โก Quick Summary
This scoping review explores the application of large language models (LLMs) and multi-agent systems in emergency medicine, analyzing 35 studies to identify current trends and gaps. The findings indicate that while LLMs show promise for task-level decision support, there is a need for more comprehensive research that reflects the dynamic workflows of emergency departments.
๐ Key Details
- ๐ Studies Analyzed: 35 studies from various databases
- ๐งฉ Focus Areas: Triage, diagnostic support, treatment decisions, documentation
- โ๏ธ System Types: Non-agentic LLMs (25), LLM-based agents (7), multi-agent systems (3)
- ๐ Evaluation Methods: Expert comparison, retrospective reviews, vignette comparisons
๐ Key Takeaways
- ๐ Dominance of Non-agentic LLMs: Most studies focused on non-agentic LLMs for task support.
- ๐ก Limited Scope: Research primarily concentrated on a narrow range of tasks.
- ๐ฉโ๐ฌ Evaluation Gaps: Few studies utilized workflow-level or prospective evaluations.
- ๐ฅ Clinical Application: LLMs have potential for improving decision support in emergency settings.
- ๐ Need for Multimodal Integration: Future research should explore integrating various data types.
- ๐ Multi-agent Coordination: Emphasis on developing systems that can coordinate roles among agents.
- ๐ Safety and Trustworthiness: Validation of safety and trustworthiness in LLM applications is crucial.
๐ Background
The integration of artificial intelligence in healthcare, particularly in emergency medicine, is rapidly evolving. Large language models (LLMs) have emerged as powerful tools capable of assisting healthcare professionals in various tasks. However, the transition from non-agentic models to more sophisticated multi-agent systems presents both opportunities and challenges that need to be addressed through comprehensive research.
๐๏ธ Study
This scoping review aimed to synthesize existing literature on the application of LLMs and multi-agent systems in emergency medicine. The researchers conducted a thorough search across multiple databases, focusing on studies published between March 2021 and March 2026. A total of 35 studies were included, categorized into application, framework, and benchmark studies.
๐ Results
Among the 35 studies analyzed, 26 were application studies, indicating a strong interest in practical implementations of LLMs. The most common system type was non-agentic LLMs, with 25 studies focusing on these models. Evaluation methods primarily relied on expert comparisons and retrospective reviews, highlighting a gap in more dynamic evaluation approaches that reflect real-world emergency department workflows.
๐ Impact and Implications
The findings of this review underscore the potential of LLMs to enhance decision-making and documentation in emergency medicine. However, the limited focus on non-agentic models suggests a need for future research to explore more complex systems that can adapt to the fast-paced environment of emergency departments. By addressing these gaps, we can improve patient outcomes and streamline workflows in critical care settings.
๐ฎ Conclusion
This scoping review highlights the significant potential of LLMs in emergency medicine, particularly for task-level support. However, to fully realize their benefits, future research must prioritize workflow-aware designs, multimodal data integration, and robust evaluations of safety and trustworthiness. The journey from non-agentic models to multi-agent systems is just beginning, and the future looks promising for the integration of AI in emergency care.
๐ฌ Your comments
What are your thoughts on the integration of AI in emergency medicine? We would love to hear your insights! ๐ฌ Leave your comments below or connect with us on social media:
From non-agentic large language models to multi-agent systems in emergency medicine: a scoping review.
Abstract
OBJECTIVE: This study aimed to conduct a scoping review of studies on non-agentic large language models (LLMs), LLM-based agents, and multi-agent systems reported in emergency medicine, and to identify current research trends and major gaps by analyzing their clinical application scope, system structures, evaluation approaches, and input data characteristics.
METHODS: The Web of Science, Scopus, PubMed, and CINAHL databases were searched for literature published from March 8, 2021, to March 7, 2026. Among English full-text articles, studies addressing the application, evaluation, or benchmarking of non-agentic LLMs, LLM-based agents, or multi-agent systems in emergency medicine or the emergency department (ED) were included. Through reference tracking, 35 studies were analyzed.
RESULTS: Of the 35 included studies, 26 were application studies, 6 were framework studies, and 3 were benchmark studies. The studies were concentrated on a limited set of tasks, including triage, diagnostic and treatment decision support, and documentation. In terms of system type, non-agentic LLMs were the most common (n=25), followed by LLM-based agents (n=7) and multi-agent systems (n=3). Inputs were predominantly text-based, and evaluation mainly relied on expert comparison, retrospective record review, vignette-based comparison, and task-specific performance metrics. In contrast, workflow-level, prospective, and safety and trustworthiness-oriented evaluation were limited.
CONCLUSION: LLMs in emergency medicine have shown potential for task-level decision support and documentation. However, current literature remains focused on non-agentic LLM-based task support, while studies reflecting the dynamic workflow of real EDs remain limited. Future research should expand toward workflow-aware design, operational evaluation, multimodal data integration, multi-agent-based role coordination, and safety and trustworthiness validation.
Author: [‘Kim H’, ‘Jo S’, ‘Lim MH’, ‘Choi DH’]
Journal: Clin Exp Emerg Med
Citation: Kim H, et al. From non-agentic large language models to multi-agent systems in emergency medicine: a scoping review. From non-agentic large language models to multi-agent systems in emergency medicine: a scoping review. 2026; (unknown volume):(unknown pages). doi: 10.15441/ceem.26.136