We first divided the patients into non‐HFpEF and HFpEF groups and compared the patient characteristics among the 3 phenogroups within each of these groups. Parametric and nonparametric variables, as well as their respective differences, were assessed using a 1‐way ANOVA or the Kruskal‐Wallis test. Significant differences between the independent categorical variables were assessed using the χ2 test. The incidence of composite events or all‐cause death was estimated using the Kaplan‐Meier survival function and compared among the 3 phenogroups using the log‐rank test. We also compared the incidence of composite events in patients who received both β‐blockers and renin‐angiotensin system inhibitors at the time of discharge and those who did not, in each of the non‐HFpEF phenogroup, using the log‐rank test. The incidence of each mode of death was estimated, accounting for competing risks. For this, cardiovascular deaths and noncardiovascular deaths were considered competing events.
Cox‐proportional hazard analyses were used to evaluate the risk for all‐cause death among phenogroups. The covariates included in the multivariable model were LVEF, Get With The Guidelines‐Heart Failure risk score at discharge, and New York Heart Association functional class ≥III at discharge. Furthermore, when analyzing patients with non‐HFpEF, we added the prescription of β‐blockers or angiotensin coenzyme inhibitor/angiotensin II receptor blockers as covariates. In addition, we conducted a weighted Cox proportional hazard model using odds ratio based on the probability for membership in each of the 3 phenogroups as a sensitivity analysis. Statistical significance was set at P<0.05. All statistical analyses were performed using R software (version 4.1.3; R Foundation for Statistical Computing, Vienna, Austria).
Cox‐proportional hazard analyses were used to evaluate the risk for all‐cause death among phenogroups. The covariates included in the multivariable model were LVEF, Get With The Guidelines‐Heart Failure risk score at discharge, and New York Heart Association functional class ≥III at discharge. Furthermore, when analyzing patients with non‐HFpEF, we added the prescription of β‐blockers or angiotensin coenzyme inhibitor/angiotensin II receptor blockers as covariates. In addition, we conducted a weighted Cox proportional hazard model using odds ratio based on the probability for membership in each of the 3 phenogroups as a sensitivity analysis. Statistical significance was set at P<0.05. All statistical analyses were performed using R software (version 4.1.3; R Foundation for Statistical Computing, Vienna, Austria).
