连续性肾脏替代治疗患者发生管路凝血预测模型的构建

Construction of a predictive model for in-circuit coagulation in patients on continuous renal replacement therapy

目的分析连续性肾脏替代治疗(continuous renal replacement therapy,CRRT)患者发生管路凝血的危险因素,构建并验证相关预测模型,为CRRT患者的治疗提供参考。方法采用目的抽样法选取2018年1月至2021年12月在江苏省人民医院行CRRT的413例患者作为研究对象,并设为训练集,根据管路凝血发生情况,将其分为凝血组(n=74)和未凝血组(n=339)。按照相同标准选取2022年1月至2022年6月在江苏省人民医院行CRRT的71例患者,设为验证集。收集患者的临床资料,使用单因素和Logistic回归分析CRRT患者发生管路凝血的危险因素,并据此建立列线图预测模型。结果本研究最终纳入血小板计数(OR=1.791)、凝血酶原时间(OR=2.543)、血流速度(OR=1.663)、治疗时间(OR=1.704)、护士工作年限(OR=2.277)5个危险因素构建预测CRRT患者发生管路凝血的列线图风险模型。模型验证结果显示:训练集和验证集的一致性指数分别为0.753和0.760,校正曲线均与理想曲线走势基本一致,受试者工作特征曲线的曲线下面积分别为0.732(95%CI:0.702~0.765)和0.751(95%CI:0.730~0.772),在7%~66%范围内,净获益值较高,表明该模型具有良好的预测能力。结论该模型预测效果较好,可帮助医务人员评估CRRT患者发生管路凝血的风险和制定适宜的干预方案。

Objective To explore the risk factors of in-circuit coagulation in patients on continuous renal replacement therapy(CRRT)and construct and verify the relevant prediction model so as to provide clinical references for CRRT.Methods From January 2018 to December 2021,objective sampling was employed for selecting 413 patients on CRRT as research subjects and training set.According to the occurrence of coagulation in pipeline,they were assigned into two groups of coagulation(n=74)and non-coagulation(n=339).According to the same criteria,71 patients on CRRT from January 2022 to June 2022 were designated as validation set.Clinical data were collected and univariate and Logistic regression employed for examining the risk factors of in-circuit coagulation in CRRT patients.Based upon the above,a nomogram prediction model was established.Results Five risk factors of platelet count(OR=1.791),prothrombin time(OR=2.543),blood flow velocity(OR=1.663),treatment time(OR=1.704)and practicing years of nurses(OR=2.277)were finally included for constructing a nomograph risk model for predicting the occurrence of pipeline coagulation in CRRT patients.The model verification results showed that C-index of training set and verification set was 0.753 and 0.760 respectively and the trend of correction curve was basically consistent with the ideal curve.The area under the ROC curve(AUC)was 0.732(95%CI:0.702~0.765)and 0.751(95%CI:0.730~0.772).In the range of 7%~66%,net benefit value was high,indicating that the model had an excellent prediction capability.Conclusion The prediction effect of this model is excellent.It helps clinicians to assess the risk of in-circuit coagulation and formulate appropriate intervention plans in CRRT patients.