MP64-14 SURPRISING SUCCESS: AI's POTENTIAL HEAVY IMPACT IN TREATING UROLOGICAL TRAUMA

You have accessJournal of UrologyReconstruction: External Genitalia and Urotrauma (Including Transgender Surgery and Traumatic Fistula) II (MP64)1 May 2024MP64-14 SURPRISING SUCCESS: AI's POTENTIAL HEAVY IMPACT IN TREATING UROLOGICAL TRAUMA Yitzchak E. Katlowitz, Michael West, Ariel Schulman, and Jacob Khurgin Yitzchak E. KatlowitzYitzchak E. Katlowitz , Michael WestMichael West , Ariel SchulmanAriel Schulman , and Jacob KhurginJacob Khurgin View All Author Informationhttps://doi.org/10.1097/01.JU.0001008824.92877.7f.14AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookLinked InTwitterEmail Abstract INTRODUCTION AND OBJECTIVE: ChatGPT is an artificial intelligence with a publicly accessible user-friendly interface. It employs deep learning, a subset of machine learning based on artificial neural networks, to "speak" to the user. It is among the more popular of several artificial intelligence modalities that will likely be a large part of information gathering sources in the near future, employed by patients and physicians alike. Considering that responses to trauma are often algorithmic in nature, one may assume artificial intelligence would be excellent at managing trauma consults. This project is designed to assess ChatGPT's clinical competency compared to the American Urological Association (AUA) guidelines regarding the evaluation and management of urological trauma. METHODS: Each key point from the AUA guidelines for urotrauma was prompted to ChatGPT as a case presentation, clinical next-step recommendations, a general question, or a request for validation of information. The answers were then sorted into 3 categories all relative to the AUA guidelines: complete and accurate (CA), incomplete but accurate (IA), and incorrect or misleading (IM). RESULTS: There are 45 key points within the AUA guidelines for urotrauma, divided into 5 sections of renal, ureteral, bladder, urethral, and genital. For renal trauma, ChatGPT obtained 8/9 CA, 0/9 IA and 1/9 IM. For ureteral trauma, the results were 10/12 CA, 2/12 IA and 0/12 IM. For bladder trauma, 5/6 were CA, 1/6 were IA and 0/6 were IM. For urethral, 6/8 were CA, 0/8 were IA and 2/8 were IM. Finally, for genital trauma, 10/10 were CA, and 0/12 were IA and IM. In total, 39/45 (87%) were CA, 3/45 (6%) were IA, and 3/45 (6%) were IM. CONCLUSIONS: Trauma is often managed according to an algorithm in order to simplify decisions in a high-stress environment. With 93% of the answers having some form of accuracy, 87% of answers being accurate and complete and only 6% being incorrect or misleading, it appears ChatGPT is almost ready for urological trauma-related queries at this point in time. Considering minor inaccuracies could have dire consequences, it should be used in accordance with clinical acumen as an assistant, rather than full guidance. The near future may regularly involve artificial intelligence incorporated algorithms for the treatment of trauma. Download PPT Source of Funding: No funding was acquired © 2024 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 211Issue 5SMay 2024Page: e1046 Advertisement Copyright & Permissions© 2024 by American Urological Association Education and Research, Inc.Metrics Author Information Yitzchak E. Katlowitz More articles by this author Michael West More articles by this author Ariel Schulman More articles by this author Jacob Khurgin More articles by this author Expand All Advertisement PDF downloadLoading ...