Statistical Analyses
Continuous variables were given as mean ± SD or median and interquartile range (IQR) as appropriate. Categorical variables are presented as percentages. Patterns of distribution of continuous variables and equality of variances across groups were tested with Shapiro-Wilk and Levene tests, respectively. For continuous variables, either one-way ANOVA test with Welch correction or Kruskal-Wallis tests were used depending on the presence of normal distribution pattern. Post-hoc analyses for variables with a normal distribution were done using Tukey’s HSD or Games-Howell tests; for the remaining post-hoc analyses Dwass-Steel-Critchlow-Fligner test was used. For categorical variables, chi-square test was used for comparisons. Kaplan-Meier curves were drawn for survival analysis, and individual groups were compared with log-rank test. Cox proportional hazards model was used to determine individual predictors of one-year mortality. All parameters that had ap value <0.10 on univariates Cox regression were included in the initial model, and a backwards selection criterion was used to construct the final model. Receiver-operator curves were drawn to analyze the predictive accuracy of ACEF-MDRD for prediction of one-year mortality. Additionally, DeLong’s test was used to determine whether ACEF-MDRD is noninferior to GWTG-HF score in terms of accuracy. Net reclassification improvement index (NRI) was calculated as described before [19]. For all analyses, a p value of <0.05 was accepted as statistically significant. All statistical analyses were done using Jamovi (The jamovi project (2020). jamovi version 1.2 for Microsoft Windows), which is a graphical user interface for R language (R Core Team (2019). R: A Language and environment for statistical computing. Version 3.6 for Microsoft Windows) and SPSS 25.0 (IBM Inc, Armonk, USA).
To avoid data loss in Cox regression and DeLong’s test, a multiple imputation procedure was used to predict missing values. A total of 5 imputations were done and results from a pooled estimate of these 5 imputations were given as the result whenever possible. For all other statistical tests, original data was used and number of cases in whom data was available was indicated in parentheses.