcGIS 9.0. We will perform GIS-based multivariate regression analyses using publicly-available data on traffic volume and composition, ambient monitoring data, land use, meteorology, and site characteristics to examine intra-urban variability in traffic-related exposures. Predictive covariates will be combined with two-dimensional thin-plate splines, capturing additional variability in a geoadditive model to produce city-wide ambient concentration estimates. Results: Preliminary analyses indicate significant heterogeneity in traffic density across sites, both within and between neighbourhoods, allowing evaluation of heterogeneity and key predictors at multiple spatial levels. Average PM2.5 concentrations were higher indoors and in summer; EC was slightly higher outdoors and in winter; NO2 was similar indoors and out, though I/O ratios varied across sites. Outdoors, PM2.5 and EC were significantly correlated in both seasons (p < 0.05), potentially displaying similar spatial patterns. EC and NO2 were marginally correlated; PM2.5 and NO2 were not. For all pollutants, outdoor concentrations span an order of magnitude, with both temporal and spatial heterogeneity. After controlling for site and roadway characteristics and regional pollution (ambient monitoring data), we expect model residuals to display significant spatial heterogeneity corresponding to local stationary sources, topographical features, or other site characteristics. Subsequent analyses will combine this ambient exposure model with home characteristics questionnaire data to address heteorogeneity in indoor exposures (and thus personal exposures) to traffic-related air pollutants.