This edition of the Occupational and Environmental Medicine includes an analysis of the link between unemployment benefits and purchased prescription drugs, and to what extent the two tended to co-occur over a 6-year period.1 The authors show that an incident prescription of a psychotropic drug was twofold to threefold more likely to occur close to a period of unemployment than the same individual's risk when unemployment was not an issue. This relationship was stronger for men, more pronounced for psychotropic drugs than other classes of prescription drugs, and was at its highest 1–3 months prior to onset of unemployment. The timing coincides with the notice period in Norway, where the study was carried out.

The results add to the well-established relationship between mental health and unemployment: on average, those who are involuntarily out of work have higher levels of psychological distress than those who have work. Previous systematic reviews conclude with a causal effect where becoming unemployed reduces mental health, and where finding good work again improves mental health.2–4 But elements of health selection may still be at play. Mental illness can be an obstacle to employment in the first place, and once having lost work, mental illness does not help in re-employment.2 As experimental designs with random allocation to unemployment are not an option, further insight into cause and selection requires carefully designed observational studies.

On this backdrop, Kaspersen et al’s study is a most welcome addition to the literature. Their detailed data allowed them to study the timing of the relationship very carefully. And importantly, the long timeframe and large sample allowed them to study changes within individuals over time. A standard observation study approach is to try to isolate the association of interest through careful statistical control of confounders and covariates. But even in many well-designed observational studies, the threat of residual confounding remains through inaccurate measurement, ignorance of important confounders and covariates, and having to leave out ‘unquantifiable’ key factors.5 By comparing individuals with themselves, the authors controlled for factors that remain unchanged over time within that individual. And having done so, their results all the more convincingly suggest that unemployment causes or at least evokes distress.

It is however important to bear in mind that purchase of prescription drugs is not fully equivalent to mental illness or elevated symptom loads. A drug purchase first requires active help-seeking from the patient, then prescription as a response from the physician, before the patient has to comply and purchase the drug. Prescription data therefore represents a “filtered” version of mental illness.

The associations for subtypes of psychotropics were of similar strength. It is possible that a sudden shock of unemployment could trigger first-time psychosis, but it is also plausible that a vulnerable subset sought medical help once struck by unemployment. Does that rule out unemployment as a cause? Epidemiology differentiates between ‘sufficient and component causes’, but we too often think about causes in terms of single, complete mechanisms—which in reality are rare. Given an understanding of ‘cause’ as an event that preceded disease and without which the disease would not have occurred at that time,6 unemployment remains a very likely component cause of mental illness given the evidence.

Decomposing the shock of unemployment requires insight into what work means to people. While work for many is far more than a pay cheque, the mere economic aspects of work remain important. This study was carried out in Norway, a country known for a strong welfare system. And while previous studies have suggested that quality welfare systems could curb some of the detrimental consequences of unemployment,7 the results obtained here remind us that unemployment remains a major challenge despite systems that ensure partial monetary compensation.

Helping people get back into good employment makes sense, and there is also some evidence for it.2 In terms of prevention, securing fair and transparent processes during downsizing is one avenue, as not just becoming unemployed, but also features of the process matter for mental health.8 ,9 The strong links between unemployment and poor mental health has eagerly been translated to ‘work is good for health’. While the slogan provides a counterbalance to a sometimes monotonous search for workplace-based stress and risk,10 it is a truth with modifications. The quality of work is of high importance,11 and an Australian study found that mental health benefits of return to work were contingent on the psychosocial qualities in the new job.12

Finally, when Kaspersen et al demonstrated these effects in such a large scale study, it reminds us that striving for a general low unemployment rate is an important component of universal mental health promoting policies.