Predicting Shoulder Function Improvement After Surgery: A Breakthrough with Machine Learning

Shoulder surgeries, specifically anatomic total shoulder arthroplasty (aTSA) and reverse total shoulder arthroplasty (rTSA), often aim to improve internal rotation (IR). However, predicting the extent of this improvement has proven challenging, especially for rTSA patients. In a recent study, researchers utilized machine learning to quantify and compare IR scores for both aTSA and rTSA patients, offering a groundbreaking way to predict shoulder function after surgery. 

In a study that analyzed clinical data from 2270 aTSA and 4198 rTSA patients. Researchers employed three supervised machine learning techniques to create predictive models for internal rotation at six different postoperative time points. The focus was on assessing accuracy in predicting IR scores using the full input feature set and two minimal input feature sets. 

Key Findings: 

  • Differences in IR Scores: rTSA patients consistently had lower mean IR scores and experienced less improvement compared to aTSA patients throughout the postoperative period. 
  • Activities of Daily Living (ADLs): While both groups showed significant improvement in performing daily activities, aTSA patients were notably more likely to do so. 
  • Machine Learning Accuracy: Using a minimal feature set of preoperative inputs, the machine learning algorithms exhibited equivalent accuracy in predicting IR scores for both aTSA (0.92-1.18 MAE) and rTSA (1.03-1.25 MAE) from 3 months to over 5 years post-surgery. 
  • Clinical Improvement Prediction: The predictive algorithms demonstrated an impressive 90% accuracy for aTSA and 85% accuracy for rTSA in identifying patients achieving minimal clinically important difference (MCID) IR score improvement. Additionally, they predicted, with 85% accuracy for aTSA and 77% accuracy for rTSA, which patients would achieve substantial clinical benefit (SCB) in IR score at 2-3 years post-surgery. 

This machine learning study showcases a significant advancement in predicting internal rotation after aTSA and rTSA surgeries. The use of minimal preoperative inputs allows accurate forecasts of shoulder function improvement at various postoperative time points.  

These predictive algorithms not only identify patients likely to experience clinical improvement but also help distinguish those surpassing critical satisfaction thresholds, providing valuable insights for both patients and healthcare professionals.

Learn more about shoulder surgery at King and Parsons here.

Success Stories

Real Patients, Real Results

Our patients are talking and sharing the good news about the excellent care and superior results they experienced our surgery center in Stratham, NH.

Watch and listen to our patients describe their individual orthopedic situations and experience with A.V.A.T.A.R®.

Have your own story to share?
Review us on Google

Don’t Live With Pain – Tell Us What’s Wrong

Guides and Advice

Patient Resources and Learning

Ongoing care is the name of the game. Recovery is a long-term process, and while our team strives to maintain a same-day outpatient surgery experience, it doesn't mean we leave you to figure out healing on your own.

First Visit Guide

Come prepared for your first visit by reading up on what to expect during your visit.

Learn More


Learn more about our prescription policy and how to ask questions about existing medications.

Learn More


Explore our payment options including bundled payments and get help regarding eligibility.

Learn More

Recent King & Parsons Insights

Recognized By