Video feedback with error-focused or correct-focused examples in surgical skills distance learning: A randomized trial

Background: Although it seems natural that surgical trainees would learn from demonstrations of a correct performance, evidence outside of surgical education has suggested that error-focused examples may promote error detection and improved procedural task performance. We hypothesized that feedback through error-focused videos would improve procedural learning more than correct-focused videos. Methods: We conducted a randomized controlled trial of video feedback comparing error-focused versus correct-focused examples. The participants were interviewees at our general surgery residency program in December 2020. All of the interviewees performed suturing and knot tying tasks on their interview day (baseline), with 70 common errors identified. For each error, we created an instructional feedback video in two formats: one video demonstrating the error and another demonstrating correct performance. The study participants received feedback videos based on baseline performance errors, with the format randomly assigned. Two blinded raters assessed the baseline and postintervention performances. Results: Thirty-seven interviewees enrolled and 17 submitted postintervention videos. The postintervention mean performance scores were significantly higher (P = .02) in the error-focused than the correct-focused example group (suturing [maximum score 18]: 16.9 vs 13.9 [difference 2.9; 95% CI 0.7, 5.1]; knot-tying [maximum score 24]: 21.6 vs 17.8 [difference 3.8; 95% CI 0.5, 7.0]). We found no between-group differences in performance time ([error-focused versus correct-focused] suturing: 246 vs 256 s; knot-tying: 170 vs 138 s; P = .08). Mean satisfaction with feedback was similar between groups (error-focused: mean = 5.3 versus correct-focused: mean = 5.2, out of 7; P =. 95). Conclusion: Feedback that highlights errors is associated with better learning of surgical skills than feedback demonstrating correct performance, confirming our hypothesis.