基于SMOTE算法的下肢创伤性骨折术后下肢深静脉血栓形成风险预警模型的构建  

Construction of An Early Warning Model for the Risk of Lower Limb Deep Vein Thrombosis after Lower Limb Traumatic Fracture Surgery Based on SMOTE Algorithm

作　　者：石颖花[1] 黄子珊[1] 蓝宁路 SHI Yinghua;HUANG Zishan;LAN Ninglu(Jiangbin Hospital of Guangxi Zhuang Autonomous Region,Nanning City,Guangxi 530021)

机构地区：[1]广西壮族自治区江滨医院,广西南宁市530021

出　　处：《医学理论与实践》2023年第19期3259-3262,3255,共5页The Journal of Medical Theory and Practice

摘　　要：目的:探索下肢创伤性骨折术后下肢深静脉血栓形成(DVT)的独立危险因素,并基于SMOTE算法建立个体化预警模型。方法:选取我院2018年1月—2021年6月收治的398例下肢创伤性骨折患者作为研究对象,根据患者术后是否发生DVT,将患者分为DVT组(62例)和非DVT组(336例)。收集两组患者临床资料,采用单因素和二元Logistic回归分析方法筛选下肢DVT的独立危险因素,并建立Logistic回归模型(P_(1)),同时基于SMOTE算法改进数据集,构建改进数据集的预警模型(P_(2)),并对比模型的预测效能。结果:患者年龄增加、术后第1天D二聚体水平升高、手术等待时间≥6d、合并冠心病及多处骨折是下肢创伤性骨折患者术后发生DVT的独立危险因素(P<0.05)。概率预测模型P_(1)=1/[1+e^(-(-13.595+0.069X1+1.106X2+1.302X3+1.000X4+1.236X5))],P_(2)=1/[1+e^(-(-10.161+0.075X1+0.808X2+1.216X3+0.787X4+0.513X5))]。P_(2)模型的决定系数和ROC曲线下面积均明显高于P_(1)模型。结论:患者年龄增加、术后第1天D二聚体水平升高、手术等待时间≥6d、合并冠心病及多处骨折与下肢创伤性骨折患者术后发生DVT密切相关,基于此建立的个体化预警模型有利于下肢DVT高危患者的识别和防治。Objective:To explore the independent risk factors of lower limb deep vein thrombosis(DVT)after lower limb traumatic fracture surgery and to establish an individualized early warning model based on SMOTE(synthetic minority over-sampling technique)algorithm.Methods:398 patients with lower limb traumatic fractures admitted to our hospital from January 2018 to June 2021 were selected as study subjects,and patients were divided into DVT group(62 cases)and non-DVT group(336 cases)according to whether they had DVT after surgery.The clinical data of patients in both groups were collected,and the independent risk factors for DVT of lower limbs were screened by univariate and binary Logistic regression analysis,and a Logistic regression model(P_(1))was established,while an early warning model(P_(2))with improved data set was constructed based on the SMOTE algorithm,and the predictive efficacy of the model was compared.Results:Increasing patient age,elevated D-dimer levels on postoperative day 1,waiting time for surgery≥6 days,combined coronary artery disease and multiple fractures were independent risk factors for the development of DVT in patients with traumatic fractures of the lower extremities after surgery(P<0.05).Probabilistic prediction model P_(1)=1/[1+e^(-(-13.595+0.069X1+1.106X2+1.302X3+1.000X4+1.236X5))],P_(2)=1/[1+e^(-(-10.161+0.075X1+0.808X2+1.216X3+0.787X4+0.513X5))].The coefficient of determination and area under the ROC curve of the P_(2) model were significantly higher than those of the P_(1) model.Conclusion:Increased patient age,elevated D-dimer levels on postoperative day 1,waiting time for surgery≥6 days,combined coronary artery disease and multiple fractures were closely associated with the occurrence of DVT in patients with traumatic fractures of the lower extremities,and the individualized early warning model based on this was beneficial for the identification and prevention of patients at high risk of DVT of the lower extremities.

关 键 词：下肢创伤性骨折 深静脉血栓形成 过抽样 预警模型 

分 类 号：R493[医药卫生—康复医学]