Working hours and ill-health — A more serious relationship than it appears?

Abstract

In a case-control study conducted in Japan Fukuoka et al. [1] report a substantial difference in mean working hours between two groups of full-time workers, myocardial infarction patients admitted to hospital (mean 58.3 h, SD 15.0) and a healthy control group (mean 50.7 h, SD 8.6). This difference was statistically significant ( p <0.01) using an unpaired t-test, indicating a positive association between working hours and disease. For reasons that will be explained, this analysis may well substantially underestimate the hazard of long working hours. The fundamental issue in assessing the link between long working hours and ill-health is the disentangling of direct causation, confounding and selection effects. Direct causal mechanisms may relate to the admittedly poorly defined construct of stress, its effect on smoking and drinking habits, and unhealthy or increased eating due to a disrupted lifestyle and late evening meals preceded by snacking resulting in increased total intake. Fukuoka et al. appropriately controlled only for age and sex, unlike other related studies in this area [2–5] which appear to have overmatched by adjusting for factors such as body mass index, smoking, hypertension and lipid levels which could well be part of the causal chain from overwork to ill-health. A correct analysis would seek to determine to what degree the crude effect of overwork may be mediated via such factors, not to use them to explain it away. Two distinct, opposite kinds of selection effects are plausible in this context. Some individuals may select themselves for long working hours including Fmoonlighting_ to help make ends meet, and deprivation, whether absolute or relative to peers, is recognised as a risk factor. This kind of selection would lead to a positive association.