Updating and prospective validation of a prognostic model for high sickness absence

C.A.M. Roelen, M.W. Heymans, J.W.R. Twisk, W. van Rhenen, S. Pallesen, B. Bjorvatn, B.E. Moen, N. Mageroy

Research output: Contribution to JournalArticleAcademicpeer-review


Objectives To further develop and validate a Dutch prognostic model for high sickness absence (SA). Methods Three-wave longitudinal cohort study of 2,059 Norwegian nurses. The Dutch prognostic model was used to predict high SA among Norwegian nurses at wave 2. Subsequently, the model was updated by adding person-related (age, gender, marital status, children at home, and coping strategies), health-related (BMI, physical activity, smoking, and caffeine and alcohol intake), and work-related (job satisfaction, job demands, decision latitude, social support at work, and both work-to-family and family-to-work spillover) variables. The updated model was then prospectively validated for predictions at wave 3. Results 1,557 (77 %) nurses had complete data at wave 2 and 1,342 (65 %) at wave 3. The risk of high SA was under-estimated by the Dutch model, but discrimination between high-risk and low-risk nurses was fair after re-calibration to the Norwegian data. Gender, marital status, BMI, physical activity, smoking, alcohol intake, job satisfaction, job demands, decision latitude, support at the workplace, and work-to-family spillover were identified as potential predictors of high SA. However, these predictors did not improve the model's discriminative ability, which remained fair at wave 3. Conclusions The prognostic model correctly identifies 73 % of Norwegian nurses at risk of high SA, although additional predictors are needed before the model can be used to screen working populations for risk of high SA. © 2014 Springer-Verlag Berlin Heidelberg.
Original languageEnglish
Pages (from-to)113-122
JournalInternational Archives of Occupational and Environmental Health
Issue number1
Publication statusPublished - 2015


Dive into the research topics of 'Updating and prospective validation of a prognostic model for high sickness absence'. Together they form a unique fingerprint.

Cite this