A Prediction Model for Surgical Decision-Making in Rotator Cuff Tears Using Anatomical and Functional Factors

This study aims to develop a clinical prediction model that integrates anatomical characteristics, functional status, and relevant clinical factors to guide surgical decision-making for rotator cuff tears.

This retrospective study included patients with rotator cuff tears treated at Peking University International Hospital between March 2019 and February 2022. A total of 337 patients meeting the predefined inclusion criteria were selected and divided into a surgical group (n = 100) and a non-surgical group (n = 237) based on whether they underwent surgical or non-surgical treatment. By systematically reviewing electronic medical records, we collected demographic information, clinical characteristics (affected shoulder, history of shoulder trauma, and duration of symptoms), Neer classification, Neer impingement test results, and Jobe test results. Based on magnetic resonance imaging data, professional physicians evaluated acromion morphology classification and measured the acromion-humeral distance. Quantitative evaluation of shoulder function was performed using the modified Constant-Murley Score, with individual component scores systematically recorded to enable a detailed functional assessment. Statistical analysis was completed using R software. Specifically, baseline characteristics of the two patient groups were compared and analyzed. A backward stepwise selection method was subsequently used in the multivariate logistic regression analysis to identify independent predictors related to surgical decisions and construct a prediction model that estimates the probability of surgical intervention for patients with rotator cuff tears. The model's performance was comprehensively evaluated across three dimensions: discrimination ability, calibration, and clinical utility.

There were significant differences between the two groups in terms of age, shape of the acromion, positive rate of the Jobe sign, Neer classification, and functional score (p < 0.05). Multivariate logistic regression analysis demonstrated that age (odds ratio [OR] = 1.070), tear depth (OR = 4.414), types II and III acromion (OR = 8.138 and 11.209), and increased abduction angle (OR = 1.800) were independent predictors of surgical intervention. In contrast, external rotation (OR = 0.566), increased internal rotation angle (OR = 0.696), and Neer classification (OR = 0.297) were negative predictors (all p < 0.05). The constructed nomogram prediction model based on these predictors displayed excellent discrimination (area under the curve = 0.934, sensitivity = 0.890, specificity = 0.840) and calibration (Hosmer-Lemeshow test, p = 0.9977). Furthermore, the decision curve analysis confirmed its clinical utility.

This study developed a prediction model based on a nomogram, utilizing selected anatomical, functional, and clinical factors to assess the probability of surgical treatment for patients with rotator cuff tears. Internal validation demonstrated the model's good discriminatory ability and acceptable calibration. These findings suggest the model may aid in risk stratification and treatment selection; however, external validation and prospective impact studies are required before routine clinical implementation.