Improving the selective use of plain radiographs in the initial evaluation of shoulder pain

In a previous retrospective study, we had found that shoulder radiographs were overused in the initial evaluation of patients presenting to the emergency department with shoulder pain. Our objective was to prospectively define clinical criteria that identify patients who have a low risk of radiographic findings that influence the management of the initial evaluation of shoulder pain in the emergency department.

All adult patients presenting to the emergency department of a university teaching hospital with shoulder pain were eligible to participate. Patients with deformities, penetrating wounds, referred pain, and those presenting for followup were excluded. Data were recorded prospectively using a standardized form. All radiographs were classified as therapeutically uninformative (TU) except fracture, acromioclavicular joint separation (> or =Grade 3), infection, or malignancy. A classification tree was used for analysis.

Two hundred six patients had shoulder radiographs, of which 88% were TU. The model yielded 3 sets of low risk patients with a high percentage of TU radiographs: (1) no fall, no swelling (99% TU: 106 TU, 1 radiograph with lytic lesions in a cancer patient); (2) a fall, but no swelling, and no pain at rest (100% TU, n = 18); (3) a fall and pain at rest, but no swelling and normal range of motion (100% TU, n = 10). Not performing radiographs on these low risk patients would have decreased radiograph utilization from 52 to 18% of all shoulder patients seen.

Our findings suggest that specific clinical criteria are able to identify patients who do not need radiographs as part of the initial management of shoulder pain. Prospective validation of these criteria are needed to confirm that these low risk patients can be safely evaluated without radiographs.