Effect of Guided ChatGPT Use on Documented Clinical Reasoning and Its Retention Among Emergency Medicine Students: A Randomized Controlled Trial

Abstract Background: Clinical reasoning is a core competency in emergency medicine, and its structured teaching remains challenging. Generative artificial intelligence (ChatGPT) may represent a useful educational tool, but its impact on learning consolidation beyond immediate benefit remains uncertain. We assessed whether guided use of ChatGPT improves immediate documented clinical reasoning performance without compromising its retention over time. Methods: A single-center, two-arm, parallel-group, exploratory randomized controlled trial was conducted in the emergency department of Mohammed VI University Hospital, Marrakech, Morocco. Sixty medical students (5th–6th year) were randomized 1:1, using sequentially numbered, opaque, sealed envelopes generated by an independent investigator, to a ChatGPT group (simulated clinical cases with guided GPT-4o assistance) or a control group (no assistance); outcome assessors were blinded to group and time point, although participants could not be blinded to the intervention received. The primary outcome was the total Revised-IDEA (R-IDEA) score (/10), measured immediately after the intervention (T0) and again one month later under unassisted conditions (T1), analyzed by analysis of covariance (ANCOVA) adjusted for the T0 score. Results: All 60 participants were followed at T0 and T1, with no loss to follow-up. At T0, the R-IDEA score was significantly higher in the ChatGPT group (median 9.50 vs 8.50; p < 0.001; r = 0.96). At T1, the pattern reversed: the ChatGPT group showed a significant decline (median Δ = −0.75), whereas the control group improved (Δ = +1.00; p < 0.001; r = 0.96). ANCOVA confirmed a group effect at T1 favoring the control group (adjusted difference = 0.48 points, 95% CI 0.16–0.80; p = 0.004; partial η² = 0.136). Perceived confidence and benefit in the ChatGPT group were not correlated with actual performance change. Conclusions: Conclusions: Guided ChatGPT assistance improves immediate documented clinical reasoning performance, but this benefit does not persist at one month and is accompanied by poorly calibrated learner confidence. These findings support structured pedagogical oversight of generative AI in medical education, combined with active learning-consolidation strategies to ensure durable skill acquisition. Trial registration: ISRCTN registry ISRCTN13261015, registered 3 August 2026. Retrospectively registered.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-17
DOI
https://doi.org/10.21203/rs.3.rs-10463281/v1
Primary Topic
Artificial Intelligence in Healthcare and Education
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Effect of Guided ChatGPT Use on Documented Clinical Reasoning and Its Retention Among Emergency Medicine Students: A Randomized Controlled Trial

Hayate Errifaiy, Hicham Nejmi, Naila Boukoub, Manal Rhezali et al.
5 links
Research Square
Artificial Intelligence in Healthcare and Education
preprint

Effect of Guided ChatGPT Use on Documented Clinical Reasoning and Its Retention Among Emergency Medicine Students: A Randomized Controlled Trial

Hayate Errifaiy, Hicham Nejmi, Naila Boukoub, Manal Rhezali, Hicham Maghtoum
preprint en

Abstract

Abstract Background: Clinical reasoning is a core competency in emergency medicine, and its structured teaching remains challenging. Generative artificial intelligence (ChatGPT) may represent a useful educational tool, but its impact on learning consolidation beyond immediate benefit remains uncertain. We assessed whether guided use of ChatGPT improves immediate documented clinical reasoning performance without compromising its retention over time. Methods: A single-center, two-arm, parallel-group, exploratory randomized controlled trial was conducted in the emergency department of Mohammed VI University Hospital, Marrakech, Morocco. Sixty medical students (5th–6th year) were randomized 1:1, using sequentially numbered, opaque, sealed envelopes generated by an independent investigator, to a ChatGPT group (simulated clinical cases with guided GPT-4o assistance) or a control group (no assistance); outcome assessors were blinded to group and time point, although participants could not be blinded to the intervention received. The primary outcome was the total Revised-IDEA (R-IDEA) score (/10), measured immediately after the intervention (T0) and again one month later under unassisted conditions (T1), analyzed by analysis of covariance (ANCOVA) adjusted for the T0 score. Results: All 60 participants were followed at T0 and T1, with no loss to follow-up. At T0, the R-IDEA score was significantly higher in the ChatGPT group (median 9.50 vs 8.50; p < 0.001; r = 0.96). At T1, the pattern reversed: the ChatGPT group showed a significant decline (median Δ = −0.75), whereas the control group improved (Δ = +1.00; p < 0.001; r = 0.96). ANCOVA confirmed a group effect at T1 favoring the control group (adjusted difference = 0.48 points, 95% CI 0.16–0.80; p = 0.004; partial η² = 0.136). Perceived confidence and benefit in the ChatGPT group were not correlated with actual performance change. Conclusions: Conclusions: Guided ChatGPT assistance improves immediate documented clinical reasoning performance, but this benefit does not persist at one month and is accompanied by poorly calibrated learner confidence. These findings support structured pedagogical oversight of generative AI in medical education, combined with active learning-consolidation strategies to ensure durable skill acquisition. Trial registration: ISRCTN registry ISRCTN13261015, registered 3 August 2026. Retrospectively registered.

Research Square
Cadi Ayyad University (MA)
Good health and well-being, Quality education
Artificial Intelligence in Healthcare and Education
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon