出血性卒中患者下肢深静脉血栓形成风险预测模型的构建与验证

Development and verification of the risk prediction model of deep venous thrombosis in the lower extremities of patients with acute hemorrhagic stroke

目的构建下肢深静脉血栓形成(DVT)的风险预测模型,以评估和预防出血性卒中患者的下肢DVT。方法采用便利抽样法收集2018年7月至2020年7月天津医科大学总医院神经外科重症监护病房收治的547例出血性卒中患者的临床资料。547例患者分为建模组(447例)和外部验证组(100例),采用彩色多普勒超声确诊下肢DVT。采用单因素和多因素logistic回归分析法判断发生下肢DVT的危险因素,并构建下肢DVT风险预测模型;绘制受试者工作特征曲线并计算曲线下面积(AUC)以及采用Hosmer-Lemeshow拟合优度检验评价风险预测模型的效能;应用外部验证法检验模型的灵敏度、特异度和一致性。结果建模组纳入的447例患者中,共112例(25.1%)确诊为下肢DVT。单因素分析结果显示,年龄、格拉斯哥昏迷评分、全身麻醉手术、镇静镇痛药物、降温毯物理降温、糖尿病史、高血压病史、下肢肌力分级及血浆D-二聚体水平是出血性卒中患者发生下肢DVT的影响因素(均P<0.05)。进一步行多因素logistic回归分析结果显示,年龄(OR=1.063,95%CI:1.036~1.092,P<0.001)、镇静镇痛药物(OR=5.115,95%CI:2.620~9.986,P<0.001)、降温毯物理降温(OR=34.991,95%CI:14.009~87.396,P<0.001)、高血压病史(OR=2.671,95%CI:1.275~5.594,P=0.009)、血浆D-二聚体水平(OR=7.026,95%CI:3.324~14.851,P<0.001)是出血性卒中患者发生下肢DVT的危险因素。风险预测模型的检验结果显示,P=0.648,AUC=0.912,约登指数最大值为0.714,对应的灵敏度为91.4%,特异度为80.0%,截断值为0.226。外部验证组100例患者的数据分析结果显示,该模型的灵敏度为91.4%,特异度为80.0%,一致性指数为0.84。结论基于临床参数所构建的下肢DVT风险预测模型简单、便捷,具有较高的特异性,可为出血性卒中患者下肢DVT风险的预防、评估及诊疗提供参考。

Objective To establish the risk prediction model of deep venous thrombosis(DVT)in the lower extremities and to evaluate and prevent the lower extremity DVT of patients with hemorrhagic stroke.Methods Using the convenience sampling method,we collected the clinical data of 547 patients with hemorrhagic stroke who were admitted to the Intensive Care Unit,Department of Neurosurgery,General Hospital of Tianjin Medical University from July 2018 to July 2020.Among 547 patients,447 were divided into the modeling group and 100 cases into the external verification group.The lower extremity DVT was diagnosed with the color Doppler ultrasound.Univariate and multivariate logistic regression analysis were applied to explore the risk factors of lower extremity DVT and construct the lower extremity DVT risk prediction model.The receiver operating characteristic curve was drawn and the area under the curve(AUC)was calculated.The goodness of fit test of Hosmer-Lemeshow was used to evaluate the effectiveness of the risk prediction model.The external validation data were used to test the sensitivity,specificity and consistency rate of the risk prediction model.Results Among the 447 cases in the modeling group,112 cases(25.1%)developed the lower extremity DVT.Univariate logistic regression analysis showed that age,general anesthetic operation,sedative and analgesic drugs,physical cooling with cooling blanket,GCS,history of diabetes,history of hypertension,muscle strength classification of lower limbs and plasma D-Dimer level were the influencing factors of DVT in patients with hemorrhagic stroke(all P<0.05).Further multivariate logistic regression analysis showed that age(OR=1.063,95%CI:1.036-1.092,P<0.001),use of sedative and analgesic drugs(OR=5.115,95%CI:2.620-9.986,P<0.001),physical cooling with cooling blanket(OR=34.991,95%CI:14.009-87.396,P<0.001),history of hypertension(OR=2.671,95%CI:1.275-5.594,P=0.009)and plasma D-Dimer level(OR=7.026,95%CI:3.324-14.851,P<0.001)were risk factors of lower extremity DVT in patients with acute hemorrhagic stroke.The test results of risk prediction model showed that the P value was 0.648,AUC value was 0.912,maximum value of Yoden index was 0.714,corresponding sensitivity was 91.4%,specificity was 80.0%,and cut-off value was 0.226.The results of the external validation showed that the predictive sensitivity was 91.4%,the specificity was 80.0%,and the consistency index was 0.84.Conclusions The lower extremity DVT risk prediction model developed in this study is simple with good sensitivity and high specificity.It could provide reference for risk assessment,prevention and treatment of lower extremity DVT in patients with hemorrhagic stroke.