The healthy worker effect Selection of workers and work forces

WILCOSKY T, WING S. The healthy worker effect: Selection of workers and work forces. Scand J Work Environ Health 13 (1987) 70-72. The favorable health status of employed populations in comparison to that of the general population is often called the "healthy worker effect," and most researchers feel that it results from the selection of healthy individuals during the hiring process. However, other mecha nisms may also contribute to the healthy worker effect. In particular, the healthy worker effect may re flect the selection, for epidemiologic study, of relativelyadvantaged workforces in terms of socioeconomic status and health. Evidence of differential health by socioeconomicstatus within the workforce is reviewed, and its possible role in the healthy worker effect is discussed.

Because epidemiologic st udies of employed population s typically find lower mortality rates than those of the general population, comparisons of workers with the general population may fail to detect excesses of occupationally related disease. The favorable health sta tus of employed populations is often called the " healthy worker effect" (9). Some researchers feel that the selection of healthy ind ividuals by employers during the hiring process is the major cause of this effect (I, 8). In addition persons in the general population who are too ill to be hired and keep a job may have a much higher mortality rate than do acti vely employed individuals, and the general population, includ ing both employed and unemployed individuals, therefore has relatively high mortality.
Differences between the mortality experience of specific groups of workers a nd the national population may, however, reflect important socioeconomic in fluences in addition to the selection of healthy persons for employment. For example, low socioeconomic status is associated with unemployment (10), as well as with poor health (3,5), and a selection process based on socioeconomic status rather than health at the time of hire could therefore still indirectly lead to health differences between the emplo yed and general populations. Low socioeconomic status could also contribute to the healthy worker effect through the traditional health selection mechanism by promoting poor health. As discussed later, another possible mechanism involving socioeconomic status emphasizes the heterogeneity of mortality rate s within the employed population.

Occupation, labormarket segmentation and econom ic sectors
Even in the absence of health selection at the time of hire, mortality variation within the work ing population could generate the relati vely low mortality rate s typically associated with the healthy worker effect. For example, the Wh itehall study (7) of British civil servants found the 1O-year mortality of workers in the lowest employment grade to be three times higher than that in the highest grade. If the rates in the general population were intermediate, workers in the highest employment grade would show a healthy worker effect, while tho se in the lowest grade would have excess mortality.
Evidence of differential health by occupation within the work force co mes from a variety of sources. In Great Britain and Scandinavia, for example, one study showed that male unskilled and semiskilled workers had a pronounced mortality excess (6), and the authors suggested that nonoccupational risk factor s associated with social class partiall y accounted for this excess. Disease risk factors like cigarette smoking also vary by occupation, such that smoking is more prevalent among blue-collar workers than among white-collar workers, a nd within industries smoking is associated with lower sta tus oc cupations (14). The possible impact of differential risk factor distributions on health ou tcome s acro ss occupations is illustrated by the high correlation (r = 0.72) between the standa rdized mortality ratio for lung cancer and the proportional ratio for cigarette smoking among the 25 occupational orders in Great Britain (13). Since high status workers tend to have relat ively low mortality, morbidity, and risk factor levels, a substantial part of the healthy worker effect in so me occupations may reflect socioeconomic differences between the workers and members of the general population.
Although variation in socioeconomic status by occupation can help explain some of the disease deficits and excesses observed among groups of workers, occupat ion by itself comprises only one of several aspects of emplo yment relate d to socioeconomic sta tus. Recent work describing the socioeconomic heterogeneity of workplaces and work forces (2,16) is especially relevant to the consideration of the healthy worker effect. This literature distinguishes economic sectors (the organization of firms and industries) from labormarket segmentation (the existence of distinct job markets within firms or ind ustries). Both concepts have potential importance for health-relevant working and living conditions. The economic center, or monopoly sector, tends to have big firms with large market shares and long range economic planning (among other characteristics) such that they are less affected by disruptions in economic cycles tha n are firms in the periphery, or compe titive sector, which are smaller, more labor intensive, and more subject to short-term disruptions . Within these sectors, primary labor markets with opportunities for promotions, job protection, and higher wages can be distinguished from secondary labor markets with more temporary employment, lower wages, and poorer work conditions.
The literature on economic sectors and labormarket segmentation provides an important guide for considering the generally favorable mortality experience of occupational cohorts. Specifically, most occupational cohort studies focus on the most favored workers in terms of both labormarket and economic sector dimensions. For example, such studies typically exclude employees who work for relatively brief periods in a given company . Nicholson (12) points out that virtually every industry includes a two-component employment pattern with a large number of transient workers and a smaller number of long-term employees. The health experience of transient workers may be worse than that of workers with stable jobs due to occupational exposures and /or living conditions; therefore epidemiologic studies of the latter group could give a misleading overview of the association between employment and health.
With regard to economic sectors and the healt hy worker effect, most occupational cohort studies are concentrated in the big firms that characterize the economic center. Although workers from small companies could be aggregated into a single large study population, the operational difficu lties of such studies are considerably greater than those for a study of a single large company with centralized records. Table I, which presents estimates from the Natio nal Occupational Hazard Survey (I I) of urban workplaces in the United States , illustrates the potential for divergent results between studies of large and small companies from different economic sectors. Workers in plants with more than 500 employees are four times more likely to have preemployment physical examinations than plants that employ fewer than 250 workers, and therefore health • These co unts are t he deno mi nato rs for the percentages that follow.
selection is more likely in large plants. Once hired, workers in large plants also have much greater access to services that protect their health. Table I shows that plants with 500 or fewer workers account for over twothirds of the employees in urban workplaces, and the health stat us of the general population is therefore heavily weighted by the healt h experience of emplo yees in smaller workplaces. To a large extent, the healthy worker effect may reflect the selection of work forces for study rather than selection of individuals into the work force .

Discussion
Although selection for good health at the time of hire may account for only part of the healthy worker effect, this explanation tends to eclipse the possib le importance of socioeconomic heterogeneity within the working popu lation. Comparisons of workers with the general population are generally based on the assumption that disease excesses among the workers result from occupational hazards and disease deficits reflect health selection, while potential confounding from socioeconomic differences receives little attention. As illustrated by the excess mortality of low-skilled workers in Great Britain and Scandi navia (6), the healthy worker effect observed in other studies may be just one end of a spectrum of biased comparisons. An understanding of the healthy worker effect has implications for nonoccupational as well as occupational researchers. Participants in genetic family studies (15), community-based cohort studies (17), and other studies of selected individuals (4) may differ socioeconomically from more heterogeneous reference populations, and their mortality will be relatively low or high. Since most measures of health and disease show a socioeconomic gradient, researchers should expect a disparity in such outcomes when relatively homogeneous groups are compared to the general population.