Prediction models to identify workers at risk of sick leave due to low-back pain in the Dutch construction industry

• Main Authors: Lisa C Bosman, Lyan Dijkstra, Catelijne I Joling, Martijn W Heymans, Jos WR Twisk, Corné AM Roelen
• Format: Article
• Language: English
• Published: Nordic Association of Occupational Safety and Health (NOROSH) 2018-03-01
• Series: Scandinavian Journal of Work, Environment & Health
• Subjects: risk assessment; absenteeism; pain; musculoskeletal disease; construction industry; prediction model; prognostic research; roc analysis; low-back pain; sick leave; construction; worker
• Online Access: https://www.sjweh.fi/show_abstract.php?abstract_id=3703

OBJECTIVE: The aim of this study was to develop a prediction model based on variables measured in occupational health checks to identify non-sick listed workers at risk of sick leave due to non-specific low-back pain (LBP). METHODS: This cohort study comprised manual (N=22 648) and non-manual (N=9735) construction workers who participated in occupational health checks between 2010 and 2013. Occupational health check variables were used as potential predictors and LBP sick leave was recorded during 1-year follow-up. The prediction model was developed with logistic regression analysis among the manual construction workers and validated in non-manual construction workers. The performance of the prediction model was evaluated with explained variances (Nagelkerke’s R-square), calibration (Hosmer-Lemeshow test), and discrimination (area under the receiver operating curve, AUC) measures. RESULTS: During follow-up, 178 (0.79%) manual and 17 (0.17%) non-manual construction workers reported LBP sick leave. Backward selection resulted in a model with pain/stiffness in the back, physician-diagnosed musculoskeletal disorders/injuries, postural physical demands, feeling healthy, vitality, and organization of work as predictor variables. The Nagelkerke’s R-square was 3.6%; calibration was adequate, but discrimination was poor (AUC=0.692; 95% CI 0.568–0.815). CONCLUSIONS: A prediction model based on occupational health check variables does not identify non-sick listed workers at increased risk of LBP sick leave correctly. The model could be used to exclude the workers at the lowest risk on LBP sick leave from costly preventive interventions.