Effects of Selection and Confounding on Mortality in an Occupational Cohort (Dow).

Abstract

This research examined the mortality experience of 8,181 men employed by The Dow Chemical Company as of March 1, 1954. A prior investigation had described mortality within this cohort for the years 1954-1972. The present study extended the observation period through 1978, examined the effects of additional occupational and environmental factors on mortality, and compared the mortality experience of the Dow cohort with that of a community-based working population not previously available. Defining the cohort at a specific point in time led to selective inclusion of long-term employees, 88% having total company service of at least 15 years. Overall and cause-specific mortality in the study cohort compared favorably with mortality in both the United States population and the community-based cohort of employed men. The effects of initial selection of an active working population on subsequent mortality were estimated indirectly by examining age at entry, interval since entry, and employment status relative to mortality outcome. From these analyses, an initial favorable selection effect was evident for non-trauma causes of death. Analysis by socioeconomic status and general job category at entry identified a number of associations, such as an increase in death rates with lower socioeconomic status that was most apparent for digestive system disease. Analyses based on cigarette smoking data obtained between 1968 and 1976 revealed cause-specific mortality patterns for cigarette smokers which were consistent with those reported for other populations. Because cigarette smoking varied with socioeconomic status and job category, confounding among these variables was apparent for certain causes of death. This research identified several important factors which need to be addressed in designing occupational mortality studies, the most noteworthy being cigarette smoking and selection for employment. In addition, an analytical strategy was described for assessing potential occupational hazards, while controlling for the influence of extraneous factors. Nevertheless, the advantages of this strategy over the traditional SMR approach will be more theoretical than practical in the small cohorts usually seen in the chemical industry.