Large Language Models as Health Information Tools: Patient Use and Trust in Otolaryngology

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Patient-Reported Outcomes

Background:

• Patients increasingly utilize large language models (LLMs) for health information seeking.
• Understanding LLM usage patterns and their impact on otolaryngology patient care is limited.

Purpose of the Study:

• To characterize the use patterns, perceptions, and impact of LLMs on clinical communication among otolaryngology patients.
• To assess patient trust, perceived helpfulness, and preparedness after using LLMs for health information.

Main Methods:

• A cross-sectional survey was administered to adult patients at an academic otolaryngology clinic.
• The survey collected data on demographics, LLM use for general and health-related queries, and perceptions of LLM-generated information.
• Primary outcomes included overall LLM use for health information and pre-visit LLM usage.

Main Results:

• 71.2% of respondents reported prior LLM use, with 28.6% using LLMs for pre-visit information about their condition.
• LLM information was perceived as helpful by all participants and largely aligned with clinician explanations.
• Most patients felt more prepared for their visit, though trust in LLMs was moderate, with common concerns regarding accuracy and privacy.

Conclusions:

• LLM use for health information is prevalent among otolaryngology patients, associated with perceived benefits like improved preparedness.
• Despite moderate trust and accuracy concerns, factors such as accessibility and perceived empathy may drive LLM engagement.
• Most patients do not disclose LLM use to clinicians, highlighting a gap in communication.