Zhong Lei, Min Jie, Zhang Jinyu, Hu Beiping, Qian Caihua
Department of Intensive Care Unit, Huzhou Central Hospital, Fifth School of Clinical Medicine of Zhejiang Chinese Medical University, Huzhou, Zhejiang Province 313000, China.
Huzhou Central Hospital, Affiliated Central Hospital of Huzhou University, Huzhou, Zhejiang Province 313000, China.
iScience. 2024 Jun 27;27(8):110397. doi: 10.1016/j.isci.2024.110397. eCollection 2024 Aug 16.
Continuous renal replacement therapy (CRRT) is a commonly utilized treatment modality for individuals experiencing severe acute kidney injury (AKI). The objective of this research was to construct and assess prognostic models for the timely discontinuation of CRRT in critically ill AKI patients receiving this intervention. Data were collected retrospectively from the MIMIC-IV database ( = 758) for model development and from the intensive care unit (ICU) of Huzhou Central Hospital ( = 320) for model validation. Nine machine learning models were developed by utilizing LASSO regression to select features. In the training set, all models demonstrated an AUROC exceeding 0.75. In the validation set, the XGBoost model exhibited the highest AUROC of 0.798, leading to its selection as the optimal model for the development of an online calculator for clinical applications. The XGBoost model demonstrates significant predictive capabilities in determining the discontinuation of CRRT.
持续肾脏替代疗法（CRRT）是治疗严重急性肾损伤（AKI）患者常用的治疗方式。本研究的目的是构建并评估用于接受该治疗的危重症AKI患者及时停止CRRT的预后模型。回顾性收集MIMIC-IV数据库中的数据（n = 758）用于模型开发，并收集湖州市中心医院重症监护病房（ICU）的数据（n = 320）用于模型验证。利用LASSO回归选择特征，开发了9种机器学习模型。在训练集中，所有模型的曲线下面积（AUROC）均超过0.75。在验证集中，XGBoost模型的AUROC最高，为0.798，因此被选为开发临床应用在线计算器的最佳模型。XGBoost模型在确定CRRT停止方面具有显著的预测能力。
