Statistical analysis
Statistical analysis was performed using the SPSS statistical software v. 25.0 (IBM Corporation, Armonk, NY, USA), and Stata v. 15.1 (StataCorp LLC, College Station, Texas, USA). Covariates and outcomes were reported as counts and percentages, and as mean and standard deviation or median and interquartile range (IQR). The Kruskal-Wallis test and the linear-by-linear association test were used for univariate analysis. The risk-adjusted mortality rates over different VA-ECMO interval durations were estimated using the direct standardization method employing thedstdize module of Stata, and the patient population were stratified according to different duration of VA-ECMO therapy: ≤3 days, 4-7 days, 8-10 days, and >10 days [3,7]. The impact of VA-ECMO duration on hospital mortality was also adjusted for the PC-ECMO score [4], multiple covariates and participating centers in logistic regression analysis. A P < .05 was set for statistical significance.