PDF(Pubmed)
Abstract：
Background and purpose: Clinically, the ability to identify individuals at risk of ischemic stroke remains limited. This study aimed to develop a nomogram model for predicting the risk of acute ischemic stroke. Methods: In this study, we conducted a retrospective analysis on patients who visited the Department of Neurology, collecting important information including clinical records, demographic characteristics, and complete hematological tests. Participants were randomly divided into training and internal validation sets in a 7:3 ratio. Based on their diagnosis, patients were categorized as having or not having ischemic stroke (ischemic and non-ischemic stroke groups). Subsequently, in the training set, key predictive variables were identified through multivariate logistic regression and least absolute shrinkage and selection operator (LASSO) regression methods, and a nomogram model was constructed accordingly. The model was then evaluated on the internal validation set and an independent external validation set through area under the receiver operating characteristic curve (AUC-ROC) analysis, a Hosmer-Lemeshow goodness-of-fit test, and decision curve analysis (DCA) to verify its predictive efficacy and clinical applicability. Results: Eight predictors were identified: age, smoking status, hypertension, diabetes, atrial fibrillation, stroke history, white blood cell count, and vitamin B12 levels. Based on these factors, a nomogram with high predictive accuracy was constructed. The model demonstrated good predictive performance, with an AUC-ROC of 0.760 (95% confidence interval [CI]: 0.736-0.784). The AUC-ROC values for internal and external validation were 0.768 (95% CI: 0.732-0.804) and 0.732 (95% CI: 0.688-0.777), respectively, proving the model\'s capability to predict the risk of ischemic stroke effectively. Calibration and DCA confirmed its clinical value. Conclusions: We constructed a nomogram based on eight variables, effectively quantifying the risk of ischemic stroke.
摘要：
背景和目的：临床，识别有缺血性卒中风险的个体的能力仍然有限.本研究旨在建立预测急性缺血性卒中风险的列线图模型。方法:在本研究中,我们对参观神经内科的患者进行了回顾性分析，收集重要信息，包括临床记录，人口特征，和完整的血液学检查.参与者以7：3的比例随机分为训练集和内部验证集。根据他们的诊断,患者被分类为患有或未患有缺血性卒中(缺血性和非缺血性卒中组).随后，在训练集中，通过多变量逻辑回归和最小绝对收缩和选择算子(LASSO)回归方法确定关键预测变量，并据此构建了列线图模型。然后通过受试者工作特征曲线下面积（AUC-ROC）分析，在内部验证集和独立的外部验证集上评估模型，Hosmer-Lemeshow适合度测试,和决策曲线分析(DCA)验证其预测效能和临床适用性。结果：确定了八个预测因素：年龄，吸烟状况,高血压,糖尿病,心房颤动,中风史,白细胞计数,和维生素B12水平。基于这些因素,构建了具有高预测准确性的列线图.该模型表现出良好的预测性能，AUC-ROC为0.760（95%置信区间[CI]：0.736-0.784）。内部和外部验证的AUC-ROC值分别为0.768（95％CI：0.732-0.804）和0.732（95％CI：0.688-0.777），分别,证明模型有效预测缺血性卒中风险的能力。校准和DCA证实了其临床价值。结论：我们基于八个变量构建了一个列线图，有效量化缺血性卒中的风险。
