OpenAlex · Aktualisierung stündlich · Letzte Aktualisierung: 18.03.2026, 13:55

Dies ist eine Übersichtsseite mit Metadaten zu dieser wissenschaftlichen Arbeit. Der vollständige Artikel ist beim Verlag verfügbar.

P1055 ChatGPT in clinical practice: A preliminary assessment against ECCO 2025 guidelines for Inflammatory Bowel Disease

2026·0 Zitationen·Journal of Crohn s and Colitis
Volltext beim Verlag öffnen

0

Zitationen

1

Autoren

2026

Jahr

Abstract

Abstract Background Inflammatory bowel diseases (IBD) present a major challenge in gastroenterology, requiring personalized management based on up-to-date guidelines, such as those published by the European Crohn’s and Colitis Organisation (ECCO) in July 2025. The emergence of artificial intelligence (AI), particularly ChatGPT, raises questions about its potential role in clinical decision support. This study aims to evaluate the reliability of ChatGPT in providing responses aligned with ECCO 2025 guidelines for IBD. Methods Forty-four questions were posed to ChatGPT across six domains: initial diagnosis (n = 7), patient follow-up (n = 9), imaging and endoscopy (n = 10), colorectal cancer screening (n = 8), disease activity scores (n = 8), and management of pregnant women (n = 2). Each response was reviewed by an IBD expert and categorized into four groups: (1) correct and complete, (2) correct but incomplete, (3) mixed correct and incorrect, (4) incorrect. Results ChatGPT’s performance varied by topic. It was highly reliable in standardized areas, with 85.7% correct and complete responses for initial diagnosis and 100% adherence to ECCO 2025 recommendations for colorectal cancer screening. Responses regarding imaging in Crohn’s disease were 80% correct. However, in areas requiring individualized approaches, such as patient follow-up, only 11% of responses were complete, with frequent errors in surveillance intervals and underestimation of intestinal ultrasound. Responses on disease activity scores were often incomplete, lacking calculation methods and thresholds. Guidance on pregnant women was generally accurate. Conclusion These findings highlight the need for regular updates of AI models to incorporate recent clinical practice evolutions, particularly for personalized patient management. ChatGPT can serve as a useful decision-support tool, but it cannot replace clinical expertise, especially for complex or individualized IBD care. Conflict of interest: Dr. Amine, Meriem: aucun conflit d’intérêt

Ähnliche Arbeiten

Autoren

Institutionen

Themen

Artificial Intelligence in Healthcare and EducationRadiomics and Machine Learning in Medical ImagingInflammatory Bowel Disease
Volltext beim Verlag öffnen