Prediction of acromial and scapular spine fractures after reverse total shoulder arthroplasty using machine learning:...

Area of Science:

• Orthopedic Surgery
• Biomedical Engineering
• Machine Learning in Healthcare

Background:

• Acromial and scapular spine fractures (ASSF) are infrequent but serious complications after reverse total shoulder arthroplasty (rTSA).
• Understanding the incidence and predictors of ASSF is crucial for improving patient outcomes.

Purpose of the Study:

• To determine the incidence and types of ASSF following rTSA.
• To develop and validate a machine learning model for predicting individual ASSF risk.
• To assess how implant design influences model performance.

Main Methods:

• Retrospective cohort study of 2,256 patients undergoing primary rTSA.
• Logistic regression machine learning model utilizing ten preoperative features, including implant design (MD, L, LD).
• Performance evaluation using sensitivity, specificity, and AUROC across the cohort and implant subgroups.

Main Results:

• Overall ASSF incidence was 4.1%, with varied Levy type distributions across implant designs (L: Type 2, MD: Type 3).
• Medialized-distalized (MD) implant design, cuff tear arthropathy, higher age, and osteoporosis were significant predictors.
• The model achieved a sensitivity of 0.71, specificity of 0.61, and AUROC of 0.71 in the test set, with fair performance across implant subgroups.

Conclusions:

• A machine learning model shows promise for preoperative identification of patients at risk for ASSF after rTSA.
• MD implant design, cuff tear arthropathy, advanced age, and osteoporosis are key predictors of ASSF.
• Personalized risk stratification can guide implant selection and postoperative monitoring.