JMIR Publications

ABSTRACT

Background:

Artificial intelligence (AI), more specifically large language models (LLMs), hold significant potential to transform the landscape of emergency care delivery. Although enthusiasm for integrating LLMs into emergency medicine (EM) is growing, the existing literature is characterized by a disparate collection of individual studies, conceptual analyses, and preliminary implementations. Given these complexities and gaps in understanding, a cohesive framework is needed to make sense of the existing body of knowledge on the application of LLMs in EM.

Objective:

This scoping review sought to map the current literature on the potential uses of LLMs in EM and identify directions for future research.

Methods:

Using PRISMA-ScR criteria, we searched Ovid MEDLINE, Embase, Web of Science, and Google Scholar for articles published between January 2018 and August 2023 that discussed LLM use in EM. We excluded other forms of artificial intelligence. Titles and abstracts were screened, and each full text article was independently reviewed by two authors. Data was abstracted independently and 5 authors performed a collaborative quantitative and qualitative synthesis of the data.

Results:

Of 1992 identified citations, 42 were included. Studies were predominantly from 2022-2023 and conducted in the USA and China. Four major themes emerged: (1) Clinical decision support, including applications in public health messaging, triage, diagnosis, treatment recommendations, and outcome predictions; (2) Workflow efficiency, through information retrieval and synthesis to reduce physician cognitive load; (3) Risks and ethics, with concerns about model accuracy, transparency, and legal implications noted; and (4) Education and communication, with potential uses in medical training, patient counseling, and knowledge dissemination identified.

Conclusions:

This review establishes an initial framework of the capabilities and limitations of LLMs in EM based on reported use cases and identifies key areas for future research. These include prospective validation of proposed applications, developing standards for responsible use, exploring provider and patient perceptions, and fostering physician literacy in artificial intelligence. Thoughtful collaboration and critical evaluation will be essential to safely and effectively integrate LLMs into emergency care.