BackgroundAfter incident heart failure (HF) admission, patients are vulnerable to readmission or death in the 90‐day post‐discharge. Although risk models for readmission or death incorporate ejection fraction (EF), patients with HF with preserved EF (HFpEF) and those with HF with reduced EF (HFrEF) represent distinct cohorts. To better assess risk, this study developed machine learning models and identified risk factors for the 90‐day acute HF readmission or death by HF subtype.Methods and ResultsApproximately 1965 patients with HFpEF and 1124 with HFrEF underwent an index admission. Acute HF rehospitalization or death occurred in 23% of HFpEF and 28% of HFrEF groups. Of the 101 variables considered, multistep variable selection identified 24 and 25 significant factors associated with 90‐day events in HFpEF and HFrEF, respectively. In addition to risk factors common to both groups, factors unique to HFpEF patients included cognitive dysfunction, low‐pulse pressure, β‐blocker, and diuretic use, and right ventricular dysfunction. In contrast, factors unique to HFrEF patients included a history of arrhythmia, acute HF on presentation, and echocardiographic characteristics like left atrial dilatation or elevated mitral E/A ratio. Furthermore, the model tailored to HFpEF (area under the curve [AUC] = 0.770; 95% confidence interval [CI] 0.767–0.774) outperformed a model for the combined groups (AUC = 0.759; 95% CI 0.756–0.763).ConclusionThe UF 90‐day post‐discharge acute HF Re admission or Death Risk Assessment (UF90‐RADRA) models help identify HFpEF and HFrEF patients at higher risk who may require proactive outpatient management.
Read full abstract