Ahmadzadeh Bahareh, Patey Christopher, Norman Paul, Farrell Alison, Knight John, Czarnuch Stephen, Asghari Shabnam
PhD Candidate in Clinical Epidemiology, Center for Rural Health Studies, Faculty of Medicine, Memorial University of Newfoundland, St John's, NL, Canada.
Assistant Professor - Family & Emergency Medicine, Discipline of Family Medicine, Faculty of Medicine, Memorial University of Newfoundland, St John's, NL, Canada.
J Emerg Med. 2025 Aug;75:174-187. doi: 10.1016/j.jemermed.2025.05.031. Epub 2025 Jun 10.
Overcrowding and long wait times in emergency departments (EDs) remain global challenges that negatively affect patient outcomes and staff satisfaction. As an emerging technology, artificial intelligence (AI) offers the potential to optimize ED operations and reduce wait times.
Establish a strategy to evaluate AI modeling as it relates to utilizing AI based strategies for ED flow.
We searched Embase, MEDLINE, CINAHL, and Scopus for English-language studies published from January 1, 1946, to August 17, 2023, and we will update our search to ensure currency. The ROBINS-I tool assessed study quality, while PROBAST examined the risk of bias and applicability.
Out of 17,569 screened studies, 65 full-text articles were evaluated for eligibility, with 16 quantitative observational studies meeting inclusion criteria. The best-performing algorithms included regression-based methods (n = 2), traditional single-model machine learning (n = 8), neural networks/deep learning (n = 3), natural language processing (n = 1), and ensemble methods (n = 2). None of the studies examined AI's impact in a real ED setting, though four simulations reported wait-time reductions ranging from 7 to 43.2 minutes.
AI integration in ED is still in its infancy. Our review found no real-world ED implementation studies, and most of the existing research lacked involvement from ED experts. This gap highlights the lack of insight into AI's practical impact. Future reviews and research must clarify these dimensions, guiding AI's effective, collaborative adoption in ED workflows.
急诊科的过度拥挤和长时间等待仍是全球性挑战，对患者治疗结果和工作人员满意度产生负面影响。作为一项新兴技术，人工智能有潜力优化急诊科运营并减少等待时间。
制定一项策略，以评估与利用基于人工智能的策略优化急诊科流程相关的人工智能建模。
我们检索了Embase、MEDLINE、CINAHL和Scopus数据库，查找1946年1月1日至2023年8月17日发表的英文研究，并将更新检索以确保时效性。使用ROBINS-I工具评估研究质量，同时使用PROBAST检查偏倚风险和适用性。
在17569项筛选出的研究中，对65篇全文进行了资格评估，其中16项定量观察性研究符合纳入标准。表现最佳的算法包括基于回归的方法（n = 2）、传统单模型机器学习（n = 8）、神经网络/深度学习（n = 3）、自然语言处理（n = 1）和集成方法（n = 2）。尽管有四项模拟研究报告等待时间减少了7至43.2分钟，但没有一项研究考察了人工智能在实际急诊科环境中的影响。
人工智能在急诊科的整合仍处于起步阶段。我们的综述未发现实际急诊科实施研究，且大多数现有研究缺乏急诊科专家的参与。这一差距凸显了对人工智能实际影响缺乏了解。未来的综述和研究必须阐明这些方面，指导人工智能在急诊科工作流程中的有效、协同应用。
