Machine Learning to Predict Return to Work After Medical Rehabilitation for Musculoskeletal Disorders: A...

Area of Science:

• Rehabilitation Medicine
• Data Science
• Occupational Health

Background:

• Musculoskeletal disorders are a leading cause of work disability.
• Predicting return to work (RTW) is crucial for patient outcomes and healthcare economics.
• Identifying factors influencing RTW aids in optimizing rehabilitation strategies.

Purpose of the Study:

• To apply machine learning for predicting return to work (RTW) after medical rehabilitation for musculoskeletal disorders.
• To quantify the contribution of various features and their interactions in RTW prediction.
• To identify novel predictive factors and interactions beyond traditional regression models.

Main Methods:

• Retrospective cohort study of 685,890 individuals undergoing rehabilitation for musculoskeletal disorders.
• Utilized the German Pension Insurance's Rehabilitation Statistics Database (RSD).
• Developed and validated a Light Gradient Boosting Machine (LightGBM) model, employing Shapley values for feature importance and interaction analysis.

Main Results:

• 65.0% achieved stable return to work (RTW).
• The LightGBM model showed strong predictive performance (AUC-ROC 0.867).
• Key predictors included employment duration, income, pre-rehabilitation work incapacity, occupation, and age; significant interactions were found.

Conclusions:

• Machine learning effectively predicts RTW post-musculoskeletal rehabilitation.
• Confirms established predictors like prior employment and work incapacity.
• Highlights the value of machine learning in uncovering complex feature interactions for RTW prediction.