Abstract:

Objective To conduct a scoping review of studies on the application of artificial intelligence-based clinical decision support systems（AI-CDSS） in the ICU domestically and internationally，providing references for future clinical nursing applications and research directions. Methods Computerized searches were performed across 9 Chinese and English databases，with the search period spanning from the inception of each database to May 2025. Studies related to the application of AI-CDSS in the ICU were screened，and the included literature was summarized and analyzed based on the scoping review framework. Results A total of 14 articles from 6 countries were included. AI-CDSS in the ICU primarily utilized machine learning，deep learning，and natural language processing technologies，extracting information from both structured and unstructured ICU data. These systems were mainly applied to patient groups such as those with infection risks，mechanical ventilation，comorbid coronary heart disease，and orthopedic trauma. Core functions included prediction of complications and disease outcomes，mortality risk prediction，decision recommendations，and user interaction. The systems demonstrated high recognition accuracy and strong transfer generalization capabilities in the ICU. Nurses reported high execution，adoption，satisfaction，and acceptance rates. Conclusion AI-CDSS can enhance the prediction of mortality and complication risks，improve predictive nursing levels，elevate the quality of care，and optimize patient outcomes in the ICU. Future efforts should focus on developing integrated AI-CDSS that incorporate multi-device data integration and strengthen human-machine collaboration to promote more efficient and reliable application in diverse ICU scenarios.

Key words: Intensive Care Unit, Artificial Intelligence, Clinical Decision Support System, Scoping Review, Critical Care

摘要：

目的 对国内外基于人工智能的临床决策支持系统在ICU中应用的相关研究进行范围综述，为未来临床护理应用和研究方向提供参考依据。方法 计算机检索9个中英文数据库，检索时限为建库至2025年5月。筛选有关基于人工智能的临床决策支持系统在ICU中应用的研究，基于范围综述的研究框架对纳入文献进行汇总和分析。结果 最终纳入14篇文献，来自6个国家。ICU中基于人工智能的临床决策支持系统主要采用机器学习、深度学习和自然语言处理技术，并从ICU的结构化和非结构化数据中提取信息，主要应用于存在感染风险、机械通气、合并冠心病、骨科创伤等患者群体，核心功能包括并发症与疾病转归预测、死亡风险预测、决策推荐以及用户交互，在ICU中具有良好的识别准确率和迁移泛化性能，护士对其具有较好的执行率、采纳率、满意度和接受度。结论 应用基于人工智能的临床决策支持系统可提升ICU患者死亡和并发症风险预测能力、预见性护理水平，提升护理质量，改善患者结局。未来应重点开发多设备数据集成、综合型人工智能临床决策支持系统，加强人机协同，推动系统更高效、可靠地应用于多样化ICU场景。

关键词: 重症监护病房, 人工智能, 临床决策支持系统, 范围综述, 危重症护理