Yield of Learning Opportunities From a Radiology Random Peer Review Program
AJR Am J Roentgenol. 2018 Sep;211(3):630-634. doi: 10.2214/AJR.17.19373. Epub 2018 Jul 17.
Trinh TW1, Shinagare AB1, Khorasani R1.
OBJECTIVE: The purpose of this study is to assess the yield and usage of a radiology random peer review program in identifying potentially meaningful discrepancies in radiology reports that could serve as learning opportunities.
MATERIALS AND METHODS: This retrospective study was performed at a 776-bed tertiary academic hospital that renders more than 620,000 radiology reports annually. A workstation-embedded peer review system was implemented on January 1, 2014, followed by radiologist training over the next 12 weeks. From April 1, 2014, through September 30, 2017 (14 quarters), randomly selected radiologic studies were peer reviewed and scored as follows: 1, agree with original interpretation; 2, minor discrepancy, not likely clinically significant; 3, moderate discrepancy, may be clinically significant; or 4, major discrepancy, likely clinically significant. Reports scored as 3 or 4 were defined as having potentially meaningful discrepancies, and the discrepancy type was characterized. The primary outcome was the quarterly rate of potentially meaningful discrepancies. The secondary outcome was program usage rate (number of reports peer reviewed / total reports generated). Chi-square trend test assessed changes in outcomes over time.
RESULTS: Of a total of 42,891 peer reviews performed, the overall potentially meaningful discrepancy rate (or yield) was 0.5% (233/42,891). The quarterly potentially meaningful discrepancy rate decreased significantly over the study period (p < 0.0001). The quarter 14 potentially meaningful discrepancy rate (0.003) was 70% lower than the quarter 1 rate (0.011). Quarterly program usage also decreased significantly over the study period (p < 0.0001). Quarter 14 usage (0.015) was 56% lower than quarter 1 usage (0.034). Among 184 potentially meaningful discrepancy reports that could be categorized, the most common discrepancy type was error of perception (87/184).
CONCLUSION: A workstation-embedded random radiology peer review program had a very low yield in identifying learning opportunities and declining usage over time.