Controlling the healthy worker effect in the presence of an intermediate variable

Abstract

The healthy worker effect is an important issue in occupational epidemiology. We propose a new statistical method to test the relationship between exposure and time to death in the presence of the healthy worker effect. The basic idea of the proposed method reflects length bias sampling caused by job changes. According to simulation studies, both the proposed test and a test based on the Cox model including job changes as a time-dependent covariate seem to be satisfactory at the upper 5% significance level. However, the Cox models involving job changes as a time-independent covariate are unsatisfactory. The proposed test is superior in power to the test based on the Cox model with time-dependent employment status.