急性白血病患儿经外周静脉穿刺的中心静脉导管置管后发生静脉血栓栓塞的影响因素分析及预测模型构建  

Analysis of influencing factors and construction of prediction model of venous thromboembolism in children with acute leukemia after peripherally inserted central venous catheter placement

作　　者：张梦可 李欢欢[1] 向明丽[1] 王晓彤 于雨萍 ZHANG Mengke;LI Huanhuan;XIANG Mingli;WANG Xiaotong;YU Yuping(Department of Pediatric Internal Medicine,the First Affiliated Hospital of Zhengzhou University,Zhengzhou 450000,He’nan,China)

机构地区：[1]郑州大学第一附属医院小儿内科,郑州450000

出　　处：《癌症进展》2025年第1期63-67,共5页Oncology Progress

摘　　要：目的分析急性白血病(AL)患儿经外周静脉穿刺的中心静脉导管(PICC)置管后发生静脉血栓栓塞(VTE)的影响因素,并构建基于随机森林算法的风险预测模型。方法收集120例行PICC置管的AL患儿的病历资料,根据是否发生VTE分为VTE组和非VTE组,比较两组患儿的临床特征。采用多因素Logistic回归模型分析AL患儿PICC置管后发生VTE的影响因素,并构建随机森林模型。采用受试者工作特征(ROC)曲线及曲线下面积(AUC)分析两种模型对AL患儿PICC置管后发生VTE的预测价值,并采用Delong检验进行不同模型预测价值的差异性分析。结果120例AL患儿PICC置管后,发生VTE 35例(29.17%),未发生VTE 85例(70.83%)。两组患儿穿刺部位、感染情况、穿刺次数、导管留置时间、导管相关并发症发生情况、抗凝药物使用情况、D-二聚体水平比较,差异均有统计学意义(P﹤0.01)。多因素Logistic回归分析结果显示,穿刺部位、穿刺次数、导管留置时间、导管相关并发症、抗凝药物使用情况、D-二聚体水平均为AL患儿PICC置管后发生VTE的影响因素(P﹤0.05)。根据Logistic回归模型中筛选出来的影响因素构建随机森林模型并进行分析,结果显示,当决策树的数目达到18时,该模型的错误率最低;AL患儿PICC置管后发生VTE影响因素的重要性排序依次为穿刺次数、导管留置时间、抗凝药物使用情况、穿刺部位、导管相关并发症、D-二聚体水平。随机森林模型预测AL患儿PICC置管后发生VTE的AUC为0.860(95%CI:0.785~0.916),高于Logistic回归模型的0.775(95%CI:0.690~0.846),差异有统计学意义(P﹤0.05)。结论穿刺部位、穿刺次数、导管留置时间、导管相关并发症、抗凝药物使用情况以及D-二聚体水平均为AL患儿PICC置管后发生VTE的影响因素。本研究构建的预测AL患儿PICC置管后发生VTE的随机森林模型具有较高的准确性,有助于医务人员更准确地评估患儿发生VTE�Objective To analyze the influencing factors of venous thromboembolism(VTE)in children with acute leukemia(AL)after peripherally inserted central venous catheter(PICC)placement,and construct a risk prediction model based on random forest algorithm.Method The medical records of 120 AL children with PICC placement were collected and divided into VTE group and non-VTE group according to whether VTE occurred.The clinical features of the two groups were compared.Multivariate Logistic regression model was used to analyze the influencing factors of VTE after PICC placement in children with AL,and a random forest model was constructed.Receiver operating characteristic(ROC)curve and area under the curve(AUC)were used to analyze the predictive value of two models for VTE in children with AL after PICC placement,and Delong test was used to analyze the difference of predictive value of different models.Result After PICC placement in 120 children with AL,35 cases(29.17%)developed VTE and 85 cases(70.83%)did not.There were significant differences in puncture site,infection,puncture times,catheter indwelling time,catheter-related complications,use of anticoagulant drugs and D-dimer levels between the two groups(P<0.01).Multivariate Logistic regression analysis showed that puncture site,puncture times,catheter indwelling time,catheter-related complications,use of anticoagulant drugs and D-dimer level were influencing factors of VTE in children with AL after PICC placement(P<0.05).According to influencing factors screened out in the Logistic regression model,a random forest model was constructed and analyzed.The results showed that when the number of decision trees reached 18,the model had the lowest error rate.The importance of influencing factors of VTE after PICC placement in children with AL was ranked in order of puncture times,catheter indwelling time,use of anticoagulant drugs,puncture site,catheter-related complications and D-dimer level.The AUC of the random forest model for predicting VTE after PICC placement in childre

关 键 词：随机森林算法 急性白血病 经外周静脉穿刺的中心静脉导管 静脉血栓栓塞 预测模型 

分 类 号：R733.71[医药卫生—肿瘤]