Abstract PR07: Latent class analysis of multipollutant exposure

Abstract

Abstract Recent work suggests that air pollutants, toxic chemicals, and other environmental exposures negatively impact health in a synergistic way. In order to measure and understand this impact, we propose a latent class mixture model of multipollutant exposure. Our model incorporates the joint behavior of individual pollutants that inform overall exposure levels and provides clear measures of multipollutant exposure. We identify low, medium, and high levels of joint exposure to volatile organic compounds (VOC), particulate matter (PM), and heavy metals (HM), and present estimated class-membership for each census tract in the US (>76,000). We use publicly available data from the EPA, including the National Air Toxics Assessment (NATA), measured at the census tract level. Model results indicate that mixtures of 5 levels of VOC, 2 levels of PM, and 4 levels of HM best fit data at the national level. Twelve percent of U.S. census tracts are in the highest levels of all 3 categories of pollutants. While frequently observed in metropolitan areas, rural areas are sometimes detected at the highest levels, as well. Our next focus is to incorporate spatial correlations as well as linkages to our earlier latent model work on socioeconomic status (SES). Taken together, these extensions can then be incorporated into a holistic, exposome modeling framework for estimating disparities in cancer survival. This abstract is also being presented as Poster A31. Citation Format: Alexandra Larsen, Viktoria Kolpacoff, Kara McCormack, Terry Hyslop. Latent class analysis of multipollutant exposure [abstract]. In: Proceedings of the AACR Special Conference on Environmental Carcinogenesis: Potential Pathway to Cancer Prevention; 2019 Jun 22-24; Charlotte, NC. Philadelphia (PA): AACR; Can Prev Res 2020;13(7 Suppl): Abstract nr PR07.