Negative bias in exposure-response trends in occupational studies: modeling the healthy worker survivor effect.
Steenland-K; Deddens-J; Salvan-A; Stayner-L
Am J Epidemiol 1996 Jan; 143(2):202-210
This study examined employment status as a confounder in studies which analyze trends between cumulative exposure and mortality. The study data were taken from 10 published mortality surveys in which no occupational risks were noted by the authors. The study populations included 89,000 workers and 1,980,000 person years. For simulated data 10,000 workers were considered over a 40 year follow up period. The empirical data indicated that there was an increasing percentage of active person years with duration of employment which accounted for the observed negative trend in mortality with duration. Active person years decreased with time since hire, resulting in an artificial positive trend in mortality with time since hire. The increased proportion of inactive person time with increasing time since hire for a cohort is the basis of the wearing off of the healthy worker effect as time increased since hire. Using the simulated data, in the null cases where cumulative exposure affects neither disease incidence nor leaving rates, employment status acts as a negative confounder of exposure response trends, and traditional adjustment eliminates this confounding. However, the authors noted, that when cumulative exposure affects disease incidence or rates of leaving, just adjusting for employment status will not be adequate to produce correct responses. The authors conclude that employment status is simultaneously a confounder and an intermediate variable.
NIOSH-Author; Epidemiology; Risk-factors; Risk-analysis; Cancer-rates; Worker-health; Mortality-surveys; Statistical-analysis;
Author Keywords: bias; cohort studies; healthy worker effect; occupations
Dr. Kyle Steenland, National Institute for Occupational Safety and Health (NIOSH), R-13, 4676 Columbia Parkway, Cincinnati, OH 45226
American Journal of Epidemiology