Development of Prediction Models for Sickness Absence Due to Mental Disorders in the General Working Population

Marieke F. A. van Hoffen, Giny Norder, Jos W. R. Twisk, Corné A. M. Roelen

Research output: Contribution to journalArticleAcademicpeer-review


Purpose This study investigated if and how occupational health survey variables can be used to identify workers at risk of long-term sickness absence (LTSA) due to mental disorders. Methods Cohort study including 53,833 non-sicklisted participants in occupational health surveys between 2010 and 2013. Twenty-seven survey variables were included in a backward stepwise logistic regression analysis with mental LTSA at 1-year follow-up as outcome variable. The same variables were also used for decision tree analysis. Discrimination between participants with and without mental LTSA during follow-up was investigated by using the area under the receiver operating characteristic curve (AUC); the AUC was internally validated in 100 bootstrap samples. Results 30,857 (57%) participants had complete data for analysis; 450 (1.5%) participants had mental LTSA during follow-up. Discrimination by an 11-predictor logistic regression model (gender, marital status, economic sector, years employed at the company, role clarity, cognitive demands, learning opportunities, co-worker support, social support from family/friends, work satisfaction, and distress) was AUC = 0.713 (95% CI 0.692–0.732). A 3-node decision tree (distress, gender, work satisfaction, and work pace) also discriminated between participants with and without mental LTSA at follow-up (AUC = 0.709; 95% CI 0.615–0.804). Conclusions An 11-predictor regression model and a 3-node decision tree equally well identified workers at risk of mental LTSA. The decision tree provides better insight into the mental LTSA risk groups and is easier to use in occupational health care practice.
Original languageEnglish
Pages (from-to)308-317
Number of pages10
JournalJournal of Occupational Rehabilitation
Issue number3
Early online date16 Aug 2019
Publication statusPublished - 1 Sep 2020

Cite this