Statistical analysis methods
Measurement data are expressed as mean and standard deviation (SD); classification and grade data are expressed as rate and constituent ratio. Chi-square test is used for classification and grade data; t-test or analysis of variance is used for measurement data; Pearson correlation and Spearmen rank correlation (rank data) are used for correlation analysis. For MACEs incidence during follow-up, we will use multivariate Cox regression models to test the difference between BVHS groups, adjusting for possible imbalanced variables at the baseline. We will also test if the intervention effect is modified by age, gender, cardiovascular risk factors and number of medications. Discrimination was evaluated by the C statistic. Calibration was assessed using a calibration plot. A P<0.05 (two-side) is statistically significant. Data inventory and management using SAS (version 3.5.1, Vienna, Austria); all statistical analysis is completed by SAS (except for the construction of multi-level model needs to be combined with professional software MLwiN2.10).