Clinical disagreements between staff radiologists are rare but can potentially have a significant impact on patient care, according to research published Thursday in JACR.

Out of nearly 1.9 million radiology reports generated at one large academic medical center, only 114 had discrepant opinions between the two imaging specialists (or about 0.006%). However, more than half (51%) of these cases were categorized as “major” discrepancies, with changes in clinical management made in 56% of the total disagreements, Harvard Medical School experts wrote Feb. 23.

Oftentimes, these disagreements can occur verbally, in more informal fashion in the reading room, at tumor board or in multidisciplinary clinics. Findings of the study may underline the need to document such discrepancies and ensure that the issue is eventually resolved.

“The low number of documented discrepant opinions may represent underreporting of such events,” radiologist Pamela J. DiPiro, MD, with the Center for Evidence-Based Imaging, Department of Radiology, at Brigham and Women’s Hospital, and co-authors noted. “Encouraging alert use for intradepartmental discrepant opinions may help optimize patient care and document what was communicated by second opinion radiologists.”

The retrospective analysis included all alerts about discrepant opinions, logged over a more than two-year period ending at the beginning of 2022. Discrepancies were defined as instances where a consultant radiologist provided a different interpretation from the one formally rendered by their peer at the institution. To address this issue, the institution had previously adopted a policy on discrepant alerts, requiring a “closed-loop notification” of the consultant’s second opinion to the original rad, with the latter needing to acknowledge the alert within 30 days.

Of the total, 13% discrepancies were labeled as minor (unlikely to be clinically significant), 36% moderate (might be significant), while the balance were major (likely). DiPiro also noted that the rate of report addendums for discrepant alerts was fourfold higher with this system (21%) when compared to using a peer learning approach (5%).

“The latter tend to be less timely, as they are sent by a radiologist interpreting a current study to a colleague regarding findings they interpreted on the prior study for that patient,” the authors advised. “By using our ‘discrepant opinion’ alert, in addition to discussion of results with the referring clinician, timely feedback is provided to the original interpreting radiologist.”