Mbadjeu Hondjeu Arnaud Romeo, Zhao Zi Ying, Newton Luka, Ajenkar Anass, Hladkowicz Emily, Ladha Karim, Wijeysundera Duminda N, McIsaac Daniel I
Department of Anesthesiology and Pain Medicine, University of Ottawa, Ottawa, ON, Canada.
The Ottawa Hospital, Ottawa, ON, Canada.
Can J Anaesth. 2025 Jun 9. doi: 10.1007/s12630-025-02980-w.
Large language models (LLMs) are a subset of artificial intelligence (AI) and linguistics designed to help computers understand and analyze human language. Clinical applications of LLMs have recently been recognised for their potential enhanced analytic capacity. Availability and performance of LLMs are expected to increase substantially over time with a significant impact on patient care and health care provider workflow. Despite increasing recognition of LLMs, insights on the utilities, associated benefits and limitations are scarce among perioperative clinicians. In this narrative review, we delve into the functionalities and prospects of existing LLMs and their clinical application in perioperative medicine. Furthermore, we summarize challenges and constraints that must be addressed to fully realize the potential of LLMs.
We searched MEDLINE, Google Scholar, and PubMed® databases for articles referencing LLMs in perioperative care.
We found that in the perioperative setting (from surgical diagnosis to discharge postoperatively), LLMs have the potential to improve the efficiency and accuracy of health care delivery by extracting and summarizing clinical data, making recommendations on the basis of these findings, as well as addressing patient queries. Moreover, LLMs can be used for clinical decision-making support, surveillance tools, predictive modelling, and enhancement of medical research and education.
The integration of LLMs into perioperative medicine presents a significant opportunity to enhance patient care, clinical decision-making, and operational efficiency. These models can streamline processes, provide personalized patient education, and offer robust decision support. Nevertheless, their clinical implementation requires addressing several key challenges, including managing hallucinations, ensuring data security, and mitigating inherent biases. If these challenges are met, LLMs can revolutionize perioperative practice, improving both patient outcomes and clinician workflow.
大语言模型（LLMs）是人工智能（AI）和语言学的一个子集，旨在帮助计算机理解和分析人类语言。大语言模型的临床应用最近因其潜在的增强分析能力而受到认可。随着时间的推移，大语言模型的可用性和性能预计将大幅提高，对患者护理和医疗保健提供者的工作流程产生重大影响。尽管对大语言模型的认识不断提高，但围手术期临床医生对其效用、相关益处和局限性的见解却很少。在这篇叙述性综述中，我们深入探讨了现有大语言模型的功能和前景及其在围手术期医学中的临床应用。此外，我们总结了要充分实现大语言模型的潜力必须解决的挑战和限制。
我们在MEDLINE、谷歌学术和PubMed®数据库中搜索了引用围手术期护理中大语言模型的文章。
我们发现，在围手术期（从手术诊断到术后出院），大语言模型有潜力通过提取和总结临床数据、根据这些发现提出建议以及回答患者疑问来提高医疗服务的效率和准确性。此外，大语言模型可用于临床决策支持、监测工具、预测建模以及加强医学研究和教育。
将大语言模型整合到围手术期医学中为提高患者护理、临床决策和运营效率提供了一个重要机会。这些模型可以简化流程，提供个性化的患者教育，并提供强大的决策支持。然而，它们的临床应用需要应对几个关键挑战，包括管理幻觉、确保数据安全和减轻固有偏差。如果这些挑战能够得到解决，大语言模型可以彻底改变围手术期实践，改善患者结局和临床医生的工作流程。
