Your browser doesn't support javascript.
loading
Artificial intelligence and ChatGPT: An otolaryngology patient's ally or foe?
Langlie, Jake; Kamrava, Brandon; Pasick, Luke J; Mei, Christine; Hoffer, Michael E.
Afiliação
  • Langlie J; University of Miami Miller School of Medicine, Miami, FL 33136, United States of America.
  • Kamrava B; Department of Otolaryngology, University of Miami Health System, Miami, FL 33136, United States of America.
  • Pasick LJ; Department of Otolaryngology, University of Miami Health System, Miami, FL 33136, United States of America.
  • Mei C; Department of Otolaryngology, University of Miami Health System, Miami, FL 33136, United States of America.
  • Hoffer ME; Department of Otolaryngology, University of Miami Health System, Miami, FL 33136, United States of America; Department of Neurosurgery, University of Miami Health System, Miami, FL 33136, United States of America. Electronic address: michael.hoffer@miami.edu.
Am J Otolaryngol ; 45(3): 104220, 2024.
Article em En | MEDLINE | ID: mdl-38219629
ABSTRACT

BACKGROUND:

As artificial intelligence (AI) is integrating into the healthcare sphere, there is a need to evaluate its effectiveness in the various subspecialties of medicine, including otolaryngology. Our study intends to provide a cursory review of ChatGPT's diagnostic capability, ability to convey pathophysiology in simple terms, accuracy in providing management recommendations, and appropriateness in follow up and post-operative recommendations in common otolaryngologic conditions.

METHODS:

Adenotonsillectomy (T&A), tympanoplasty (TP), endoscopic sinus surgery (ESS), parotidectomy (PT), and total laryngectomy (TL) were substituted for the word procedure in the following five questions and input into ChatGPT version 3.5 "How do I know if I need (procedure)," "What are treatment alternatives to (procedure)," "What are the risks of (procedure)," "How is a (procedure) performed," and "What is the recovery process for (procedure)?" Two independent study members analyzed the output and discrepancies were reviewed, discussed, and reconciled between study members.

RESULTS:

In terms of management recommendations, ChatGPT was able to give generalized statements of evaluation, need for intervention, and the basics of the procedure without major aberrant errors or risks of safety. ChatGPT was successful in providing appropriate treatment alternatives in all procedures tested. When queried for methodology, risks, and procedural steps, ChatGPT lacked precision in the description of procedural steps, missed key surgical details, and did not accurately provide all major risks of each procedure. In terms of the recovery process, ChatGPT showed promise in T&A, TP, ESS, and PT but struggled in the complexity of TL, stating the patient could speak immediately after surgery without speech therapy.

CONCLUSIONS:

ChatGPT accurately demonstrated the need for intervention, management recommendations, and treatment alternatives in common ENT procedures. However, ChatGPT was not able to replace an otolaryngologist's clinical reasoning necessary to discuss procedural methodology, risks, and the recovery process in complex procedures. As AI becomes further integrated into healthcare, there is a need to continue to explore its indications, evaluate its limits, and refine its use to the otolaryngologist's advantage.
Assuntos
Palavras-chave

Texto completo: 1 Coleções: 01-internacional Base de dados: MEDLINE Assunto principal: Otolaringologia / Inteligência Artificial Tipo de estudo: Guideline Limite: Humans Idioma: En Revista: Am J Otolaryngol Ano de publicação: 2024 Tipo de documento: Article País de afiliação: Estados Unidos

Texto completo: 1 Coleções: 01-internacional Base de dados: MEDLINE Assunto principal: Otolaringologia / Inteligência Artificial Tipo de estudo: Guideline Limite: Humans Idioma: En Revista: Am J Otolaryngol Ano de publicação: 2024 Tipo de documento: Article País de afiliação: Estados Unidos