基于SII和CRP构建颅内血管狭窄风险预测模型

Constructing a Model to Predict the Risk of Intracranial Vascular Stenosis Using SII and CRP

摘要: 目的：颅内血管狭窄与脑梗死的严重程度相关，基于患者入院时病史资料､实验室指标开发预测模型识别颅内血管狭窄高危人群。方法：收集2022年9月至2023年4月于青岛大学附属医院住院行MRA的839例患者，使用单因素和多因素Logistics回归分析，筛选与颅内动脉狭窄相关的预测因子，构建颅内动脉狭窄风险预测评估表，并通过模型的区分度、校准度，来评估预测模型的临床实用性，采用1000次Bootstrapping法进行内部验证。结果：根据多因素Logistics筛选得到年龄、高血压、糖尿病、吸烟、体重指数、高脂血症、SII、CRP与颅内血管狭窄相关，构建风险预测模型，该模型有良好的区分度(AUC = 0.816 95%CI 0.787~0.845)和临床预测性(HL = 0.834)，内部验证时模型也表现出良好的区分度(AUC = 0.820 95%CI 0.791~0.849)和临床预测性(HL = 0.530)。结论：基于年龄、高血压、糖尿病、吸烟、体重指数、高脂血症、SII、CRP等因素构建的评分表可以用于预测颅内血管狭窄的风险，具有潜在临床应用价值。

Abstract: Objective: Intracranial vascular stenosis is associated with the severity of cerebral infarction. The aim of this study was to develop a user-friendly model for predicting the risk of intracranial arterial stenosis using admission history and laboratory indices. Methods: We conducted a retrospective analysis of clinical data from 839 patients who underwent MRA between September 2022 and April 2023 in Affiliated Hospital of Qingdao University. We identified risk factors associated with intracranial stenosis through univariate and multivariate logistic regression analyses. The prediction and evaluation table of intracranial artery stenosis risk was constructed, and the clinical practicability of the prediction model was evaluated by the degree of differentiation and calibration of the model. The 1000 times Bootstrapping method was used for internal verification. Results: According to multi-factor Logistics screening, age, hypertension, diabetes, smoking, body mass index, hyperlipidemia, SII and CRP are related to intracranial vascular stenosis, and a risk prediction model is constructed. The model had good differentiation (AUC = 0.816 95%CI 0.787~0.845) and clinical prediction (HL = 0.834), the model also showed good differentiation (AUC = 0.820 95%CI 0.791~0.849) and clinical prediction (HL = 0.530) during internal validation. Conclusion: A score scale based on age, hypertension, diabetes, smoking, body mass index, hyperlipidemia, SII, CRP and other factors can be used to predict the risk of intracranial vascular stenosis, which has potential clinical application value.