Shoulder MRI examinations are regularly not ordered according to existing appropriateness criteria, according to a new single-site study published by the Journal of the American College of Radiology.

Scott E. Sheehan, MD, MS, department of radiology at William S. Middleton Memorial Veterans Hospital in Madison, Wis., and colleagues reviewed data from 237 consecutive shoulder MRI exams performed at the Department of Veterans Affairs (VA) tertiary care hospital in 2013. The research team observed the medical records of each patient and, using established imaging guidelines, determined its appropriateness.

Overall, 45 percent of the exams were inappropriately ordered. Forty-four percent of exams ordered by nonorthopedic providers were inappropriate, and that number for orthopedic specialists was 17 percent.

“Our data suggest that efforts to improve ordering practices targeting both primary care and other nonorthopedic providers would yield the highest returns toward ordering appropriateness, though orthopedic providers would also benefit from such interventions,” the authors wrote.

Sheehan et al. also found that ultrasound (US) would have been appropriate for 66 percent of the patients who received MRI exams. Eighty-five percent of those patients could have had “all relevant pathologies characterized” after being combined with a radiograph.

Using US instead of MRI in all of those cases, the team said, could have resulted in significant savings for the hospital.

“The potential cost savings of substituting musculoskeletal US for MRI are substantial,” the authors wrote. “Our estimated cost differential of $321.59 per substituted examination is similar to values reported in the literature.”

The authors also noted that implementation of clinical decision support as part of a computerized physician order entry system can improve communication and overall exam appropriateness by “ensuring that the most relevant clinical questions have been addressed before consideration of advanced imaging.”

“These tools can also be targeted toward specific providers or clinical settings where overutilization is more prevalent, and can be designed to minimize their obtrusiveness to ordering providers,” the authors wrote.

Sheehan and colleagues said limitations to their study included the potential lack of important clinical context, although electronic health records were closely reviewed.