Semiparametric Adjustment for Temporal Confounding of Acute Air Pollution Effects in Cohort Studies

• Activity Number: 530
• Type: Contributed
• Date/Time: Wednesday, August 4, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Adam A. Szpiro*+ and Lianne Sheppard and Sara D. Adar
• Companies: University of Washington and University of Washington and University of Michigan School of Public Health
• Keywords: Environmental Epidemiology ; Semiparametric Modeling ; Air Pollution ; Cohort Studies

Abstract

A challenge in studying acute health effects of air pollution is temporal confounding due to seasonal or meteorological trends. There is extensive literature addressing this problem for time series studies with balanced exposure and outcome measurements. The typical approach is to include semiparametric temporal terms in the regression model. This methodology can also be applied in a cohort study, but it may not be optimal due to the imbalance between exposure and outcome data. In many cohort studies there are exposure data for every day while health outcome data are only available on a more limited set of follow-up days. In addition, cohort outcome data typically include significant inter-subject variability. We develop approaches to utilizing the more extensive and less variable exposure data in order to improve efficiency of inference for the health effect parameter of interest.