经连续性肾脏替代治疗的横纹肌溶解致急性肾损伤患者死亡预测模型的构建与验证

Development and validation of prediction models for death in patients with rhabdomyolysis-induced acute kidney injury treated with continuous renal replacement therapy

目的识别经连续性肾脏替代治疗(continuous renal replacement therapy,CRRT)的横纹肌溶解致急性肾损伤患者死亡的危险因素,开发临床预测模型并验证其效能。方法从MIMIC-IV 2.2数据库中提取2008年-2019年横纹肌溶解致急性肾损伤且经CRRT治疗患者的临床资料及预后信息。将入组患者按照7∶3分为训练集及测试集,在训练集中通过LASSO回归、随机森林及极端梯度提升法(extreme gradient boosting,XGBoost)识别影响患者28 d死亡的危险因素并建立logistic回归模型、随机森林模型、支持向量机模型与XGBoost模型,在测试集中计算上述预测模型的准确率及受试者操作特征曲线下面积。结果共纳入175例患者,入重症监护病房时测的乳酸、年龄、急性生理评分Ⅲ、血红蛋白、平均动脉压及体质量指数为LASSO回归、随机森林与XGBoost共同识别出的影响患者生存最重要的6个危险因素。通过以上危险因素构建的logistic回归模型、随机森林模型、支持向量机模型与XGBoost模型在测试集中的预测准确率分别为0.75、0.79、0.79与0.81,受试者操作特征曲线下面积分别为0.82、0.85、0.87与0.87。结论通过入重症监护病房时测的乳酸、年龄、急性生理评分Ⅲ、血红蛋白、平均动脉压及体质量指数6个危险因素构建的XGBoost模型的临床预测效能较好,有利于医生临床决策。

Objective To identify risk factors for death in patients with rhabdomyolysis-induced acute kidney injury(RI-AKI)treated with continuous renal replacement therapy(CRRT),then to develop and validate the efficacy of prediction models based on these risk factors.Methods lsClinical data and prognostic information of patients with RI-AKI requiring CRRT from 2008 to 2019 were extracted from the MIMIC-IV 2.2 database.The enrolled patients were divided into a training set and a test set at a ratio of 7:3.LASSO regression,random forest(RF)and extreme gradient boosting(XGBoost)were used to identify the risk factors affecting patients'28-day survival in the training set,then to develop logistic model,RF model,support vector machine(SVM)model and XGBoost model.The accuracy of above prediction models and the area under the receiver operating characteristic curve(AUC)were calculated in the test set.ResultsA total of 175 patients were included.Lactic acid,age,Acute Physiology Score Il,hemoglobin,mean arterial pressure and body mass index measured at intensive care unit admission were identified as the six risk factors affecting 28-day survival of enrolled patients by LASSO regression,RF and XGBoost.The accuracy of the logistic model,RF model,SVM model and XGBoost model in the test set was 0.75,0.79,0.79 and 0.81,with the AUC of 0.82,0.85,0.87 and 0.87,respectively.Conclusion The XGBoost model,incorporating six risk factors including lactic acid,age,Acute Physiology ScoreⅢ,hemoglobin,mean arterial pressure,and body mass index assessed at the time of admission to the intensive care unit,demonstrates superior clinical predictive performance,thereby enhancing the clinical decision-making process for healthcare professionals.