本研究旨在探讨静脉溶栓时间窗外,急性缺血性脑卒中(Acute Ischemic Stroke,AIS)患者入院时常规临床资料对其短期不良结局的影响.该研究通过收集桂平市中医医院神经内科住院治疗的溶栓时间窗外AIS患者(774例)临床资料,比较短期不同预后患者入院时的临床资料,并利用机器学习分类(Classification)算法建模分析患者90 d内预后不良的影响因素,同时采用受试者工作特征曲线(ROC)和校准图验证模型的预测效能及准确度.研究结果显示:(1)3个月随访期内预后不良患者占比13.95％(108例).(2)预后不良患者的年龄、血白细胞(WB C)水平、血C反应蛋白(C-Reactive Protein,CRP)水平、血甘油三酯(TG)水平、入院美国国立卫生研究院卒中量表(National Institutes of Health Stroke Scale,NIHSS)评分、梗死灶体积和发病-治疗时间大于或高于预后良好患者;预后不良患者的入院格拉斯哥昏迷(Glasgow Coma Scale,GCS)评分低于预后良好组,均差异显著(P＜0.05).机器学习分类算法中,极限梯度提升(Extreme Gradient Boosting,XGB)算法的效果最佳[受试者工作特征曲线下面积(Area Under Curve,AUC)=0.81],且校准曲线显示模型评估预测风险与实际发生风险的一致程度高.(3)XGB算法筛选的预测变量影响权重居前6位的依次为发病-治疗时间、血CRP水平、年龄、梗死灶体积、血TG水平与NIHSS评分.上述结果说明XGB算法可用于预测静脉溶栓时间窗外AIS患者短期不良预后的影响因素.
Establishment and Validation of a Short-term Prognostic Model for Acute Ischemic Stroke Patients Beyond the Time Window for Intravenous Thrombolysis Based on Machine Learning
The objective of this study is to investigate the effects of clinical data at admission on the short-term adverse outcomes of Acute Ischemic Stroke(AIS)patients beyond the time window for intravenous throm-bolysis.The clinical data of 774 AIS patients beyond the time window for thrombolysis treated in the depart-ment of neurology in the Traditional Chinese Medical Hospital of Guiping City were collected.The clinical in-dicators at admission of patients with different short-term prognoses were compared.Machine learning classi-fication algorithms were used to model and analyze the factors influencing poor prognosis within 90 d.The prediction performance and accuracy of the nomogram model were evaluated based on the Receiver Operating Characteristic(ROC)curve and calibration plot.The results showed:(1)13.95％(108 cases)of the patients had poor prognosis within the 3-month follow-up period.(2)The patients with poor prognosis had higher age,levels of White Blood Cell(WBC),C-Reactive Protein(CRP),and Triglyceride(TG),National Institutes of Health Stroke Scale(NIHSS)score at admission,infarct volume,and onset-to-treatment time than the pa-tients with good prognosis.The patients with poor prognosis had lower Glasgow Coma Scale(GCS)score at admission than the patients with good prognosis(P＜0.05).The Extreme Gradient Boosting(XGB)algo-rithm performed best,with the area under the ROC curve(AUC)of 0.81,and the calibration curve showed a high degree of consistency between the predicted and actual occurrence of risk.(3)The top six predictive vari-ables selected by the XGB algorithm were onset-to-treatment time,CRP level at admission,age,infarct vol-ume,TG level,and NIHSS score.These results indicate that the XGB algorithm can be used to predict the factors influencing short-term poor prognosis in AIS patients beyond the time window for intravenous throm-bolysis.
acute ischemic strokebeyond the time window for thrombolysisprognosismachine learningclinical prediction model
万方数据
维普
