Estimation of radiation doses for workers without monitoring data for retrospective epidemiologic studies.

Dosimetry data are required for many epidemiologic studies of radiation workers. For several reasons, these data may not be available for some workers for some periods of time. An estimate of a worker's dose for such a time period can be made using data from nearby time periods for the worker or using data obtained for other workers during the same time period. This paper reports the results of an evaluation of a procedure for estimating external doses for workers included in retrospective epidemiologic studies. This evaluation was conducted using data for workers who had recorded doses for each year in a 5-y period. The dose for the middle year of the 5-y period was treated as unavailable and the dose was estimated several ways. These estimates were compared with the recorded dose. It was observed that averaging a worker's dose data for the years preceding and following a year for which data hypothetically were not available most often yielded the best estimate of the dose. At one facility studied, the use of nearby data for the same worker was a more reliable estimate for a dose than was the department or facility median or mean dose. However, at another facility, use of the department median or mean dose produced a more accurate estimate than did nearby data for the worker. These results indicate that a single procedure for estimating doses is not best for all facilities, and it is recommended that a test of the type reported in this paper be performed before a decision is made on how to estimate doses. The use of department medians or means to estimate doses consistently produced better estimates than were obtained using facility medians or means.