You are here:
Regional and temporal trends in exposure to ambient air pollution: Findings from the Strong Heart Study
Citation:
Weaver, A., J. Brooks, M. Li, B. Zou, J. Reese, K. Malloy, Y. Zhang, J. Yracheta, A. Navas-Acien, C. Ward-Caviness, G. Currin, N. Franceschini, AND G. Corbie-Smith. Regional and temporal trends in exposure to ambient air pollution: Findings from the Strong Heart Study. International Society of Environmental Epidemiology, New York, NY, August 23 - 26, 2021.
Impact/Purpose:
PM2.5 air pollution exposure is increasingly recognized as a risk factor for cardiovascular disease (CVD), a leading cause of death among American Indians (AIs). However, PM2.5 estimates for AIs remain largely unknown due to sparse monitoring. We describe ambient PM2.5 concentrations and regional/temporal trends in the Strong Heart Study (SHS) communities inArizona [AZ], Oklahoma [OK], and North and South Dakota [ND and SD] from 2001 to 2009 using an ensemble model that combined neural network, random forest, and gradient boosting algorithms. The median 30-day ambient PM2.5 in the different study sites were as follows: 9.2 μg/m3 (IQR: 2.1) in AZ, 9.1 μg/m3 (IQR: 2.4) in OK, and 5.5 μg/m3 (IQR: 2.8) in ND and SD. We did not detect a seasonal effect on PM2.5 exposure in AZ; however, PM2.5 in the wintertime was slightly higher in OK and substantially higher in ND and SD.
Description:
Background and Aim: Fine particulate matter (PM2.5) pollution exposure is increasingly recognized as a risk factor for cardiovascular disease (CVD), a leading cause of death among American Indians (AIs). PM2.5 estimates for AIs, however, remain largely unknown due to sparse monitoring. We describe ambient PM2.5 concentrations and regional/temporal trends in the Strong Heart Study (SHS) communities (Arizona [AZ], Oklahoma [OK], and North and South Dakota [ND and SD]) from 2001 to 2009. Methods: We used SHS phase IV (2001-2003, n=2,769) and phase V (2006-2009, n=2,478) data. We estimated ambient PM2.5 exposure using an ensemble model of daily mean PM2.5 concentrations at 1x1 km resolution, combining PM2.5 estimates from neural network, random forest, and gradient boosting algorithms. Participant exposure was estimated at the ZIP code level and averaged over 30 days. We performed descriptive analyses to depict how population-based concentrations of PM2.5 vary over the three SHS regions and across the two phases. Results: There were significant differences in 30-day median PM2.5 exposure between study sites. The median 30-day ambient PM2.5 in the different study sites were as follows: 9.2 μg/m3 (IQR: 2.1) in AZ, 9.1 μg/m3 (IQR: 2.4) in OK, and 5.5 μg/m3 (IQR: 2.8) in ND and SD. We did not detect a seasonal effect on PM2.5 exposure in AZ; however, PM2.5 in the wintertime was slightly higher in OK and substantially higher in ND and SD. These trends hold over both visit periods and all study sites. Annual median PM2.5 remained stable between phases IV and V for the OK and ND and SD regions; in AZ, PM2.5 first declined in phase V relative to phase IV but began to increase in 2008-2009. Conclusion: Understanding regional and temporal trends in ambient PM2.5 concentrations, which vary across tribal populations, may help to understand environmental exposures and CVD in AIs. This abstract does not reflect EPA policy.