Accuracy of Visual Estimation of Active Shoulder Elevation Among Orthopedic Physicians: An Exploratory Video-Based Study.

Introduction Visual estimation of shoulder range of motion (ROM) is rapid and widely used, but its accuracy may vary across the arc of elevation. This study quantified angle-specific error in the visual estimation of active shoulder elevation by orthopedic physicians. Materials and methods In this cross-sectional video-based study, 50 physicians at different stages of orthopedic training viewed seven approximately 10-second videos showing active shoulder elevation at six reference angles: 30° (recorded twice, once in the coronal plane and once in the sagittal plane), 60°, 70°, 90°, 115°, and 160°. The reference angles were measured with a universal goniometer and checked with the phyphox smartphone application. Participants used a slider ranging from 0° to 180° to record each estimate. The primary outcome was signed estimation error, calculated as the estimated angle minus the reference angle. Exact agreement was analyzed descriptively. Angle-specific one-sample t-tests assessed whether the mean signed error differed from zero, with Holm correction for seven comparisons. Mean absolute error per participant was compared across training levels. Results The 50 participants generated 350 ratings. Mean estimates were closest to the reference value at 90° and 115°, with a mean signed error of -0.7° for both angles. After Holm correction, the only angle showing a statistically significant bias was 160°, which was underestimated by a mean of 21.4° (95% CI, -25.7° to -17.2°; adjusted p < 0.001). The 70° target was overestimated by 5.2° (95% CI, +1.3° to +9.2°; adjusted p = 0.063), and the 30° target recorded in the sagittal plane was overestimated by 5.8° (95% CI, -0.5° to +12.2°; adjusted p = 0.361). Neither comparison remained statistically significant after Holm adjustment for seven angle-specific tests. Exact agreement was highest at 90° (43/50, 86%), followed by 30° in the sagittal plane (23/50, 46%) and 30° in the coronal plane (19/50, 38%). Mean absolute error did not increase with less experience: it was lowest among third-year residents and did not differ between residents, fellows, and specialists as groups. Conclusion In this video-based setting, visual estimation of active shoulder elevation was not uniformly accurate and showed substantial angle-dependent bias, particularly marked underestimation at high elevation. Visual estimation may provide a rapid gross impression of shoulder elevation; however, standardized goniometric or validated digital measurement should be preferred when precise documentation or detection of change is required.