A study of health effect estimates using competing methods to model personal exposures to ambient PM2.5

Various methods have been developed recently to estimate personal exposures to ambient particulate matter less than 2.5 microm in diameter (PM2.5) using fixed outdoor monitors as well as personal exposure monitors. One class of estimators involves extrapolating values using ambient-source components of PM2.5, such as sulfate and iron. A key step in extrapolating these values is to correct for differences in infiltration characteristics of the component used in extrapolation (such as sulfate within PM2.5) and PM2.5. When this is not done, resulting health effect estimates will be biased. Another class of approaches involves factor analysis methods such as positive matrix factorization (PMF). Using either an extrapolation or a factor analysis method in conjunction with regression calibration allows one to estimate the direct effects of ambient PM2.5 on health, eliminating bias caused by using fixed outdoor monitors and estimated personal ambient PM2.5 concentrations. Several forms of the extrapolation method are defined, including some new ones. Health effect estimates that result from the use of these methods are compared with those from an expanded PMF analysis using data collected from a health study of asthmatic children conducted in Denver, Colorado. Examining differences in health effect estimates among the various methods using a measure of lung function (forced expiratory volume in 1 s) as the health indicator demonstrated the importance of the correction factor(s) in the extrapolation methods and that PMF yielded results comparable with the extrapolation methods that incorporated correction factors.