"Healthy-Worker Effect" Can Skew True Picture of Workers’ Health, UB Study Shows

BUFFALO, N.Y. -- Employees should be skeptical of any report boasting that their health as a group is better than that of the general population, an occupational epidemiologist at the University at Buffalo warns.

Such a comparison always will make the group and the company look good, said Gregg S. Wilkinson, Ph.D., a professor in the UB Department of Social and Preventive Medicine, because of an innate bias called the "healthy worker effect."

Wilkinson's study of female nuclear-weapons workers, presented recently at the annual meeting of the Society for Epidemiologic Research, illustrated a robust example of the healthy-worker effect, well-known among male workers but studied far less frequently among female workers.

"The healthy-worker effect exerts its influence when fewer deaths are observed for workers in an industry, compared to the U.S. population," Wilkinson said. "The understanding is that people must have at least a minimum level of health to hold down a job, whereas the general population includes everyone -- healthy and sick."

Wilkinson compared death statistics for female workers at 12 nuclear-weapons sites with the number of deaths expected to occur in the female population of the United States. The study covered a total of 67,976 women who had worked at any of the sites before Jan. 1, 1980. It was one of the largest analyses to date of female workers.

Results showed that at all of the sites, the number of deaths was either similar to or lower than that of the female population at large.

"Industries tend to use such findings to show how well they're doing," Wilkinson said. "But if you are going to get the right answers, you have to use appropriate comparisons. An appropriate comparison would be workers who are similar: either unexposed workers in the same plant, or workers who are similar but work in another plant where the exposure of interest does not exist."

However, the healthy-worker effect also can skew results of health studies involving workers within the same factory. Wilkinson compared death statistics for female nuclear-weapons workers who wore badges to monitor radiation exposure with deaths of women who did not wear badges because they weren't exposed to radiation.

These results showed that there were 25 percent more deaths from all causes and 17 percent more deaths from all cancers among the unbadged workers. Wilkinson said this counterintuitive result may exist either because less robust workers "self-select" themselves out of jobs that involve exposure to radiation, or because department heads make those selections themselves.

Whatever the cause, Wilkinson said the study suggests that the healthy-worker effect also exists for female workers and must be taken into account in any studies of occupational health.

"Studies of radiation effects among nuclear workers must be restricted to those workers who are monitored for radiation exposures," he stated. "Inclusion of unmonitored workers will introduce a bias and give an inaccurate result."

Additional researchers on the study were Roger L. Priore, Sc.D., and Michael Fries, Ph.D., of the UB Department of Social and Preventive Medicine, and Robin Graham, Ph.D., and B. Lauren Young of the UB Department of Family Medicine.

Also Norman Trieff, Ph.D., Jennifer Jones and Laura Ray of the University of Texas Medical Branch, Galveston, and Jeanne Loughlin of the Epidemiology Research Institute in Newton Lower Falls, Mass.

The study was funded by the National Institute for Occupational Safety and Health.