Longitudinal analysis of effects of operator experience on accuracy for ultrasound detection of supraspinatus tears.

Background: Magnetic resonance imaging (MRI) is highly accurate in diagnosing full-thickness rotator cuff tears but has the limitations of high cost and patient-based inconveniences. Diagnostic ultrasound of the rotator cuff has become popular, but its accuracy is operator dependent. We hypothesized that the incremental learning associated with initiation of shoulder ultrasonography in the orthopedic office setting could be quantified.

Materials And Methods: Ultrasound was performed by 2 surgeons in different patient populations each consisting of 200 patients within the clinical setting. Analysis of the operator learning curves based on sensitivity, specificity, positive predictive value, negative predictive value, and κ agreement scores was performed with MRI used as the reference standard.

Results: Surgeon 1 detected 67 (ultrasound) of 72 (MRI) full-thickness supraspinatus tears, and surgeon 2 detected 92 (ultrasound) of 95 (MRI) full-thickness tears. Agreement between ultrasound and MRI improved with increasing number of ultrasounds performed. The maximal rate of increase in sensitivity occurred by the 50th ultrasound for both surgeons. The maximal rate of increase in specificity occurred by the 75th ultrasound for both surgeons. Overall accuracy to detect full-thickness supraspinatus tears plateaued after 100 ultrasounds.

Conclusions: The majority of learning to detect full-thickness supraspinatus tears by ultrasound occurred in the surgeons' first 50 cases. The ability to accurately distinguish a normal supraspinatus tendon by ultrasound took longer, occurring within the first 75 cases. The majority of learning before accuracy plateaued occurred within the first 100 ultrasounds. The recommended operator experience necessary to evaluate the supraspinatus tendon before clinical application was 100 ultrasounds.

Level Of Evidence: Level I, Diagnostic Study.