Disentangling the associations between employment, income and antidepressant use: an application of the parametric G-formula

Maarten J. Bijlsma, Max-Planck Institute for Demographic Research
Lasse Tarkiainen, University of Helsinki
Mikko Myrskylä, Max-Planck Institute for Demographic Research and London School of Economics and Political Science (LSE)
Pekka Martikainen, University of Helsinki

Background: Employment, income and health are intertwined. Unfortunately, the existing literature typically does not treat them as such. This hinders estimation of the causal effects of these variables on one another. Data & methods: We use annual Finnish register data on a cohort of individuals aged 15+ (n = 42,172) in the period 1996-2007. We model the time-dependent relations between employment, income, antidepressant purchases, other drug purchases, and education, while adjusting for potential confounders. We perform mediation analysis using the G-formula; reciprocal time-dependent relations are modeled and life-course trajectories under particular interventions are simulated. Results: In the data, 75% of observed person-years were employed, 12% unemployed and the remainder being some other status. By intervening on employment so that all unemployed person-years were employed at all times, the population-averaged hazard ratio of first antidepressant use was reduced by 5.8% (95% CI: 1.5% to 9.4%). Roughly 67% of this reduction is a direct effect of employment, 20% is due to income, while the remainder goes through other pathways. Conclusion: Being employed lowers the risk of depression. A large part of this component is likely psychological, since only part of the employment effect is mediated by material factors such as income.