摘要
目的探讨建立脓毒症相关急性肾损伤患儿持续肾替代治疗停机时机选择的临床预测模型并进行验证。方法选取2020年1月至2021年12月入住广东医科大学附属东莞儿童医院PICU的82例脓毒性相关急性肾损伤(septic associated acute kidney injury,SA-AKI)并行连续性肾替代治疗(continuous renal replacement therapy,CRRT)的患儿,在CRRT治疗前及停止治疗时,结合床边肾脏超声测定其肾叶间动脉阻力指数(renal resistive index,RRI)和肾血流半定量评分(semiquantitative power doppler ultrasound,PDU)、尿量、平均动脉压,利尿剂、血管活性药使用及相关生化指标,进行二元Logistic回归分析,建立临床预测模型并对模型进行Bootstrap重复自抽样内部验证。结果按CRRT停机是否成功分成两组,组间比较提示CRRT开始时间、RRI、PDU、平均动脉压、β2-微球蛋白(β2-microglobulin,β2-MG)及尿量有统计学意义(P<0.05),进而纳入二元Logistic回归得出预测方程,其ROC曲线的AUC=0.877(95%CI:0.786~0.939),敏感度0.96,特异度0.67,使用Bootstrap重复自抽样进行内部验证提示模型具有良好的区分度及校准度,临床决策曲线也提示有较好的临床适用度。结论由RRI、β2-MG、尿量及CRRT开始时间整合的预测模型有助于SA-AKI患儿CRRT治疗停机时机的选择。
Objective To establish and validate the predictive model of the optimal discontinue timing of renal replacement therapy in sepsis associated with acute kidney injury in children.Methods This observational cohort study enrolled 82 children treated with continuous renal replacement therapy(CRRT)for septic associated acute kidney injury(SA-AKI)from January 2020 to December 2021 in the department of pediatric intensive care medicine(PICU)in Dongguan Children’s Hospital affiliated to Guangdong Medical University.When the patients were treated with CRRT,we used renal ultrasound to measure renal resistive index(RRI)and semiquantitative power doppler ultrasound(PDU),and quickly record the urine volume,mean arterial pressure(MAP),diuretics,the use of vasoactive drugs and renal biochemical indicators before and after CRRT treatment.These indicators were incorporated into the binary logistic regression model to establish a clinical prediction model and internal validation was performed with Bootstrad repeated sampling.Results All patients were divided into two groups according to the CRRT treatment was ended successful or not.The comparison between the two groups showed that CRRT initiation time,RRI,PDU,MAP,β2-MG and urine volume were statistical significance(P<0.05).Then they were put into the binomial Logistic regret analysis and the prediction model was obtained.The ROC curve of the prediction model were AUC=0.877(95%CI:0.786~0.939),sensitivity 0.96,specificity 0.67.Internal validation of the model using Bootstrap repeated self-sampling suggested that the model had good differentiation and calibration,and its clinical decision curve also suggested that the model had good clinical fitness.Conclusions The prediction model integrating RRI,β2-MG,urine volume and the CRRT initiation time,could help to determine the discontinue timing of CRRT treatment of children with SA-AKI.
作者
赖志君
杨文海
李燕
马可泽
Lai Zhijun;Yang Wenhai;LiYan;Ma Keze(Department of Pediatric Intensive Care Unit,Dongguan Children's Hospital Affiliatedto Guangdong Medical University,Dongguan,China)
基金
东莞市社会科技发展项目面上项目(项目编号:20211800900672)
关键词
脓毒症
急性肾损伤
持续肾替代治疗
停机时机
预测模型
Sepsis
Acut kidney injury
Continuous renal replacement therapy
Discontinue timing
Prediction model
作者简介
通信作者:马可泽,E-mail:azema2005@163.com
