Abstract：
BACKGROUND: Ischaemic stroke is the leading cause of death worldwide, and early neurological deterioration(END) occurs in 20%-40% of patients, which is the main cause of severe neurological deficits and disability, and even increased mortality. The occurrence of END is closely related to the poor prognosis of the patients, so it is important to identify the risk factors for the occurrence of END in patients with AIS and target intervention at an early stage factors and targeted intervention is of great significance.
METHODS: Up to December 20, 2023, a comprehensive search was conducted across PubMed, Embase, Web of Science, MedLine, and The Cochrane Library for studies focusing on predictive models for END in acute stroke patients. Included studies either developed or validated predictive models. The Prediction Model Risk of Bias Assessment tool was utilized to assess bias in these prediction models. Pooled area under the curve values were calculated using DerSimonian and Laird random-effects model.
RESULTS: Nineteen studies, each presenting an original model, were identified. Predominantly constructed through logistic multiple regression, these models demonstrated robust predictive performance (area under the curve ≥0.80). Key predictors of END in acute ischemic stroke patients included blood glucose levels, baseline National Institute of Health Stroke Scale scores, extent of cerebral infarction, and stenosis in the carotid and middle cerebral arteries.
CONCLUSIONS: Clinical practitioners should closely monitor high-frequency predictors of END in patients. However, the varying quality of current models necessitates the selection of models that balance performance with operational simplicity in clinical practice.
METHODS: Up to December 20, 2023, a comprehensive search was conducted across PubMed, Embase, Web of Science, MedLine, and The Cochrane Library for studies focusing on predictive models for END in acute stroke patients. Included studies either developed or validated predictive models. The Prediction Model Risk of Bias Assessment tool was utilized to assess bias in these prediction models. Pooled area under the curve values were calculated using DerSimonian and Laird random-effects model.
RESULTS: Nineteen studies, each presenting an original model, were identified. Predominantly constructed through logistic multiple regression, these models demonstrated robust predictive performance (area under the curve ≥0.80). Key predictors of END in acute ischemic stroke patients included blood glucose levels, baseline National Institute of Health Stroke Scale scores, extent of cerebral infarction, and stenosis in the carotid and middle cerebral arteries.
CONCLUSIONS: Clinical practitioners should closely monitor high-frequency predictors of END in patients. However, the varying quality of current models necessitates the selection of models that balance performance with operational simplicity in clinical practice.
摘要：
目的：本综述旨在通过检查缺血性卒中患者早期神经功能恶化的危险因素和预测模型，为未来的研究提供临床指导和指导。
方法：截至2023年12月20日，对PubMed进行了全面搜索，Embase,WebofScience,MedLine,和Cochrane图书馆用于研究急性中风患者早期神经系统恶化的预测模型。纳入的研究开发或验证了预测模型。PROBAST工具用于评估这些预测模型中的偏差。使用DerSimonian和Laird随机效应模型计算曲线下的集合面积(AUC)值。
结果：19项研究，每个人都展示一个原始模型，已确定。主要通过逻辑多元回归构建，这些模型表现出稳健的预测性能(AUC≥0.80)。急性缺血性卒中患者早期神经功能恶化的关键预测因子包括血糖水平,美国国立卫生研究院卒中量表(NIHSS)基线评分,脑梗死的程度，颈动脉和大脑中动脉狭窄.
结论：临床医生应密切监测患者早期神经功能恶化的高频预测因子。然而,当前模型的质量参差不齐，因此需要选择在临床实践中平衡性能和操作简单性的模型。
方法：截至2023年12月20日，对PubMed进行了全面搜索，Embase,WebofScience,MedLine,和Cochrane图书馆用于研究急性中风患者早期神经系统恶化的预测模型。纳入的研究开发或验证了预测模型。PROBAST工具用于评估这些预测模型中的偏差。使用DerSimonian和Laird随机效应模型计算曲线下的集合面积(AUC)值。
结果：19项研究，每个人都展示一个原始模型，已确定。主要通过逻辑多元回归构建，这些模型表现出稳健的预测性能(AUC≥0.80)。急性缺血性卒中患者早期神经功能恶化的关键预测因子包括血糖水平,美国国立卫生研究院卒中量表(NIHSS)基线评分,脑梗死的程度，颈动脉和大脑中动脉狭窄.
结论：临床医生应密切监测患者早期神经功能恶化的高频预测因子。然而,当前模型的质量参差不齐，因此需要选择在临床实践中平衡性能和操作简单性的模型。
