Assessing the Importance of Different Exposure Metrics and Time-Activity Data to Predict 24-H Personal PM 2.5 Exposures

Abstract

Personal PM 2.5 data from two recent exposure studies, the Scripted Activity Study and the Older Adults Study, were used to develop models predicting 24-h personal PM 2.5 exposures. Both studies were conducted concurrently in the summer of 1998 and the winter of 1999 in Baltimore, MD. In the Scripted Activity Study, 1-h personal PM 2.5 exposures were measured. Data were used to identify significant factors affecting personal exposures and to develop 1-h personal exposure models for five different micro-environments. By incorporating the time-activity diary data, these models were then combined to develop a time-weighted microenvironmental personal model (model M1AD) to predict the 24-h PM 2.5 exposures measured for individuals in the Older Adults Study. Twenty-four-hour time-weighted models were also developed using 1-h ambient PM 2.5 levels and time-activity data (model A1AD) or using 24-h ambient PM 2.5 levels and time-activity data (model A24AD). The performance of these three models was compared to that using 24-h ambient concentrations alone (model A24). Results showed that factors affecting 1-h personal PM 2.5 exposures included air conditioning status and the presence of environmental tobacco smoke (ETS) for indoor micro-environments, consistent with previous studies. ETS was identified as a significant contributor to measured 24-h personal PM 2.5 exposures. Staying in an ETS-exposed microenvironment for 1 h elevated 24-h personal PM 2.5 exposures by approximately 4 w g/m 3 on average. Cooking and washing activities were identified in the winter as significant contributors to 24-h personal exposures as well, increasing 24-h personal PM 2.5 exposures by about 4 and 5 w g/m 3 per hour of activity, respectively. The ability of 3 microenvironmental personal exposure models to estimate 24-h personal PM 2.5 exposures was generally comparable to and consistently greater than that of model A24. Results indicated that using time-activity data with 1-h exposure information, either as micro-environment-specific exposures (model M1AD) or as ambient concentrations (model A1AD), improves our ability to estimate 24-h personal PM 2.5 exposure over the model using 24-h averaged ambient levels alone (model A24). Model performance was higher in the summer than in the winter season. In addition, higher crude R 2 values were reported for subjects participating in both seasons, where the R 2 values equaled .53, .55, .46, and .38 for models M1AD, A1AD, A24AD, and A24, respectively. The low predictive ability of the microenvironmental exposure models in the winter might, in part, be attributed to the narrow dynamic range of personal PM 2.5 exposures.