Background: It is unknown which combination of patient information and clinical tests might be optimal for the diagnosis of rotator cuff tears. This study aimed to determine the diagnostic value of nine individual clinical tests for evaluating rotator cuff tear and to develop a prediction model for diagnosing rotator cuff tear.
Methods: This prospective cohort study included 169 patients with shoulder complaints. Patients who reported a previous shoulder dislocation were excluded from the analysis (N = 69). One experienced clinician conducted 25 clinical tests of which 9 are specifically designed to diagnose rotator cuff pathology (empty can, Neer, Hawkins-Kenney, drop arm, lift-off test, painful arc, external rotation lag sign, drop sign, infraspinatus muscle strength test). The final diagnosis, based on magnetic resonance arthrography (MRA), was determined by consensus between the clinician and a radiologist, who were blinded to patient information. A prediction model was developed by logistic regression analysis.
Results and discussion: In this cohort, 38 patients were diagnosed with rotator cuff tears. The individual overall accuracy of the rotator cuff clinical tests was 61%-75%. After backward selection, the model determined that the most important predictors of rotator cuff tears were higher age and a positive Neer test. This internally validated prediction model had good discriminative ability (area under the receiver operating characteristic curve (AUC) = 0.73).
Conclusion: Our results showed that individual clinical shoulder tests had moderate diagnostic value for diagnosing rotator cuff tear. Our prediction model showed improved diagnostic value. However, the prediction value is still relatively low, supporting a low threshold for additional diagnostic tests for the diagnosis of rotator cuff tears.
Level of evidence: Study of diagnostic test: level I.