Some confounding factors in the study of mortality and occupational exposures

With the recent interest in the study of occupational exposures, the impact of certain selective biases in the groups studied is a matter of some concern. In this paper, data from the Hanford nuclear facility population (southeastern Washington State, 1947-1976), which includes many radiation workers, are used to illustrate a method for examining the effect on mortality of such potentially confouding variables as calendar year, length of time since entering the industry, employment status, length of employment, job category, and initial employment year. The analysis, which is based on the Mantel-Haenszel procedure as adapted for a prospective study, differs from most previous studies of occupational variables which have relied primarily on comparing standardized mortality ratios (utilizing an external control) for various subgroups of the population. Results of this analysis confirm other studies in that reduced death rates are observed for early years of follow-up job category). In addition, workers employed less than two years and especially terminated workers are found to have elevated death rates as compared with the remainder of the study population. It is important that such correlations be taken into account in planning and interpreting analyses of the effects of occupational exposure.