Xu Xianrong, Chen Mou, Chen Lvjing, Huang Kaixing, Cao Shiqi, Gao Wenwen, Liu Kang, Wu Buyun, Mao Huijuan
Department of Nephrology, Jiangsu Province Hospital, The First Affiliated Hospital of Nanjing Medical University, Guangzhou Road, #300, Nanjing, 210029, PR China.
BMC Nephrol. 2025 Jul 24;26(1):414. doi: 10.1186/s12882-025-04206-z.
To construct an intelligent Continuous Kidney Replacement Therapy nursing feedback model to predict extracorporeal coagulation risk and support timely clinical decisions.
Analyzed 1,354 treatment records from 143 patients to identify new data features relevant to Continuous Kidney Replacement Therapy nursing. The data was preprocessed, and new variables were derived to serve as model inputs. A hybrid machine learning algorithm was applied to predict the timing of Continuous Kidney Replacement Therapy initiation for patients with acute kidney injury. Extensive numerical experiments were conducted to optimize model parameters and evaluate performance, ensuring high accuracy, stability, and superior AUC values for reliable predictions.
Univariate analysis identified eight factors significantly affecting coagulation risk, including treatment mode, anticoagulation method, blood pump stoppage, and insufficient blood flow ( < 0.001). Logistic regression analysis indicated that treatment mode and anticoagulation method were key factors influencing extracorporeal coagulation during Continuous Kidney Replacement Therapy, with the highest regression coefficient observed for heparin-free anticoagulation (β = 2.209). The prediction model achieved an AUC of 0.87 ( < 0.001) and an accuracy rate of 99.21%, significantly outperforming the performance of other models.
The intelligent Continuous Kidney Replacement Therapy nursing feedback model improves prediction accuracy while reducing redundant information. This model helps mitigate the risk of missing urgent conditions in patients under limited healthcare resources and lowers the frequency of extracorporeal coagulation events during Continuous Kidney Replacement Therapy.
Not applicable.
构建智能连续性肾脏替代治疗护理反馈模型，以预测体外凝血风险并支持及时的临床决策。
分析143例患者的1354份治疗记录，以识别与连续性肾脏替代治疗护理相关的新数据特征。对数据进行预处理，并导出新变量作为模型输入。应用混合机器学习算法预测急性肾损伤患者开始连续性肾脏替代治疗的时机。进行了广泛的数值实验以优化模型参数并评估性能，确保模型具有高准确性、稳定性和卓越的AUC值，以进行可靠的预测。
单因素分析确定了八个显著影响凝血风险的因素，包括治疗模式、抗凝方法、血泵停止和血流量不足（<0.001）。逻辑回归分析表明，治疗模式和抗凝方法是影响连续性肾脏替代治疗期间体外凝血的关键因素，无肝素抗凝的回归系数最高（β = 2.209）。预测模型的AUC为0.87（<0.001），准确率为99.21%，显著优于其他模型的性能。
智能连续性肾脏替代治疗护理反馈模型提高了预测准确性，同时减少了冗余信息。该模型有助于在医疗资源有限的情况下降低患者漏诊紧急情况的风险，并降低连续性肾脏替代治疗期间体外凝血事件的发生频率。
不适用。
