Defining Clinically Meaningful Subgroups in Patients Undergoing Arthroscopic Rotator Cuff Repair Using Unsupervised...

Area of Science:

• Orthopedic Surgery
• Machine Learning in Medicine
• Patient Outcomes Research

Background:

• Clinically significant outcomes (CSOs) are crucial for evaluating arthroscopic rotator cuff repair (RCR) success.
• Predicting global CSO achievement after RCR remains a challenge.

Purpose of the Study:

• To identify distinct patient subgroups based on CSO achievement after elective arthroscopic RCR using unsupervised machine learning.
• To uncover factors influencing CSO achievement in RCR patients.

Main Methods:

• Analysis of a prospectively collected database of patients undergoing arthroscopic RCR.
• Utilized an unsupervised random forest algorithm to partition patients into subgroups based on CSO achievement.
• Employed stepwise multivariable logistic regression to identify predictors of optimal CSO achievement.

Main Results:

• Two distinct subgroups (optimal and suboptimal CSO achievement) were identified among 333 patients.
• Increased rotator cuff tear size (sagittal dimension >1.9 cm), number of tendons involved, and subscapularis involvement predicted suboptimal outcomes.
• Subpectoral biceps tenodesis was associated with a higher likelihood of CSO achievement compared to biceps tenotomy.

Conclusions:

• Unsupervised machine learning effectively identified clinically meaningful subgroups after arthroscopic RCR.
• Tear size, number of tendons involved, and subscapularis involvement are significant predictors of suboptimal CSO achievement.
• Biceps tenodesis demonstrates a protective effect, increasing the likelihood of achieving CSOs compared to tenotomy.