Scenario-based comparative evaluation of ChatGPT-4o and physician groups in pediatric minor head trauma
Irish Journal of Medical Science, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Basım Tarihi: 2026
- Doi Numarası: 10.1007/s11845-026-04512-x
- Dergi Adı: Irish Journal of Medical Science
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, EMBASE, MEDLINE, Biomedical Reference Collection: Corporate Edition (EBSCO), Health Research Premium Collection (ProQuest)
- Anahtar Kelimeler: Pediatric minor head trauma, ChatGPT-4o, PECARN, Scenario-based assessment, Clinical decision support
- Sağlık Bilimleri Üniversitesi Adresli: Evet
Özet
Background: Interest in the use of ChatGPT-4o in scenario-based clinical assessment has increased substantially in recent years. However, studies evaluating ChatGPT-4o in pediatric head trauma scenarios and comparing it with different physician groups remain limited. Aims: To evaluate the scenario-based performance of ChatGPT-4o in pediatric head trauma and compare it with that of emergency physicians, neurosurgeons, and pediatricians. Methods: This study included 60 pediatric patients who presented between 15 December 2024 and 15 June 2025 and met the inclusion criteria. After clinical follow-up, cases were converted into multiple-choice case scenarios and classified into red, yellow, and green zones according to PECARN. These scenarios were answered by 42 physicians from emergency medicine, neurosurgery, and pediatrics (n=14 per group) and by ChatGPT-4o. Concordance of scenario-based management responses with PECARN recommendations was compared statistically. Results: Of the 60 cases, 25.0% (n=15) were classified as red zone, 50.0% (n=30) as yellow zone, and 25.0% (n=15) as green zone. ChatGPT-4o showed lower scenario-based performance than all physician groups in red-zone cases. When non-contrast brain CT was accepted as the correct option in the yellow zone, ChatGPT-4o had the lowest overall accuracy (median: 24.50). When observation was accepted as correct, ChatGPT-4o showed the highest accuracy both in the yellow zone (median: 17.00; p=0.001) and overall (median: 35.50; p<0.001). ChatGPT-4o showed the highest accuracy in green-zone cases (median: 8.50). Conclusion: ChatGPT-4o did not demonstrate adequate scenario-based performance in critical pediatric head trauma cases. However, it may have potential as a supportive tool in non-critical case scenarios.