Statistical Analysis

Depending on the distribution of continuous variables as assessed by the Shapiro-Wilk test, the results are shown as either arithmetic means with standard deviations or as median values with inter-quartile ranges. The significance of differences between the mean values was evaluated using the Student’s t-test. Categorical variables are shown as frequencies and percentages. Comparisons of the proportional categorical data from two different groups were performed using the chi-squared test with the Yate’s correction for continuity, or the Fisher’s exact test if the minimum expected count in the cell was less than 5. Agreement between the binary variables of the paired samples was analyzed using the kappa coefficient and McNemar’s test. The relationships between frailty and the other examined variables are listed in Table 1. Outcome (all-cause death and unscheduled rehospitalization) was assessed with the Cox’s proportional hazard model using univariate and backward multivariate procedures. The multivariate analysis included variables for which the level of statistical significance in univariate analysis was P < 0.1 ( Table 2). Variables included in the multivariate analysis of all-cause death were gender, age, BMI, previous CKD, malignant disease, hypercholesterolemia or hyperlipidemia, hemoglobin (Hgb) on admission to hospital, and left ventricular ejection fraction (LVEF) assessed during hospitalization. The variables included in the multivariate analysis of unscheduled rehospitalization were age, previous MI, CHF with EF ≥ 50%, CHF with EF < 50%, previous AF, previous CKD, previous stroke/TIA, Hgb on admission to hospital, LVEF assessed during hospitalization, and coronary angiography results. Each of the analyzed frailty scores were assessed in separate multivariate models. The goodness of fit models were evaluated with Harrell’s C-index. A significance level of P< 0.05 was required for the variable to remain in the multivariate model. A test for non-proportionality of hazards based on Schoenfeld residuals did not reveal significant violations of the proportionality assumptions. The probabilities of survival and survival free of unscheduled rehospitalization were estimated using the Kaplan–Meier method. The homogeneities of the curves with a different status of frailty syndrome were assessed with the log-rank test. All tests were two-sided and aP-value of < 0.05 was considered statistically significant. All statistical analyses were performed using SAS 9.4 software (SAS Institute, Cary, NC, USA).