Methods development for epidemiologic investigations of the health effects of prolonged ozone exposure. Part III. An approach to retrospective estimation of lifetime ozone exposure using a questionnaire and ambient monitoring data (U.S. sites).

Methods are needed for retrospective estimation of long-term ozone exposures in epidemiologic studies. The overall objective of this study was to evaluate whether data from available U.S. ozone monitoring sites are useful for estimating lifetime ozone exposures of young adults (for example, college students). Several aspects of this question were evaluated. First, we applied and (compared several spatial interpolation methods to a set of long-term average ozone data from all U.S. monitoring sites in operation from 1981 through 1990. Interpolation methods included simple and weighted averages, linear regression, and, in an exploratory way, kriging. The comparison of methods was carried out for five different metrics of ozone concentration: the daily one-hour maximum (MAX1) and eight-hour maximum (MAX8), the average ozone concentrations between 10 a.m. and 6 p.m. (MID8) and between 10 a.m. and 10 p.m. (MID12), and the sum of all hourly ozone concentrations greater than or equal to 60 parts per billion (ppb) (SUM06). We also tested whether interpolations were improved by modeling the influence of covariates such as population density, elevation, and weather on ozone concentrations. We analyzed the reliability of a set of newly developed questions about past activity levels among a group of 52 freshmen students at Yale University. This was done by analyzing the agreement between answers to the same questionnaire administered two times, one month apart (test and retest), to the same students. Finally, we combined the interpolation models with residential history information obtained by questionnaire to derive long-term ozone exposure estimates for a group of 200 Yale freshmen. Results of our study showed that the density of available monitoring sites appears to be adequate for estimating spatial patterns of long-term average ambient ozone concentrations. A simple regression-based interpolation on the three nearest sites produced consistently good results. Including covariates in the interpolation models did not substantially improve the estimates. The largest estimation errors occurred for areas where ozone concentrations were highest. The newly developed activity history questions exhibited fair to moderate reliability, The results of this work imply that reasonably precise estimates of long-term ambient ozone concentrations for use in large-scale epidemiologic studies can be achieved by interpolating ozone concentrations between available U.S. monitoring sites. This study did not address the issues of whether and how retrospective data on factors that modify exposure or dose (e.g., indoor/outdoor penetration of ozone and time outdoors) can be used to derive estimates of long-term personal ozone exposures and contribute to the assessment of received dose.