基于机器学习的心血管手术患者术后谵妄预测模型的范围综述

A Scoping Review of Machine Learning-Based Prediction Models for Postoperative Delirium in Cardiovascular Surgery Patients

摘要: 本研究通过系统分析应用机器学习预测心血管患者术后谵妄的相关文献，对其预测模型的性能进行了评估。本研究共纳入中、英文文献9篇，样本量87~4476。分析结果显示，随机森林模型在众多机器学习算法中预测性能最优，其AUC值达0.92。术后谵妄预测因子覆盖患者基本资料、术前、术中及术后等多方面因素。机器学习预测模型对术后谵妄的发生具有良好的预测价值，然而现有研究存在样本量小，缺乏外部验证及模型可解释性不足等局限性。未来研究需扩大样本量，选择合理预测特征，增强模型的可解释性和临床应用价值。

Abstract: This study systematically analyzed the literature on the application of machine learning in predicting postoperative delirium in cardiovascular patients and evaluated the performance of the prediction models. A total of 9 Chinese and English articles were included, with sample sizes ranging from 87 to 4476. The analysis showed that among various machine learning algorithms, the random forest model had the best prediction performance, with an AUC value of 0.92. The predictors for postoperative delirium covered multiple factors, including patient basic information, preoperative, intraoperative, and postoperative factors. Machine learning prediction models have good predictive value for the occurrence of postoperative delirium. However, existing studies have limitations such as small sample sizes, lack of external validation, and insufficient model interpretability. Future research needs to increase the sample size, select appropriate prediction features, and enhance the interpretability and clinical application value of the models.