Healthy Worker Effect

Epidemiologic studies of occupationally exposed cohorts are useful in identifying and quantifying environmental risks because workers are generally exposed to higher levels of toxic materials than the general population. Although higher exposures make it easier to detect and characterize modest elevations in risk, such studies are plagued by a particularly pervasive form of selection bias, referred to as the healthy worker effect (HWE). The HWE is reflected in better health status of exposed workers relative to the general population. This bias arises from two sources: (i) workers were initially healthy enough to be hired, whereas the general population includes persons unfit for work, and (ii) workers who stay healthy (survivors) remain exposed, whereas less healthy coworkers reduce exposure via time off work, job transfer, or early termination. Unlike standard methods, G-estimation has been shown to reduce healthy worker survivor bias. Keywords: selection bias; G-estimation; occupational health; statistical methods; time-varying confounders; feedback