A formal neuroradiology second opinion service changes management in one in five cases: a retrospective cohort study of discrepancies and types of errors.

AIM: To quantify discrepancy rates between referring reports and subspecialist neuroradiology second-opinion reports in a UK tertiary neuro centre, describe the clinical impact of the service, and classify discrepancies using the Kim-Mansfield modification of the Renfrew error taxonomy. MATERIAL AND METHODS: Retrospective review of consecutive formal neuroradiology second opinion reports over 12 months (April 2024 to March 2025) at a tertiary neuro centre. Agreement between referring reports and second-opinion reports was classified into: major discrepancy, minor discrepancy, or complete agreement. Discrepancies were further categorised by error type using the Kim-Mansfield classification. RESULTS: A total of 580 studies were reviewed (magnetic resonance imaging [MRI]: 487/580, 84%; computed tomography [CT]: 93/580, 16%). The overall discrepancy rate was 42.2% (245/580), including 21.4% major discrepancies (124/580) and 20.9% minor discrepancies (121/580). Vascular studies were the commonest major discrepancies (61/124, 49.2%). Error typing across all discrepancies demonstrated predominantly cognitive/interpretive categories: faulty reasoning was the most frequent (109/245, 44.5%), followed by complacency (54/245, 22.0%), underreading (37/245, 15.1%), and satisfaction of search (28/245, 11.4%). In major discrepancies, faulty reasoning remained the most common (53/124, 42.7%), followed by complacency (34/124, 27.4%) and underreading (26/124, 21%). Major discrepancies were associated with high clinical impact, including a change in diagnosis (123/124, 99.2%), a change in treatment (116/124, 93.5%), and a change in investigation change (85/124, 68.5%). CONCLUSION: Subspecialist neuroradiology second-opinion reporting identified discrepancies in over two-fifths of cases, including management-altering discrepancies in approximately one in five. Faulty reasoning was the dominant error type, supporting targeted quality improvement interventions focused on interpretive/cognitive pitfalls.