Kang Jihoon, Noh Maengseok, Lee Juneyoung, Lee Youngjo, Bae Hee-Joon
Department of Neurology, Seoul National University College of Medicine Seoul National University Bundang Hospital Seongnam Republic of Korea.
Department of Statistics Pukyong National University Busan Republic of Korea.
J Am Heart Assoc. 2025 Apr;14(7):e036287. doi: 10.1161/JAHA.124.036287. Epub 2025 Mar 26.
Using the significant link between blood pressure fluctuations and neurological deterioration (ND) in patients with ischemic stroke, this study aims to develop a predictive model capable of tracking ND risk in real time, enabling timely detection of high-risk periods.
A total of 3906 consecutive patients with ischemic stroke were recruited. To develop an initial predictive model, we employed a multinomial logistic regression model incorporating clinical parameters. This model estimates the probability of ND occurring within 2 distinct time windows relative to hospital arrival: within the first 12 hours and between 12 and 72 hours. To refine ND risk assessments over time, we subsequently introduced an iterative risk-tracking model that uses continuously updated blood pressure measurements. We endeavored to integrate these models, assessing their combined discriminative capacity and clinical utility, with a particular emphasis on pinpointing time periods of increased ND risk. ND rates were observed at 6.1% within the first 12 hours and 7.3% between 12 and 72 hours, presenting the variation over time. Multinomial logistic models encountered disparities in significant predictors across different time zones. The iterative risk-tracking model was successfully set up to forecast ND within a 12-hour window at every measurement. The integrated models achieved an area under the receiver operating characteristic curve ranging from 0.68 to 0.76 for narrowing down ND risk identification within 12 hours without sacrificing predictive accuracy across diverse patient groups. At 90% and 70% sensitivity settings, the combined model presented slightly higher or comparable specificity and positive predictive values relative to conventional models.
This study presents a novel approach for real-time monitoring of ND risk in patients with ischemic stroke, using blood pressure trends to identify critical periods for potential intervention.
利用缺血性中风患者血压波动与神经功能恶化（ND）之间的显著关联，本研究旨在开发一种能够实时追踪ND风险的预测模型，以便及时发现高危期。
共招募了3906例连续的缺血性中风患者。为了开发初始预测模型，我们采用了包含临床参数的多项逻辑回归模型。该模型估计相对于入院时间在2个不同时间窗内发生ND的概率：在最初12小时内以及在12至72小时之间。为了随时间完善ND风险评估，我们随后引入了一种迭代风险追踪模型，该模型使用不断更新的血压测量值。我们努力整合这些模型，评估它们的综合判别能力和临床实用性，特别强调确定ND风险增加的时间段。在最初12小时内观察到ND发生率为6.1%，在12至72小时之间为7.3%，呈现出随时间的变化。多项逻辑模型在不同时区的显著预测因子方面存在差异。成功建立了迭代风险追踪模型，以在每次测量时预测12小时窗内的ND。整合模型在不牺牲不同患者群体预测准确性的情况下，将12小时内ND风险识别范围缩小的受试者工作特征曲线下面积在0.68至0.76之间。在90%和70%的灵敏度设置下，联合模型相对于传统模型呈现出略高或相当的特异性和阳性预测值。
本研究提出了一种用于实时监测缺血性中风患者ND风险的新方法，利用血压趋势来确定潜在干预的关键时期。
