RETURN TO ABSTRACT LISTING


Objective:
Our Radiology department recently transitioned from peer review to peer learning across the hospital enterprise, which includes both teaching service lines (radiologists and residents/fellows) and non-teaching service lines (radiologists working alone). The purpose of this study was to assess whether the pathway of case interpretation (teaching vs. non-teaching service lines) impacted the distribution of cases marked for peer learning.

Materials and Methods:
This was an IRB-approved quality improvement retrospective study with exemption status. Cases were identified by a primary reader with adjudication by a second reader falling into one of four categories: concur, great call, constructive feedback, and discrepancy. All peer learning cases from 2020 to 2021 were identified via a retrospective query of our PACS system across 9 hospitals. Cases were classified as either using a teaching or non-teaching service line. Additional variables included: peer learning category (1-4), modality and years of experience for the attending. Cases in which service line classification could not be defined were excluded. Cases that fell into great call, constructive feedback, or discrepancy were used. The fraction of cases within the study group of categories were reported as percentages with 95% confidence intervals. Our primary (N0) hypothesis was there would be no difference in the fraction of cases within each of the above 3 categories between the two models. An exploratory sub-analysis of cases categorized as constructive feedback or discrepancy was also performed. Chi-square test of proportions was used to look for differences in the primary outcome measure (p < 0.05 was considered significant). Significance following multiple testing was inferred after Bonferroni corrections. All analysis was done using IBM SPSS Statistics.

Results:
The data review flagged 66,678 for potential peer-learning, 64,870 of which were categorized as “concur with report” and 91 cases with unclear service line designation both of which were not included in the analysis. Of the remaining 1,717 peer-learning cases that fell in categories 2, 3 and 4, 65.2% [n = 1120/1717] were from the non-teaching service line vs. 34.8% [n = 597/1717] from the teaching service line (p < 0.001). Analyzing only the cases with great call, constructive feedback and discrepancy by modality, we found a significant increase in the proportion of cases across all categories when the initial reads were completed from the non-teaching service line for CT (63.6% vs. 36.4% completed with a trainee, p < 0.001), and no association for other modalities (MRI, ultrasound, and radiography; p = 0.213).

Conclusion:
The number of “great calls” identified during peer-learning were higher when the primary reports were dictated in the non teaching workflow. However, the proportion of cases with constructive feedback or discrepancy were lower in the teaching workflow. This suggests a benefit in incorporating both workflows into one peer learning system and may also direct future efforts in identifying improvement initiatives for specific case classifications.