Abstract:Background: Reliable prediction of postoperative dislocation after reverse total shoulder arthroplasty (RSA) would inform patient counseling as well as surgical and postoperative decision making. Understanding interactions between multiple risk factors is important to identify those patients most at risk of this rare but costly complication. To better understand these interactions, a game theory-based approach was undertaken to develop machine learning models capable of predicting dislocation-related 90-day readmission following RSA. Material & methods: A retrospective review of the Nationwide Readmissions Database was performed to identify patients who underwent RSA between 2016 and 2018 with a subsequent readmission for prosthetic dislocation. Of the 74,697 index procedures included in the data set, 740 (1%) experienced a dislocation resulting in hospital readmission within 90 days. Five machine learning algorithms were evaluated for their ability to predict dislocation leading to hospital readmission within 90 days of RSA. Shapley additive explanation (SHAP) values were calculated for the top-performing models to quantify the importance of features and understand variable interaction effects, with hierarchical clustering used to identify cohorts of patients with similar risk factor combinations. Results: Of the 5 models evaluated, the extreme gradient boosting algorithm was the most reliable in predicting dislocation (C statistic = 0.71, F2 score = 0.07, recall = 0.84, Brier score = 0.21). SHAP value analysis revealed multifactorial explanations for dislocation risk, with presence of a preoperative humerus fracture; disposition involving discharge or transfer to a skilled nursing facility, intermediate care facility, or other nonroutine facility; and Medicaid as the expected primary payer resulting in strong, positive, and unidirectional effects on increasing dislocation risk. In contrast, factors such as comorbidity burden, index procedure complexity and duration, age, sex, and presence or absence of preoperative glenohumeral osteoarthritis displayed bidirectional influences on risk, indicating potential protective effects for these variables and opportunities for risk mitigation. Hierarchical clustering using SHAP values identified patients with similar risk factor combinations. Conclusion: Machine learning can reliably predict patients at risk for postoperative dislocation resulting in hospital readmission within 90 days of RSA. Although individual risk for dislocation varies significantly based on unique combinations of patient characteristics, SHAP analysis revealed a particularly at-risk cohort consisting of young, male patients with high comorbidity burdens who are indicated for RSA after a humerus fracture. These patients may require additional modifications in postoperative activity, physical therapy, and counseling on risk-reducing measures to prevent early dislocation after RSA.

Predictors of inpatient mortality following reverse shoulder arthroplasty

Survivorship after reverse total shoulder arthroplasty and predictors of 1-year and overall mortality

Reverse versus anatomic total shoulder arthroplasty: A large matched cohort analysis

Understanding risk for early dislocation resulting in reoperation within 90 days of reverse total shoulder arthroplasty: extreme rare event detection through cost-sensitive machine learning

Reverse Shoulder Arthroplasty in Patients with Rheumatoid Arthritis: Early Outcomes, Pitfalls, and Challenges

Predictors of Return to Emergency Department and Readmission Following Primary Elective Total Shoulder Arthroplasty

A Comprehensive Analysis of the Causes and Predictors of 30-Day Mortality Following Hip Fracture Surgery

Enhanced Outcomes and Safety of Reverse Shoulder Arthroplasty in Rotator Cuff Deficiency and Proximal Humerus Fractures: A 22-Patient Retrospective Analysis

Relationship between the preoperative grip strength and postoperative shoulder strength of patients treated via reverse shoulder arthroplasty

Predictors of Mortality in Elder Patients With Proximal Humeral Fracture

Outcomes After Reverse Shoulder Arthroplasty for the Treatment of Glenohumeral Osteoarthritis in Patients Under and Over 70 Years of Age: A Propensity Score- Matched Analysis

A Pooled Analysis and Systematic Review

Predictors of 30-day mortality following hip/pelvis fractures

Reverse Shoulder Arthroplasty for Proximal Humerus Fractures and Reverse Shoulder Arthroplasty for Elective Indications Should Have Separate Current Procedural Terminology (CPT) Codes

Functional and radiographic outcomes of reverse shoulder arthroplasty with a minimum follow-up of 10 years.

Prediction of Non-Home Discharge Following Total Hip Arthroplasty in Geriatric Patients

Reverse shoulder arthroplasty versus hemiarthroplasty for acute proximal humeral fractures. A blinded, randomized, controlled, prospective study

Preoperative Predictors for Mortality Following Hip Fracture Surgery: A Systematic Review and Meta-Analysis

Reverse total shoulder arthroplasty compared to stemmed hemiarthroplasty for proximal humeral fractures: a registry analysis of 5946 patients

Stemless Anatomic and Reverse Shoulder Arthroplasty in Patients under 55 years of age with Primary Glenohumeral Osteoarthritis: An analysis of the Australian Orthopaedic Association National Joint Replacement Registry at five years

Predictors of 30-day mortality, unplanned related readmission and reoperation among isolated closed femoral shaft fractures