Related Experiment Video
Updated: Jul 12, 2025
05:20
Development of a Rabbit Chronic-Like Rotator Cuff Injury Model for Study of Fibrosis and Muscular Fatty Degeneration
Published on: March 31, 2023
1.3K
Using Machine Learning to Predict Nonachievement of Clinically Significant Outcomes After Rotator Cuff Repair
Rafael Krasic Alaiti1,2, Caio Sain Vallio3, Jorge Henrique Assunção4,5
1Research, Technology, and Data Science Office, Grupo Superador, São Paulo, Brazil.
Orthopaedic Journal of Sports Medicine
|October 23, 2023
Summary
Machine learning models can predict if patients will not achieve the minimal clinically important difference (MCID) after rotator cuff repair surgery. While performance was modest, some algorithms outperformed traditional methods.
Area of Science:
- Orthopaedic Surgery
- Machine Learning
- Prognosis Prediction
Background:
- Limited research exists on machine learning for predicting patient-reported outcomes after rotator cuff repair.
- Machine learning may offer advantages over classical statistical methods in orthopaedic prognosis.
Purpose of the Study:
- To evaluate machine learning algorithms' ability to predict the nonachievement of the minimal clinically important difference (MCID) in disability.
- To assess prediction performance using preoperative data for rotator cuff repair patients at 2-year follow-up.
Main Methods:
- Utilized routinely collected clinical, demographic, and imaging data from 500 rotator cuff repairs.
- Trained and evaluated 8 machine learning algorithms, including random forest and LightGBM, using a 70/30 split for training and testing.
- Assessed model performance using the area under the receiver operating characteristic curve (AUC).
Main Results:
- The highest AUCs were achieved by the random forest classifier (0.68) and LightGBM (0.67).
- Most machine learning algorithms showed better performance than logistic regression (AUC, 0.59).
- Model performance was generally lower than reported in other orthopaedic machine learning studies.
Conclusions:
- Machine learning algorithms show potential in predicting the nonachievement of the MCID for the American Shoulder and Elbow Surgeons (ASES) score post-rotator cuff repair.
- Further research is needed to optimize machine learning models for improved predictive accuracy in this patient population.
