脑血管血液动力学参数脑卒中预测模型的建立

目的 根据脑血管血液动力学参数(CVHI)和脑卒中的主要危险因素建立脑卒中预测模型。方法 选择全国六大行政区脑卒中研究队列人群25355例，将基线调查时的CVHI检测结果 进行主成分分析，再以各主成分和主要脑卒中危险因素为自变量，以随访中脑卒中发病为应变量进行回归分析，根据回归系数建立脑卒中预测模型，计算发病概率，绘制ROC曲线，确定最佳截断点，评价预测模型的预测效能。结果 四个主成分的累积贡献率依次为58．1％、79．4％、88．4％和94．6％，被筛检进入logistic回归方程的变量分别为第一至第四主成分、高血压病史、年龄和性别，ROC曲线下面积为0．855，最佳截断点为预测概率≥0．05，预测脑卒中的敏感度、特异度和准确度分别为踟．7％，78．5％，78．5％。结论 通过主成分回归分析，可以建立具有良好效能的脑卒中预测模型。

Principal component analysis and integral methods of cerebral vascular hemodynamic parameters

Objective To establish a predicting model for stroke according to cerebral vascular hemodynamic indexes and major risk factors of stroke. Methods Participants selected from a stroke cohort with 25 355 population in China. The first step was to carry out principal component analysis using CVHI. Logistic regression with principal component and main risk factors of stroke were then served as independent variables and stroke come on as dependent variables. The predictive model was established according to coefficient of regression and probability of each participant was also estimated. Finally, ROC curve was protracted and predictive efficacy was measured. Results The accumulative contribution rates of four principal components were 58.1 %, 79.4 %, 88.4 % and 94.6 % respectively.Seven variables were being selected into the equation with the first to fourth principal component as history of hypertension, age and sex. Area under ROC curve was 0.855 and optimal cut- off point was probability over 0.05 . Sensitivity, specificity and accuracy of stroke prediction were 80.7 %, 78.5 % and 78.5 % respectively. Conclusion The model established by principal component and regression could effectively predict the incidence of stroke coming on.