Development and validation of a risk prediction model for work disability: multicohort study

Scientific Reports. 2017;7(1):1-12 DOI 10.1038/s41598-017-13892-1


Journal Homepage

Journal Title: Scientific Reports

ISSN: 2045-2322 (Online)

Publisher: Nature Publishing Group

LCC Subject Category: Medicine | Science

Country of publisher: United Kingdom

Language of fulltext: English

Full-text formats available: PDF, HTML



Jaakko Airaksinen (Finnish Institute of Occupational Health)

Markus Jokela (Department of Psychology and Logopedics, Faculty of Medicine, University of Helsinki)

Marianna Virtanen (Finnish Institute of Occupational Health)

Tuula Oksanen (Finnish Institute of Occupational Health)

Jaana Pentti (Department of Public Health, University of Turku)

Jussi Vahtera (Department of Public Health, University of Turku)

Markku Koskenvuo (Clinicum, Faculty of Medicine, University of Helsinki)

Ichiro Kawachi (Harvard T H Chan School of Public Health)

G. David Batty (Department of Epidemiology and Public Health, University College London)

Mika Kivimäki (Finnish Institute of Occupational Health)


Blind peer review

Editorial Board

Instructions for authors

Time From Submission to Publication: 20 weeks


Abstract | Full Text

Abstract Work disability affects quality of life, earnings, and opportunities to contribute to society. Work characteristics, lifestyle and sociodemographic factors have been associated with the risk of work disability, but few multifactorial algorithms exist to identify individuals at risk of future work disability. We developed and validated a parsimonious multifactorial score for the prediction of work disability using individual-level data from 65,775 public-sector employees (development cohort) and 13,527 employed adults from a general population sample (validation cohort), both linked to records of work disability. Candidate predictors for work disability included sociodemographic (3 items), health status and lifestyle (38 items), and work-related (43 items) variables. A parsimonious model, explaining > 99% of the variance of the full model, comprised 8 predictors: age, self-rated health, number of sickness absences in previous year, socioeconomic position, chronic illnesses, sleep problems, body mass index, and smoking. Discriminative ability of a score including these predictors was high: C-index 0.84 in the development and 0.83 in the validation cohort. The corresponding C-indices for a score constructed from work-related predictors (age, sex, socioeconomic position, job strain) were 0.79 and 0.78, respectively. It is possible to identify reliably individuals at high risk of work disability by using a rapidly-administered prediction score.