Peer-to-Peer Case Review as a Strategy to Improve Sepsis Education in Graduate Medical Education

We conducted an educational intervention using a peer-to-peer case review process to improve emergency medicine (EM) resident sepsis knowledge and diagnostic confidence. Local quality audits in our academic emergency department (ED) identified poor compliance in addressing adult cases of severe sepsis and septic shock, suggesting an educational gap. We evaluated whether a facilitated, peer-to-peer case review process improved EM resident sepsis knowledge and diagnostic confidence.

We conducted a single-site, pre-test/post-test educational QI evaluation among senior EM residents. Using routinely audited severe sepsis and septic shock cases that did not meet bundle expectations, residents completed brief, asynchronous peer case reviews with a standardized checklist and provided structured written feedback to the original care teams. Educational impact was evaluated at Kirkpatrick Level 2 (learning) using paired pre-post assessments: a 12-item single-best-answer knowledge test and a 4-point self-reported diagnostic confidence survey. Paired pre/post knowledge scores were compared with the Wilcoxon signed rank test; confidence was summarized descriptively. The institutional review board determined this not human research.

Mean knowledge accuracy improved from 70% (8.36/12) to 85% (10.14/12), an increase of 14.9 percentage points (p = 0.003). The proportion scoring in the 76-100% band increased by 29%, and the number of scores over 90% increased by 53%. Confidence increased for sepsis (71.4% to 78.6%) and remained constant for severe sepsis at 78.6% but decreased for septic shock (92.9% to 85.7%).

A resident-led, asynchronous peer case review process was associated with significantly improved sepsis knowledge and minimal gains in diagnostic confidence for sepsis and severe sepsis. This low-resource approach is feasible for unit-level implementation and may be adaptable to other time-sensitive conditions with complex bundle requirements.