Observer variability in measuring elevation and external rotation of the shoulder. Primary Care Rheumatology Society Shoulder Study Group.

The aim of the study was to assess the observer variation between trained primary care physicians in their assessment of two key shoulder movements: elevation and external rotation. Six observers each examined and recorded their visual estimate of the range of movements in six patients assessed in random order. There was good agreement on the range of passive elevation assessed to the start of pain (if present): intraclass correlation coefficient (ICC) = 0.84, and to the point of maximum elevation: ICC = 0.95. There was no evidence of an important systematic bias between observers. By contrast, external rotation was poorly reproducible: ICC = 0.43, with important systematic differences between observers. In the second experiment, six observers simultaneously witnessed a range of movements in a single volunteer subject, and the agreement on their visual estimation of the angles achieved was assessed. There was a marked reduction in the systematic bias in external rotation, but agreement was still poor. Agreement for elevation remained high with a reduction in the small amount of bias observed in the first experiment when variability in both examination and visual assessment had been investigated. We conclude that shoulder elevation is a reliable measurement for use in multicentre studies by trained primary care physicians. By contrast, external rotation is poorly reproducible because of systematic variation in examination technique and random variation in visual assessment.