Predicting musculoskeletal sick leave in tertiary hospital staff using classification trees for targeted ergonomic intervention

BackgroundWork-related musculoskeletal disorders (WMSD) are major occupational injury and sick leave drivers in healthcare settings. However, predictive approaches for identifying high-risk worker profiles and estimating sick leave duration remain limited.ObjectiveTo identify risk profiles associated with WMSD-related sick leave among hospital staff and predict both the occurrence and duration of sick leave using classification tree models.MethodsA retrospective observational study was conducted using 705 overexertion-related WMSD incidents reported during 2021-2024 in three tertiary public hospitals. Classification and Regression Tree models were developed to predict sick leave occurrence and duration based on sociodemographic, organisational, and task-related variables. Model performance was evaluated using 10-fold cross-validation.ResultsSick leave occurred in 64.8% of cases, with task type, age, work shift, and anatomical region affected being the main predictors. Meanwhile, medium-term absences (10-29 days) accounted for 24.5% of cases, with age and hospital unit being the strongest predictors of duration. The models achieved an overall accuracy of 68.1% and 60% for predicting sick leave occurrence and duration, respectively.ConclusionsClassification tree models provide an interpretable and practical approach for predicting WMSD-related sick leave in hospital settings. The identified risk profiles may support targeted ergonomic interventions and preventive strategies in healthcare environments.