JMIR Publications

ABSTRACT

Background:

Recent surveys indicate that 58% of consumers actively use generative AI for health-related inquiries. Despite widespread adoption and potential to improve healthcare access, scant research examines the performance of AI chatbot responses regarding emergency care advice.

Objective:

We assessed the quality of AI chatbot responses to common emergency care questions. We sought to determine qualitative differences in responses from four free-access AI chatbots, for ten different serious and benign emergency conditions.

Methods:

We created 10 emergency care questions that we fed into the free-access versions of ChatGPT 3.5, Google Bard, Bing AI Chat, and Claude AI on November 26, 2023. Each response was graded by five board-certified emergency medicine (EM) faculty for eight domains of percentage accuracy, presence of dangerous information, factual accuracy, clarity, completeness, understandability, source reliability, and source relevancy. We determined the correct, complete response to the 10 questions from reputable and scholarly emergency medical references. These were compiled by an EM resident physician. For readability of the chatbot responses, we used the Fleischer-Kincaid Grade Level (FKGL) of each response from readability statistics embedded in Microsoft Word. Differences between chatbots were determined by Chi-square test.

Results:

Each of the four chatbots’ responses to the 10 clinical questions were scored across eight domains by five EM Faculty, for 400 assessments for each chatbot. Together, the four chatbots had the best performance in clarity and understandability (both 85%), intermediate performance in accuracy and completeness (both 50%), and poor performance (10%) for source relevance and reliability (mostly unreported). Chatbots contained dangerous information in 5-35% of responses, with no statistical difference between chatbots on this metric. ChatGPT, Google Bard, and Claud AI had similar performances across 7/8 domains. Only Bing AI performed better with more identified/relevant sources (40%, others 0-10%). Fleischer-Kincaid Reading level was 7.7-8.9 grade for all chatbots, except ChatGPT at 10.8, all too advanced for average emergency patients. Responses included both dangerous (e.g. start CPR with no pulse check) and generally inappropriate advice (e.g. loosen the collar to improve breathing without evidence of airway compromise).

Conclusions:

AI Chatbots, though ubiquitous, have significant deficiencies for emergency medicine patient advice, despite relatively consistent performance. Information for when to seek urgent/emergent care is frequently incomplete and inaccurate, and patients may be unaware of misinformation. Sources are not generally provided. Patients who use AI to guide healthcare assume potential risk. AI Chatbots for health may exacerbate disparities in social determinants of health and should be subject to further research, refinement, and regulation. We strongly recommend proper medical consultation to prevent potential adverse outcomes.