Abstract: Objective To establish a screening model of cerebrovascular stenosis based on Logistic regression. Methods This study selected 1,118 physical examiners without cerebrovascular stenosis by magnetic resonance angiography(MRA)as the control group, and 1,329 with cerebrovascular stenosis diagnosed by digital subtraction angiography(DSA)as the case group. Distribution of the sites of cerebrovascular stenosis was described. Routine and economic examination indicators were adopted to construct a Logistic screening model of cerebrovascular stenosis in both male and female. The model was evaluated with area under the receiver operator characteristic curve(AUC). Results Among the 1,329 cerebrovascular stenosis patients, 985 were male and 344 were female. In the Logistic screening model with age adjustment, 8 variables were screened out for male, including age, aspartate transaminase, albumin, high-density lipoprotein, neutrophil value, hypertension, smoking and drinking, while 5 variables were screened out for female, including albumin, high-density lipoprotein, fasting blood glucose and hypertension. The high-density lipoprotein, albumin, hypertension and neutrophil value were involved in both male and female. The AUC was 0.927 and 0.888 for male and female, respectively. After 10-fold cross validation, the AUC was 0.928 and 0.885, respectively. Conclusion The model has good screening ability and is useful for identifying high-risk individuals with cerebrovascular stenosis.

Key words: Cerebrovascular stenosis, Logistic regression, Screening model