"Objectives
Previous studies have consistently reported evidence of large significant associations between measures of psychological health and sickness absence. Some of this association, however, may be confounded by relevant covariates that have not been controlled. By using data with repeated observations from the same individuals, this study aimed to quantify the bias due to unobserved characteristics that are time invariant.
Methods
Longitudinal data from the Household, Income, and Labour Dynamics in Australia (HILDA) Survey were used to estimate negative binomial regression models of the number of annual paid sickness absence days. Observations spanning the period 2005–2012, and covering all employed persons aged 15–64 years, were used (56 348 observations from 13 622 individuals).
Results
Significant associations between the number of paid sickness absence days taken each year and scores on the mental health subscale of the SF-36 (MHI-5) were found. Inclusion of correlated random effects (which effectively control for unobserved person-specific factors that do not vary over time), however, resulted in a marked decline in the magnitude of this association. For persons with severe depressive symptoms (MHI-5 ?52), the estimated incidence rate ratios were in the range 1.13–1.14 for men and 1.10–1.12 for women.
Conclusions
Poor mental health is a risk factor affecting work attendance, but the magnitude of this effect, at least in a country where the rate of sickness absence is relatively low, is modest."

Support: Digital