Skip to main content
AAO-HNSF 2023 Annual Meeting & OTO Experience Main banner
Icon Legend
This session is not in your schedule.
This session is in your schedule. Click again to remove it.
Back
Rate ePoster

Patient Safety and Quality Improvement

(085) ChatGPT, the New ENT?

Monday, October 2, 2023
2:45 PM – 3:45 PM East Coast USA Time
Has Audio

    Senior Author(s)

  • Brian E. Benson, MD

    Chair, Department of Otolaryngology/Head & Neck Surgery
    Hackensack Meridian School of Medicine
    Hackensack, New Jersey, United States

    • Presenting Author(s)

  • Ellen Hong

    Medical Student
    Hackensack Meridian School of Medicine
    Yorba Linda, California, United States

    • Co-Author(s)

  • DB

    Daniel Basta, N/A

    Hackensack Meridian School of Medicine, USA

  • Bethany Ho

    Medical Student
    Hackensack Meridian School of Medicine
    Bernardsville, New Jersey, United States

  • NS

    Nihal Sriramaneni, n/a

    HMSOM, USA

  • JW

    Jordan Windsor, N/A

    Medical Student
    Hackensack Meridian School of Medicine
    Nutley, New Jersey, United States

  • ES

    Ezra Schneier

    Student
    Hackensack Meridian School of Medicine
    teaneck, New Jersey, United States

Disclosure(s):
Brian E. Benson, MD: No relevant relationships to disclose.
Ellen Hong: No relevant relationships to disclose.
Bethany Ho: No relevant relationships to disclose.
Jordan Windsor, N/A: No relevant relationships to disclose.
Ezra Schneier: No relevant relationships to disclose.


Introduction: ChatGPT is a popular new Artificial Intelligence (AI) computer program designed to simulate conversation with humans. Because human users may seek medical diagnoses from Chat GPT, we evaluated the ability of ChatGPT to make appropriate medical diagnoses based on a prompt containing common symptoms of a variety of otolaryngic disorders.

Methods: The signs and symptoms of 29 common otolaryngic disorders were obtained from reputable sources and verified by a board-certified otolaryngologist. These symptoms were provided to Chat GPT in the format, "I have (symptoms 1), (symptom 2), and (symptom 3). What do I have?" The results of each chat were recorded in a database and assessed for the presence or absence of the otolaryngic disorder associated with the constellation of symptoms described in the chat.

Results: Many responses from ChatGPT provided a list of possible diagnoses. ChatGPT was able to include the correct diagnosis in 19 of the 29 queries, and did include the correct organ system in two of the incorrect answers. The program frequently encouraged follow-up with a human physician, and in some cases of a diagnosis requiring urgent medical care, ChatGPT recommended immediate medical attention.

Conclusions: ChatGPT is a powerful new AI program that demonstrates some ability to communicate appropriate diagnoses when prompted with a clinical scenario. The problem of incorrect information provided by Large Language Model AI programs remains significant, and patient preference for chatbots versus traditional web searches remains to be seen.