Artificial Intelligence in Pediatric Education by SNAPPS and Clinical Reasoning Among Postgraduate Residents in a Tertiary Care Hospital in Central India

Background Competency-based medical education (CBME) emphasizes clinical reasoning as a core postgraduate competency, yet structured teaching strategies such as SNAPPS remain under-optimized in pediatric postgraduate training. Artificial intelligence (AI) has been proposed as a supervised cognitive scaffold within SNAPPS, but empirical outcome-level evidence, particularly from pediatric postgraduate education in India, is limited. Objective To evaluate the effect of AI-assisted SNAPPS on clinical reasoning, structured case-presentation quality, and knowledge acquisition among pediatric postgraduate residents, and to assess learner and faculty perceptions of the intervention. Methods A prospective, single-arm, within-subject educational interventional study was conducted over six months among 30 pediatric postgraduate residents (10 each from JR1, JR2, and JR3) in the Department of Pediatrics, NKP Salve Institute of Medical Sciences & Research Centre and Lata Mangeshkar Hospital, Nagpur. Following sensitization, all participants underwent baseline assessment, a traditional-SNAPPS teaching phase, and an AI-assisted SNAPPS phase using a standardized SNAPPS-mapped AI prompt protocol with mandatory faculty validation at every step. Outcomes were measured using the Script Concordance Test (SCT), a validated 30-point SNAPPS analytical rubric, a multiple-choice question (MCQ) knowledge test, and five-point Likert perception surveys for residents and faculty. Paired comparisons were analyzed using appropriate parametric and non-parametric tests; p < 0.05 was considered significant. Results Mean SCT scores improved significantly with AI-assisted SNAPPS (12.0 ± 2.0 to 19.6 ± 2.8, p < 0.001), a gain that exceeded the improvement achieved with traditional SNAPPS alone (11.8 ± 2.1 to 15.2 ± 2.5, p < 0.01). The total SNAPPS rubric score rose from 19.1 ± 3.2 to 27.6 ± 2.9 (p < 0.001), with significant gains across all six domains. MCQ knowledge scores improved similarly (12.7 ± 2.5 to 18.8 ± 1.9, p < 0.001). SCT and SNAPPS rubric scores were strongly and positively correlated (r = 0.72, p < 0.001). Both residents and faculty reported significantly higher satisfaction with AI-assisted SNAPPS across every measured domain (all p < 0.001). Conclusion AI-assisted SNAPPS, implemented as a faculty-supervised cognitive scaffold, is feasible, well accepted, and educationally effective for strengthening clinical reasoning, structured presentation and knowledge retention among pediatric postgraduate residents. The model merits validation in larger, multicentric, competency-based curricula.