EUVIMEDEUVIMED entdecken
CrossrefInfokarte

Accuracy of ChatGPT in Answering Parental Questions about Posterior Urethral Valves: An Observational Study

Mohd. S. Ansari, Alok Dutta, Anupam Shukla, Apoorva Saxena, Prakhar Pratap

Journal of Indian Association of Pediatric Surgeons · 2026-11 · 31 (6) · 955-960

DOI: 10.4103/jiaps.jiaps_144_26 ↗
AbstractBibliografische Angaben

Abstract

A BSTRACT Importance: Parents of children with posterior urethral valves (PUVs) frequently seek accessible information regarding diagnosis, treatment, and long-term outcomes. Artificial intelligence (AI) tools such as ChatGPT are increasingly used for this purpose; however, the accuracy, completeness, and consistency of such information remain uncertain. Objective: The objective of this study is to evaluate the accuracy and completeness of ChatGPT-generated responses to common parental questions about PUV. Design, Setting, and Participants: This cross-sectional and observational study assessed ChatGPT-generated responses to 14 commonly asked parental questions regarding PUV. Questions were derived from clinical interactions at a tertiary care pediatric urology center. Responses were evaluated by 25 pediatric urology experts with at least 5 years of postqualification experience. Experts independently rated each response using a four-category scale (misleading, helpful but incomplete, accurate, or accurate and comprehensive) and were blinded to each other’s assessments. Main Outcomes and Measures: The primary outcome was the distribution of expert ratings across response categories. Secondary analyses explored the patterns of agreement and the frequency of incomplete or potentially misleading responses. Results: All questions were considered relevant. Among 350 expert evaluations, 67.7% of responses were rated as accurate and comprehensive, 18.9% as helpful but incomplete, and 13.4% as potentially misleading. Eight responses received unanimous high-quality ratings, whereas six demonstrated variability in expert assessment, primarily due to omission of clinically important details such as key investigations and aspects of long-term management. Inter-rater reliability demonstrated substantial agreement (Fleiss’ κ =0.70). Conclusions and Relevance: ChatGPT-generated responses demonstrated variable accuracy, with a substantial proportion lacking important clinical details. While the model may provide general explanations to common parental queries, its outputs can be incomplete and may vary across instances, limiting reliability for independent medical guidance. These findings support a limited, supplementary role for AI-generated information and reinforce the continued necessity of clinician oversight in patient education.

Wissen, das zu Ihrer Arbeit passt.

Vom kostenlosen Einstieg über die persönliche Recherche bis zur vollständigen Research Suite.

Tarife ansehen →

Sind alle Verlagsvolltexte enthalten?

Nein. Bibliografische Nachweise, Abstracts und Volltexte sind unterschiedliche Inhalte. Ob ein Volltext verfügbar ist, hängt von der Quelle, Open Access und Ihren vorhandenen Lizenzen ab. Ein EUVIMED-Tarif ersetzt keine zusätzlichen Verlagsrechte.

Tarife ansehen →