急性缺血性脑卒中病人吞咽障碍风险预测模型的构建与验证

Construction and validation of a risk prediction model for dysphagia in patients with acute ischemic stroke

目的:探讨急性缺血性脑卒中(AIS)病人发生吞咽障碍的危险因素,构建列线图风险预测模型并进行验证。方法:采用便利抽样法,选取2023年2月—10月在重庆市某三级甲等医院神经内科住院的644例AIS病人,分为建模组(n=448)和验证组(n=196),采用Logistic回归分析建立预测模型并绘制列线图,以Bootstrap法进行内部验证,验证集行外部验证;采用ROC曲线下面积、Hosmer-Lemeshow检验、C-index、校准曲线以及临床决策曲线评估模型的预测价值。结果:年龄≥60岁、NIHSS评分≥7分、脑干梗死、病灶多发、肢体最差肌力≤3级是AIS病人吞咽障碍的独立影响因素。列线图模型验证结果显示:建模集和验证集的AUC分别为0.83,0.80;C-index分别为0.82,0.78;Hosmer-Lemeshow拟合优度检验分别为χ2=5.05,P=0.65;χ2=12.30,P=0.14;两校准曲线基本吻合且斜率接近1;决策曲线结果显示阈值概率范围在0.00~0.89时,临床净收益较大。结论:本研究构建的列线图模型具有较好的预测能力,可为早期识别与管理AIS吞咽障碍病人提供依据。

Objective:To explore the risk factors for dysphagia in patients with acute ischemic stroke(AIS),to construct a nomogram risk prediction model,and to conduct validation.Methods:A total of 644 AIS patients hospitalized in the Department of Neurology of a tertiary Grade A hospital in Chongqing from February 2023 to October 2023 were selected by convenience sampling and divided into a modeling set(n=448)and a validation set(n=196).Logistic regression analysis was used to establish a prediction model and draw a nomogram.The Bootstrap method was used for internal validation,and the validation set was used for external validation.The area under the receiver operating characteristic curve(ROC),Hosmer-Lemeshow test,C-index,calibration curve,and clinical decision curve were used to evaluate the predictive value of the model.Results:Age≥60 years,NIHSS score≥7 points,brainstem infarction,multiple lesions,and the worst limb muscle strength≤3 were independent influencing factors for dysphagia in AIS patients.The validation results of the nomogram model showed that the area under curve(AUC)of the modeling set and the validation set were 0.83 and 0.80,respectively;the C-index was 0.82 and 0.78,respectively;the Hosmer-Lemeshow goodness of fit test wasχ2=5.05,P=0.65;χ2=12.30,P=0.14,respectively;the two calibration curves were basically consistent and the slope was close to 1;the decision curve results showed that when the threshold probability range was 0.00-0.89,the clinical net benefit was greater.Conclusions:The nomogram model constructed in this study has good predictive ability which could provide a basis for the early identification and management of AIS patients with dysphagia.