Unemployment and subsequent depression: A mediation analysis using the parametric G-formula

Abstract

The effects of unemployment on depression are difficult to establish because of confounding and limited understanding of the mechanisms at the population level. In particular, due to longitudinal interdependencies between exposures, mediators and outcomes, intermediate confounding is an obstacle for mediation analyses. Using longitudinal Finnish register data on socio-economic characteristics and medication purchases, we extracted individuals who entered the labor market between ages 16 and 25 in the period 1996 to 2001 and followed them until the year 2007 (n = 42,172). With the parametric G-formula we estimated the population-averaged effect on first antidepressant purchase of a simulated intervention which set all unemployed person-years to employed. In the data, 74% of person-years were employed and 8% unemployed, the rest belonging to studying or other status. In the intervention scenario, employment rose to 85% and the hazard of first antidepressant purchase decreased by 7.6%. Of this reduction 61% was mediated, operating primarily through changes in income and household status, while mediation through other health conditions was negligible. These effects were negligible for women and particularly prominent among less educated men. By taking complex interdependencies into account in a framework of observed repeated measures data, we found that eradicating unemployment raises income levels, promotes family formation, and thereby reduces antidepressant consumption at the population-level.

Modelling the Socio-Economic Determinants of Fertility: A Mediation Analysis using the Parametric g-formula

“…Our method employs the g-formula, which was developed by Robins to estimate time-varying effects in the presence of time-varying confounding and mediation (Daniel et al, 2013;Bijlsma et al, 2017;Robins, 1986;Robins and Hernán, 2009). The g-formula is implemented in a series of steps.…”

“…Similarly to other causal inference approaches, the g-formula uses counterfactual theories of causal inference to enable researchers to estimate the effects of a cause (Greenland et al, 1999), such as the hypothetical effect of treatment on the chance of experiencing a cardiac event in the presence of time-varying confounding (Daniel et al, 2013). Another advantage of the g-formula is that it was developed to estimate time-varying effects in the presence of time-varying confounding and mediation (Robins, 1986) and is therefore capable of testing hypothetical changes in life course processes while accounting for their interdependence (Robins and Hernán, 2009;Daniel et al, 2013;Bijlsma et al, 2017).…”

The Impact of Unemployment on Antidepressant Purchasing

“…4 One determinant of depression is unemployment. 5,6 Unemployment can result in loss of financial means, social contacts, and purposeful activity, all of which increase the risk of depression. [5][6][7] From the reverse perspective, depression is a risk factor for unemployment, as individuals suffering from depression may experience greater difficulty in finding and retaining employment.…”

“…5,6 Unemployment can result in loss of financial means, social contacts, and purposeful activity, all of which increase the risk of depression. [5][6][7] From the reverse perspective, depression is a risk factor for unemployment, as individuals suffering from depression may experience greater difficulty in finding and retaining employment. 8,9 Because of the interrelationships between unemployment and depression, many methods will struggle to estimate an unbiased causal effect of unemployment on depression.…”

“…[14][15][16] The g-formula has been used before to model the complex interrelationships between unemployment and antidepressant purchasing, an indicator of diagnosis of depression. 5 An earlier study found that, among young adults entering the labor market for the first time, reducing unemployment is estimated to result in less antidepressant purchasing, especially for low-educated men. 5 However, unobserved time-constant confounding may affect the estimates in that study.…”