Arabfard Masoud, Jeffery Alvin D, Moradian SeyedTayeb, He Hong-Gu, Pandian Vinciya, Vahedian-Azimi Amir
Chemical Injuries Research Center, Systems Biology and Poisonings Institute, Baqiyatallah University of Medical Sciences, Tehran, Iran.
Biomedical Informatics and Nursing, Vanderbilt University Medical Center. Nashville, TN, USA.
Intensive Crit Care Nurs. 2025 Sep 1:104213. doi: 10.1016/j.iccn.2025.104213.
Intensive Care Units (ICUs) present a high-stakes environment where timely decision-making is critical for managing patients with life-threatening conditions. The continuous influx of complex data often challenges clinicians, increasing the risk of errors. Artificial Intelligence (AI) offers transformative potential to enhance ICU care by supporting data analysis, decision-making, and workflow efficiency.
This review aims to explore current applications of AI in ICUs, assess their impact on clinical outcomes, workflow optimization, and ethical considerations, and propose an Integrated AI Framework for enhanced critical care delivery.
A literature search was conducted across PubMed, Scopus, and Web of Science, focusing on studies published between 2014 and 2024. The data were synthesized using an inductive thematic analysis approach to evaluate AI's impact on clinical outcomes and to identify key barriers to its integration.
AI has demonstrated significant advancements in ICU care, including early detection of sepsis, prediction of cardiac arrest, and workflow optimization through decision support systems. Predictive models reduced sepsis-related mortality by up to 20%, while workflow enhancements improved medication accuracy by 30% and reduced adverse events by 25%. Advanced techniques such as natural language processing (NLP), large language models (LLMs), and multimodal data integration have further streamlined ICU operations. However, challenges remain, including algorithmic bias, data privacy concerns, and integration barriers.
The Ideal Human Care in Green ICU model integrates advanced AI technologies with multidisciplinary collaboration to provide personalized, evidence-based, and patient-centered care. This model emphasizes ethical AI practices, transparency, and family engagement to ensure responsible implementation.
AI can transform ICU care by improving outcomes and workflows, but ethical, practical, and explainable challenges must be addressed through diverse, validated research.
AI integration in ICUs improves patient outcomes and workflows by enabling early detection and precise treatment, but ethical issues and real-world validation are crucial.
重症监护病房（ICU）是一个高风险环境，及时决策对于管理危及生命的患者至关重要。复杂数据的持续涌入常常给临床医生带来挑战，增加了出错风险。人工智能（AI）通过支持数据分析、决策制定和工作流程效率，为改善ICU护理提供了变革性潜力。
本综述旨在探讨AI在ICU中的当前应用，评估其对临床结果、工作流程优化和伦理考量的影响，并提出一个综合AI框架以加强重症护理服务。
在PubMed、Scopus和Web of Science上进行文献检索，重点关注2014年至2024年发表的研究。使用归纳主题分析方法对数据进行综合，以评估AI对临床结果的影响并确定其整合的关键障碍。
AI在ICU护理方面取得了显著进展，包括脓毒症的早期检测、心脏骤停的预测以及通过决策支持系统优化工作流程。预测模型将脓毒症相关死亡率降低了20%，而工作流程的改进使用药准确性提高了30%，不良事件减少了25%。自然语言处理（NLP）、大语言模型（LLMs）和多模态数据集成等先进技术进一步简化了ICU操作。然而，挑战依然存在，包括算法偏差、数据隐私问题和整合障碍。
绿色ICU中的理想人文护理模型将先进的AI技术与多学科协作相结合，以提供个性化、基于证据且以患者为中心的护理。该模型强调符合伦理的AI实践、透明度和患者家属参与，以确保负责任的实施。
AI可以通过改善结果和工作流程来改变ICU护理，但必须通过多样、经过验证的研究来解决伦理、实践和可解释性方面的挑战。
在ICU中整合AI通过实现早期检测和精确治疗来改善患者结果和工作流程，但伦理问题和实际验证至关重要。
