Work

AI-enabled supervised case-based teaching and end-of-rotation assessment performance in neurosurgery residency training: a single-centre non-randomized educational implementation study

Year: 2026

Type: article

Abstract: Standardized neurosurgery residency training requires repeated case practice, timely feedback and supervised clinical reasoning. Lecture-based learning (LBL) delivers knowledge efficiently but offers limited process-level evidence of learner performance. We evaluated whether AI-enabled, mentor-supervised case-based teaching was associated with end-of-rotation performance and exploratory learner-reported experience in routine implementation. This single-centre, parallel-group non-randomized implementation study included 62 residents or rotating trainees completing neurosurgery training between 1 June 2024 and 31 May 2026. Allocation followed departmental arrangements and AI-platform exposure; no random sequence or allocation concealment was used. The AI-human collaboration group ( n = 29) received conventional teaching plus AI-platform case training and mentor debriefing; the LBL group ( n = 33) received conventional teaching only. No pre-intervention theory, technical-skill or clinical-reasoning scores were collected. The primary outcome was composite score; secondary outcomes comprised domain scores, exploratory learner-reported experience, teaching satisfaction and platform metrics. Measured demographic and training characteristics did not differ between groups, but baseline outcome scores were unavailable. The AI-human collaboration group had higher theory, technical-skill, clinical-reasoning and composite scores. The composite-score MD was 4.36 points (95% CI 2.58 to 6.14; Cohen’s d = 1.25; adjusted P < 0.001). Exploratory learner-reported experience was higher (26.07 ± 2.60 vs. 21.70 ± 1.90), but its questionnaire had limited internal consistency (Cronbach’s alpha = 0.555). Teaching satisfaction was also higher. Averaged faculty ratings showed good inter-rater reliability for technical skill (ICC(2,2) = 0.786) and clinical reasoning (ICC(2,2) = 0.806). In this single-centre implementation study, AI-enabled, mentor-supervised case-based teaching was associated with higher end-of-rotation assessment scores than LBL, particularly in knowledge acquisition and case-based clinical reasoning. The workflow may support cognitive training through structured virtual cases, formative feedback and mentor debriefing. Given the non-randomized, post-only, dose-imbalanced design, these findings need confirmation in baseline-controlled, intensity-matched studies that assess retention, clinical transfer and operative competence.

Source: BMC Medical Education

DOI: 10.1186/s12909-026-10225-7