急性缺血性卒中患者入院时临床资料对患者短期预后结局的预测模型建立

Establishment of predictive model with clinical data at admission on short⁃term prognosis of patients with acute ischemic stroke

目的探讨急性缺血性脑卒中(acute ischemic stroke,AIS)患者入院时临床资料对患者结局的影响。方法回顾性分析2015年1月至2018年12月在我院神经内科住院治疗的AIS患者(n=1 476)入院时临床资料,比较不同预后患者的临床指标,将有价值的指标纳入多因素Logistic回归分析,筛选出影响患者不良结局的指标并绘制列线图。采用ROC曲线和校准图验证列线图模型的预测效能及准确度。结果男性(OR=1.081,95%CI:1.039~1.125)、年龄(OR=1.138,95%CI:1.056~1.226)、发病-治疗时间(OR=4.992,95%CI:3.447~7.229)、梗死体积(OR=1.016,95%CI:1.005~1.027)、NIHSS(OR=1.138,95%CI:1.056~1.226)、GCS(OR=0.864,95%CI:0.777~0.961)、CRP(OR=1.148,95%CI:1.049~1.256)和CYS(OR=0.862,95%CI:0.796~0.934)是短期不良预后的独立影响因素(P <0.05)。训练集ROC曲线AUC=0.91,验证集AUC=0.89。校准图显示,预测模型校准曲线与标准曲线接近,差异无统计学意义(P> 0.10)。结论通过入院时临床资料(性别、年龄、发病-治疗时间、梗死体积、NIHSS、GCS、CRP、CYS)建立的预测AIS患者不良结局风险的预测模型具有良好的诊断效能和准确度,对鉴别不良结局发生的风险人群,制定干预决策有重要价值。

Objective To investigate the efficacy of clinical data at admission on the short outcomes of patients with acute ischemic stroke(AIS).Methods Retrospective analysis of the clinical data at admission of AIS patients(n=1476)admitted to our hospital from January 2015 to December 2018 in neurology department,clinical data of patients with different prognosis were compared and valuable features were put into multiple⁃factor Logistic Regression analysis to screen out which influence the outcomes and nomogram were plotted.ROC curve and calibration chart were used to verify the performance and accuracy of the predictive model.Results Male(OR=1.081,95%CI:1.039~1.125),age(OR=1.138,95%CI:1.056~1.226),onset⁃treatment interval(OR=4.992,95%CI:3.447~7.229),Infarct volume(OR=1.016,95%CI:1.005~1.027),National Institute of Health stroke scale(NIHSS)(OR=1.138,95%CI:1.056~1.226),Glasgow Coma Scale(GCS)(OR=0.864,95%CI:0.777~0.961),C⁃reactive protein(CRP)(OR=1.148,95%CI:1.049~1.256)and Cysteine(CYS)(OR=0.862,95%CI:0.796~0.934)were all independent factors of short⁃term poor prognosis(P<0.05).The AUC of ROC curve of training cohort was 0.91 and the which of the test cohort was 0.89.The calibration test showed that the calibration curve of the predictive model was close to the standard curve,and there was no statistically significant difference(P>0.10).Conclusions The predictive model established on clinical data at admission(gender,age,onset⁃treatment interval,infarct volume,NIHSS,GCS,CRP and CYS)to predict the risk of short⁃term adverse outcomes in AIS patients has good diagnostic power and accuracy,and it′s valuable to identify the high⁃risk patients with adverse outcomes and clinical management.