基于列线图模型的急性缺血性脑卒中后自发性出血转化风险预测研究

Predicting the risk of spontaneous hemorrhage conversion after acute ischemic stroke based on a columnar graph model

目的建立急性缺血性脑卒中(AIS)后自发性出血转化(sHT)的定量化、可视化预测模型,并通过列线图进行效能验证。方法选取AIS患者共240例,收集患者一般资料、血清学检测及影像学检查结果。将患者随机分为建模组(175例)和验证组(65例),同时根据影像学结果分为非HT组和HT组。采用R 4.1.1软件包和rms程序包建立列线图模型,同时应用Bootstrap方法重复抽样1000次进行内、外部验证,分别采用H-L拟合优度检验、临床决策曲线和ROC曲线评估列线图模型的校准度和区分度。结果240例AIS患者中,发生HT为60例(25.0%)。在建模组中,多因素Logistics回归结果显示,既往是否有心房颤动(房颤)病史、发病时NIHSS评分、Hb、高密度脂蛋白(HDL)及梗死面积是AIS后sHT的重要影响因素。建模组和验证组的H-L拟合优度检验的χ2值分别为5.61和0.74,对应的P值为0.13和0.69,表示所建立的列线图模型有较好的预测精确度;列线图预测建模组和验证组的ROC曲线下面积分别为0.963(95%CI:0.926~1.000)和0.977(95%CI:0.950~1.000),结果提示模型具有良好区分度。结论房颤史、NIHSS评分、Hb、HDL及梗死面积是AIS后sHT的独立影响因素,在此基础上综合评估建立可视化的列线图模型可有效预测AIS后sHT的风险。

Objective To establish a quantitative and visual prediction model for spontaneous hemorrhagic transformation(sHT)after acute ischemic stroke(AIS)and validate the efficacy by nomogram.Methods A total of 240 patients with AIS were selected,and the general data,serological tests and imaging findings were collected.The patients were randomly divided into modeling group(175 cases)and validation group(65 cases).The patients were also divided into non-HT group and HT group according to the imaging results.The R 4.1.1 software and the rms package were used to build the column line graph model,while Bootstrap method was applied to repeat sampling 1000 times for internal and external validation,and the H-L goodness-of-fit test,clinical decision curve and ROC curve were used to assess the calibration and discrimination of the column line graph model,respectively.Results Among 240 patients with AIS,bleeding conversion occurred in 60 cases(25.0%).In the modeling group,the results of multifactorial Logistic regression showed that the presence or absence of previous history of atrial fibrillation,NIHSS score at the onset,Hb,high-density lipoprotein(HDL)and infarct area were significant influencing factors for sHT after AIS.Theχ~2 values of the H-L goodness-of-fit test for the modeling and validation groups were 5.61 and 0.74,respectively,corresponding to P values of 0.13 and 0.69,indicating that the established column line graph model had good prediction accuracy;the area under the ROC curve for the column line graph prediction modeling group and validation group were 0.963(95%CI:0.926-1.000)and 0.977(95%CI:0.950-1.000),and the results suggested that the model had good discrimination.Conclusions Previous history of atrial fibrillation,NIHSS score size at onset,Hb,HDL and the size of infarct area are independent influencing factors of sHT after AIS.Establishing the visual nomogram model based on the above factors can effectively predict the risk of sHT after AIS.