Entrustable professional activities were developed as a cornerstone of competency-based medical education. Rather than asking whether a resident has attended enough lectures or logged enough hours, EPAs ask a deceptively simple question: can this trainee be trusted to perform this core professional activity, unsupervised? In general surgery, the American Board of Surgery has rolled out a national EPA framework covering activities such as performing an operation, managing critically ill patients, and communicating with families. Faculty rate trainees on scales of supervision, from direct observation to full independence, and the assessments are meant to be formative, feeding forward into the resident’s ongoing development rather than serving merely as summative judgments. The mobile platform developed for the EPA project makes these assessments quick to complete in the middle of a busy clinical day, which is precisely why the accompanying narrative comments matter so much: they are the mechanism by which a rating becomes a lesson.

The research team, led by Gabrielle Moore of the University of Utah Spencer Fox Eccles School of Medicine, together with colleagues at the University of Utah and Wellstar MCG Health in Augusta, Georgia, set out to interrogate those narratives systematically. They collected EPA assessments of general surgery trainees completed by faculty between July 2023 and June 2025 across multiple institutions, yielding 896 assessments that included written narrative feedback. Each comment was then analyzed using deductive coding, a qualitative method in which two independent researchers categorize text according to a predefined framework drawn from previously established feedback literature. The framework captured six characteristics: whether the feedback was specific, whether it was coaching-oriented, whether it was formative, whether it was appreciative, whether it was evaluative, and whether it included person-focused descriptors such as comments on a trainee’s confidence or character rather than their observable performance.

The coding revealed a striking imbalance in what faculty choose to write. Nearly four in five assessments, 78.5 percent, contained specific feedback, and 75.7 percent were appreciative, praising what the resident had done well. But the developmental elements were far scarcer. Only 27.2 percent of comments were formative, meaning they were designed to guide future improvement, and only 21.4 percent contained coaching language that actively worked through how the trainee could get better. Just 8 percent included person-focused descriptors. In other words, the typical EPA narrative was specific and complimentary but rarely instructional, a pattern consistent with earlier studies suggesting that faculty feedback in surgery tends toward praise rather than critique. Because EPAs are explicitly intended to provide formative feedback for core activities of general surgery, the scarcity of formative and coaching content strikes at the heart of what the assessment tool was designed to do.

The most consequential finding, however, was not simply that feedback was thin, but that its quality varied systematically with the trainee’s level of entrustment. Residents judged to be practice-ready, meaning they could plausibly function at the level expected of a new independent surgeon, were dramatically less likely to receive coaching or formative feedback than their less advanced peers. After adjusting for other factors using mixed-effects logistic regression, practice-ready trainees had roughly 92 percent lower odds of receiving coaching feedback, with an odds ratio of 0.07, and about 92 percent lower odds of formative feedback, with an odds ratio of 0.08, compared with trainees who were not yet practice-ready. Both differences were highly statistically significant. The pattern is intuitive on one level, since faculty may feel there is little left to teach a resident who is nearly independent, but it inverts the logic of deliberate practice. Even expert surgeons continue to refine their technique, and the transition to independent practice is arguably the moment when candid, forward-looking feedback carries the highest stakes.

Case complexity produced a similar paradox. When residents participated in complex cases, their feedback was less likely to include coaching, with an odds ratio of 0.50, or formative language, with an odds ratio of 0.38, than when they participated in straightforward cases. Complex operations are precisely where technical nuance, decision-making under uncertainty, and teamwork under pressure are most instructive, yet the written narratives attached to those experiences offered less developmental content. The authors suggest this may reflect the cognitive load on faculty during demanding cases, or an implicit assumption that a resident entrusted with a complex case has already proven their competence. Either way, the result means that the richest learning opportunities in the operating room may be generating the poorest written guidance.

The study also uncovered gender-linked patterns that echo a growing body of literature on bias in medical assessment. Male trainees were significantly less likely than their female counterparts to receive evaluative feedback, with an odds ratio of 0.65, meaning their assessments were less likely to contain explicit judgment about their performance. There were no significant differences in feedback characteristics by trainee race or ethnicity, a null finding the authors report transparently. On the faculty side, the direction of the effect reversed: male faculty were less likely than female faculty to provide specific feedback, with an odds ratio of 0.33, or coaching language, with an odds ratio of 0.47. Previous work has documented gender differences in operative feedback and in the language used to describe surgical trainees, including studies of vascular surgery trainees and analyses of gender dyads between attending surgeons and residents. The new findings add general surgery EPA narratives to that evidence base and point toward faculty development as a concrete intervention point.

Methodologically, the study is notable for combining qualitative rigor with quantitative modeling. The deductive coding framework, applied independently by two researchers, grounds the text analysis in established constructs from the feedback literature rather than ad hoc categories, and the use of mixed-effects logistic regression allows the authors to estimate the association between each feedback characteristic and factors such as phase of care, case complexity, entrustment level, and demographics while accounting for the clustering of assessments within institutions, faculty, and residents. That clustering matters, because feedback is shaped by relationships: a given attending’s style, a given resident’s trajectory, and a given program’s culture all influence what gets written. The retrospective design and the reliance on voluntarily written narratives impose limits, since the analysis can identify associations but cannot establish why faculty write what they write, and comments may reflect the constraints of a quick mobile assessment as much as the faculty member’s intent.

The practical implications are direct. The authors argue that future faculty development should focus on increasing constructive feedback given to practice-ready residents and on mitigating gender bias in assessment. That could mean training attendings to treat high-entrustment assessments not as a signal to stop teaching but as an opportunity to coach at a higher level of performance, and it could mean building prompts or structured templates into the EPA platform that nudge faculty toward formative language. Prior research on feedback quality in surgery, including studies comparing end-of-rotation evaluations with workplace-based assessments, has repeatedly found that narrative comments are often generic and non-actionable, and national validity studies of EPAs have shown that the tool works best when the narratives do their job. As competency-based medical education continues to replace time-based training across surgical specialties, the words attached to each assessment may ultimately matter as much as the numbers, and this study makes clear that those words still fall short of the coaching that surgical education promises.

Subject of Research: Narrative feedback quality in entrustable professional activity assessments of general surgery residents

Article Title: What do entrustable professional activities really tell us? Evaluating narrative feedback in general surgery

Article References: Moore, G., Hegeholz, D., Sun, T., Weaver, M. L., Ward, E. P., Mabes, E. S., & Jaiswal, K. (2026). What do entrustable professional activities really tell us? Evaluating narrative feedback in general surgery. Global Surgical Education – Journal of the Association for Surgical Education, 5(1), Article 137. https://doi.org/10.1007/s44186-026-00542-2

Image Credits: AI Generated

DOI: 10.1007/s44186-026-00542-2

Keywords: entrustable professional activities, general surgery, surgical education, narrative feedback, competency-based medical education, formative assessment, coaching, entrustment, gender bias, faculty development, residency training, mixed-effects regression