老年人急性前循环大面积脑梗死发生恶性脑水肿预测模型的建立与评价

Establishment and validation of a nomogram model for predicting malignant cerebral edema in elderly patients with acute large hemispheric infarction of the anterior cerebral artery

目的构建老年人急性前循环大面积脑梗死(LHI)发生恶性脑水肿(MCE)的预测模型并进行验证.方法回顾性分析2018年1月至2023年4月苏州大学附属第三医院神经内科收治的301例老年急性前循环LHI患者临床、实验室及影像学资料.以简单随机抽样法将患者按照7∶3的比例分为建模组(211例)和验证组(90例).根据是否发生MCE,对建模组数据进行单因素和多因素Logistic回归分析,筛选出发展为MCE的独立预测因子.运用R软件建立列线图并进行内部验证,并以验证组的数据进行外部验证,分别通过受试者工作特征(ROC)曲线、校准图和临床决策曲线分析(DCA)评估模型效果.结果建模组和验证组间MCE发生率和基线资料差异均无统计学意义,具有可比性.建模组多因素Logistic分析显示心房颤动史(OR=3.459,95%CI:1.202~9.955,P=0.021)、急性生理学与慢性健康状况评分系统Ⅱ(APACHEⅡ)评分(OR=1.202,95%CI:1.052~1.373,P=0.007)、美国国立卫生研究院卒中量表(NIHSS)评分(OR=1.163,95%CI:1.039~1.3013,P=0.008)、Alberta卒中项目早期CT评分(ASPECTS)(OR=0.782,95%CI:0.639~0.958,P=0.018)和侧支循环(CS)评分(OR=0.414,95%CI:0.221~0.777,P=0.006)是老年LHI患者发生MCE的独立预测因子.基于独立预测因子所构建的列线图模型,建模组发生MCE风险的ROC值为0.912(95%CI:0.867~0.957),验证组发生MCE风险的ROC值为0.957(95%CI:0.902~0.997).列线图在建模组及验证组中预测概率与实际概率均接近,具有较好的校准度;验证组DCA曲线显示该预测模型具有良好的临床实用性.结论本研究所构建的列线图模型对于MCE的预测具有良好的区分度及校准度,有一定的预测价值.

Objective To construct and validate a predictive model for the occurrence of malignant cerebral edema(MCE)in the elderly with acute large hemispheric infarction(LHI)of the anterior cerebral artery.Methods Clinical,laboratory and imaging data of 3o1 elderly patients with acute LHI of the anterior cerebral artery admitted to the Department of Neurology of the Third Affiliated Hospital of Soochow University between January 2018 and April 2023 were retrospectively analyzed.Patients were divided into a modeling group(211 cases)and a validation group(90 cases)by the simple random sampling method with a ratio of 7:3.According to the occurrence of MCE,univariate and multivariate Logistic regression analyses were performed with data from the modeling group to screen for independent predictors of the development of MCE.Nomograms were created and internally validated using R software.Additionally,external validation was performed with data from the validation group,and the performance of the model was assessed by receiver operating characteristic(ROC)curves,calibration plots,and clinical decision curve analysis(DCA),respectively.Results The MCE incidence and baseline data between the modeling and validation groups were not statistically significantly different and were actually comparable.Multivariate Logistic analysis in the modeling group showed that a history of atrial fibrillation(OR=3.459,95%CI:1.202-9.955,P=O.021),Acute Physiology and Chronic Health Evaluation II(APACHE II)score(OR=1.202,95%CI:1.052-1.373,P=0.007),National Institutes of Health Stroke Scale(NIHSS)score(OR=1.163,95%CI:1.039-1.3013,P=0.008),Alberta Stroke Program Early CT Score(ASPECTS)(OR=0.782.95%CI:0.639-0.958,P=0.018),and collateral score(CS)(OR=0.414,95%CI:0.221-0.777,P=0.006)were independent predictors of the occurrence of MCE in the elderly patients with LHI.Based on the nomogram model constructed using the independent predictors,the ROC value for the risk of developing MCE was 0.912(95%CI:0.867-0.957)in the modeling group and 0.957(95%CI:0.902-0.997)in the validation group.The predicted probabilities from the nomograms in the modeling and validation groups were close to the actual probabilities,indicating good calibration.The DCA curves in the validation group showed that the predictive model had good clinical utility.ConclusionsThe nomogram model established in this study exhibits good discrimination and calibration for the prediction of MCE,and has some predictive value.