THE FORGOTTEN SIDE OF THE EXPOSURE MODEL: USE OF AUTOBIOGRAPHICAL MEMORY IN RETRIEVAL OF PERSONAL MOBILITY PATTERNS Chris Peterson Brus, Burton C. Kross, and Charles F. Lynch University of Iowa Department of Preventive Medicine and Environmental Health Iowa City, IA ABSTRACT Precise residential radon exposure assessment is, by necessity, dependent upon the coupling of in-home radon measurements and personal mobility data. To evaluate the impact of collecting individual occupancy data rather than assigning a constant for percent time 'in-home', thirty study participants were queried using a standard decade table format. Over 50% of participants refused to attempt completion of the initial format. A second methodology was then devised in which autobiographical memory events, supplied by the participant, were used to create a contextual framework in which retrieval of mobility data could be accomplished. Ninety-two percent of those retested (n=26) found Method Two more acceptable, reporting a significant increase in accuracy (Wilcoxon Sign Rank Test, p c 0.01). Although percent time 'in-home' across the entire sample approximated the 70% standard routinely used in 222Rn exposure modeling (mean 73.7% * 12.4) the range of reported 'in-home' time (53.0 with a minimum and maximum of 41.6 and 94.6% respectively) was large enough to warrant assessing the impact of percent time variability on exposure classification. Time-weighted average exposure for years at latest residence (33.7 yrs  10.1) were calculated for each participant using reported hours spent 'in-home' and secondarily, incorporating reported hours spent 'outside' and 'in another building' to yield a 'total lime' exposure. The mean for 'in-home' exposure (3.4 pCi/1: 3.8) was then compared with that for 'total time' exposure (2.7 pCih 2.7) using a paired (-Test and found to be significantly different (p c 0.01). The rank order of individual data points did not remain stable after weighting exposure by individually reported percent time in the 'total time' exposure model. INTRODUCTION With the increase in epidemiologic research of chronic diseases with long latent periods, use of the case- control study design has become more prominent. This design relies on creating a two-sided exposure model: one side consisting of quantitative, mechanistically derived data; the other utilizing participant reported levels of the exposures and confounders of interest. Often this model relies heavily on the subjects' ability to accurately retrieve pertinent information over a time span of several decades. Differential memory retrieval then becomes a major impediment to correct categorization of study participants leading to substantial misclassification bias. It is often difficult to determine the diictionality of misclassification with any degree of assuredness, and although directional misclassification may be reported as a possible source of bias, differential memory retrieval is often accepted as non- directional, having only the effect of demonstrating a less significant relationship between exposure and disease. Precise residential radon exposure assessment is, by necessity, dependent upon the coupling of in-home radon measurements and personal mobility data. Advancements in radon measurement technology are driven by a need to minimize measurement error and enhance precision. Collection techniques and devices are closely monitored by an external regulatory agency to ensure that quality assurance/quality control guidelines arc being met. There are no such safeguards for (lie other side of the model. Although researchers often state that time spent in different areas of the home may affect their proposed model (1.2). it is standard practice to factor in a 'constant' for percent of occupancy. This 'constant' can vary substantially between studies. A review of present literature reveals use of a 'constant' occupancy factor anywhere from 65% to 80% (3,4,5.6,7), with Lccs, et al.(8) reporting a dichotoinous occupancy factor based on 'worker' (65%) vs. 'non-worker' status (85%). It would seem that a factor as integral to accurate radon dose assessment as mobility would be as important to retrieve on an individual basis as is data concerning smoking, occupational exposures and other confounders. This paper will compare two methodologies that were used to retrieve individual mobility pattern data and express die finding as they relate to methodologic acceptability, ease of retrieval, and participant reported accuracy. The 1993 International Radon Conference HIP 35