医学
列线图
接收机工作特性
狭窄
逻辑回归
内科学
队列
曲线下面积
试验预测值
放射科
心脏病学
外科
作者
Xianjun Zhang,Xiaoliang Wang,Teng Ma,Wen-tao GONG,Yong Zhang,Naidong Wang
标识
DOI:10.1136/jnis-2024-022022
摘要
Background Hyperperfusion-induced cerebral hemorrhage (HICH) is a rare but severe complication in patients with carotid stenosis undergoing stent placement for which predictive models are lacking. Our objective was to develop a nomogram to predict such risk. Methods We included a total of 1226 patients with carotid stenosis who underwent stenting between June 2015 and December 2022 from three medical centers, divided into a development cohort of 883 patients and a validation cohort of 343 patients. The model used LASSO regression for feature optimization and multivariable logistic regression to develop the predictive model. Model accuracy was assessed via the receiver operating characteristic curve, with further evaluation of calibration and clinical utility through calibration curves and decision curve analysis (DCA). The model underwent internal validation using bootstrapping and external validation with the validation cohort. Results Older age (OR 1.07, p=0.005), higher degrees of carotid stenosis (OR 1.07, p=0.006), poor collateral circulation (OR 6.26, p<0.001), elevated preoperative triglyceride levels (OR 1.27, p=0.041) and neutrophil counts (OR 1.36, p<0.001) were identified as independent risk factors for HICH during hospitalization. The nomogram constructed based on these predictive factors demonstrated an area under the curve (AUC) of 0.817. The AUCs for internal and external validation were 0.809 and 0.783, respectively. Calibration curves indicated good model fit, and DCA confirmed substantial clinical net benefit in both cohorts. Conclusion We developed and validated a nomogram to predict HICH in patients with carotid stenosis post-stenting, facilitating early identification and preventive intervention in high-risk individuals.
科研通智能强力驱动
Strongly Powered by AbleSci AI