Estimating risk under varying models of occupational exposure.
Sullivan-PA; Eisen-E; Kriebel-D; Woskie-S; Odencrantz-J
Occup Hyg 1996 Jan; 3(1-3):185-190
This study was designed to assess how changes in cut points for categorizing exposure variables in case/control studies affect the estimates of risk and to identify structural features of data which determine the extent of the effect on risk estimates resulting from changing categorical cut points. Matched case/control data sets were generated with a fixed data structure characterized by a known distribution of exposure among the controls and a known underlying distribution of the odds ratio. Preliminary analysis indicated that the average mean square error increased as the prevalence of exposure in the comparison group increased. The authors note that the choice of cut points is less likely to bias the results as the strength of the association decreases. The authors conclude that the variation in the odds ratio estimate may result from varying the choice of cut points between categorical exposure groups. The authors suggest that when analyzing case/control study data, bias could be minimized by categorizing exposure to achieve an equal distribution of exposed cases across the exposure groups.
NIOSH-Author; Epidemiology; Risk-factors; Risk-analysis; Analytical-models; Worker-health; Occupational-exposure; Simulation-methods; Statistical-analysis