Original articlesHealthy Volunteer Effect in Industrial Workers
Introduction
Increased mortality rates have been repeatedly observed among non-responders to invitations to participate in epidemiological surveys. Mortality among Seventh-Day Adventist's [1] or among Japanese men in Honolulu [2] who did not respond to a mailed questionnaire, was higher than that among those who did. This “healthy volunteer effect” (HVE) has been confirmed in several other studies 3, 4, 5, 6, 7, 8 and emerges as an important source of bias in epidemiological research.
The workplace is a convenient setting for epidemiological research. Testing is usually done on site and does not require from participants to invest leisure or work time. However, occupationally active persons are a selected population capable of holding a job. They have been found to be healthier than the general population, thereby producing the “healthy worker effect” [9]. Whether the HVE applies also to such cohorts is uncertain. The only attempt we know of, to detect a HVE in occupationally active persons was carried out in 1957, and found no differences in mortality between participants and non-participants in an epidemiological survey [10].
The objective of the following study was to determine whether the HVE exists in a cohort of industrial workers. We compared the mortality and causes of death during 8 years of 3,795 male workers who responded in 1985 to an invitation to participate in a prospective study, with those of 1,507 workers who did not.
Section snippets
Study Population
An effort was made in 1985 to identify all male industrial employees of 21 plants throughout Israel. A total of 5547 employees of textile, heavy industry (tire and metal), light industries (printing and home products manufacture), and others (furniture, electronics, and food), were located and offered free of charge screening examinations for selected risk factors for cardiovascular disease (CVD). Nearly 70% were blue-collar workers. Of the 5547 workers, 3816 (68.8%) responded and were examined
Healthy Worker Effect
The 8-year cumulative mortality from cancer, CVD and other causes in the entire cohort of employees who were offered the examinations increased with age (Table 2). The age-adjusted standardized all cause mortality ratio in the entire cohort was 78 (95% CI: 68–89), that for CVD mortality was 75 (95% CI: 60–92), and that for cancer mortality was 94 (95% CI: 73–119).
Healthy Volunteer Effect
The SMRs were 99% (95% CI: 78–124) for non-participants and 71% (95% CI: 60–83) for participants (Table 3). The differences in
Discussion
The main finding of this study is the reduced mortality risk among subjects who volunteered to participate in screening examinations as compared to those who did not, and to the general population. While this “healthy volunteer effect” has previously been observed in other populations 1, 2, 3, 4, 5, 6, 7, 8, this study is the first to demonstrate such an effect in a cohort of industrial workers, a select population that was shown to be healthier than the general population to begin with. As
Acknowledgements
The authors thank the Committee for Preventive Action and Research in Occupational Health, Ministry of Labor and Social Welfare, Jerusalem, Israel for their support, and also Professor Manfred Green who was instrumental in instituting this study.
References (12)
- et al.
Healthy volunteer effect in a cohort study; temporal resolution in the Adventist Health Study
J Clin Epidemiol
(1996) - et al.
Some methodological problems in the long-term study of cardiovascular diseaseObservations on the Framingham Study
J Chron Dis
(1959) - et al.
The effect of non-response on risk ratios in a cardiovascular disease study
J Chron Dis
(1979) - et al.
Characteristics of respondents and non-respondents in a prospective study of osteoporosis
J Clin Epidemiol
(1991) - et al.
Response bias in the Honolulu Heart Program
Am J Epidemiol
(1989) - et al.
Differences between respondents and non-respondents in a population-based cardiovascular disease study
Am J Epidemiol
(1978)
Cited by (94)
Wearables, smartphones, and artificial intelligence for digital phenotyping and health
2020, Digital Health: Mobile and Wearable Devices for Participatory Health ApplicationsIn utero metal exposures measured in deciduous teeth and birth outcomes in a racially-diverse urban cohort
2019, Environmental ResearchCitation Excerpt :We have previously shown that children from WHEALS who donated a tooth to the study were of slightly older gestational age at delivery (39.1 ± 1.3 weeks compared to 38.7 ± 1.8 weeks) and higher birth weight Z-score (0.06 ± 0.95 compared to −0.15 ± 0.99 weeks) than WHEALS children who did not donate a tooth to the study (Cassidy-Bushrow et al., 2017). As reported elsewhere, the propensity for participation in research studies by healthier volunteers (with respect to both outcomes of interest and potential exposures) may bias the results of a study (Froom et al., 1999), potentially toward the null. There are known sex differences in the vulnerability to prenatal exposures (DiPietro and Voegtline, 2017); we found evidence for sex-specific effects of tooth manganese on birth weight Z-score in stratified models but not when tested with interaction terms, which may suggest that sex acts as an effect modifier rather than an interaction variable (VanderWeele, 2009).
A cohort study of mammography screening finds that comorbidity measures are insufficient for controlling selection bias
2018, Journal of Clinical EpidemiologyUsefulness of Non–High-Density Lipoprotein Cholesterol as a Predictor of Cardiovascular Disease Mortality in Men in 22-Year Follow-Up
2017, American Journal of CardiologyCitation Excerpt :However, data collected during the first and second phases of the CORDIS study were crosschecked to ensure reliability. A nondifferential information bias in outcome analysis may stem from mortality data obtained from NDRs; however, verification of NDR coding has previously indicated a coding accuracy of 91%.28 The authors have no conflicts of interest to disclose.