Statistical analyses
Normally distributed continuous data was presented as mean ± standard deviation and analysed using ANOVA, whereas skewed data was presented as median (interquartile range) and analysed using Wilcoxon rank-sum test. Categorical data was presented as number (percent) and analysed using either chi-square or Fisher’s exact test as appropriate.
To take into account the clustering of patients under surgeon, univariate and multivariate mixed effects logistic regression were performed to examine the associations between characteristics and the binary outcomes. As some surgeons operated in multiple hospitals the clustering of hospitals was not included in the models.
Due to the low incidence of 30-day mortality, the multivariate mortality analysis was constrained by the number of variables which could be included in the model. Surgeon and hospital volume were analysed as continuous rather than categorical variables. Age and gender were selected as a priori variables in addition to mitral valve repair.
The odds ratios (OR) for continuous variables including age and Body Mass Index (BMI) were presented in increments of 5 units. Stata version 14.2 was used for analysis. Statistical tests were carried out at a 2-sided 0.05 level of significance.