Leveraging ChatGPT to address patient queries on fever of unknown origin: opportunities and limitations

Abstract

Fever of unknown origin (FUO) remains one of the most intriguing and diagnostically challenging syndromes in internal medicine. The term FUO itself implies a passive/negative approach to diagnosis, suggesting a state of uncertainty that may perpetuate medical inertia. FUO is defined as a temperature ≥38.3°C (101°F) on several occasions for at least three weeks without an established diagnosis despite appropriate evaluation.¹ In resource-limited settings an active and pragmatic approach is essential to expedite diagnosis and mitigate burden of undiagnosed fever cases.² In contrast, access to large language models like ChatGPT has been increasing over past few years.³ Developed by OpenAI, ChatGPT is a large language model capable of giving human-like answers to queries related to health by compiling information from medical literature.⁵ Patients are increasingly using AI to clarify medical terminologies and physicians’ advice.³ Patients suffering from FUO search online for explanations of symptoms and investigations, where information may be incomplete or misleading.