Background: Few simple risk models, without echocardiography have been developed for patients with heart failure (HF) and preserved left ventricular ejection fraction (LVEF) (HFpEF). Methods: To develop a risk score to predict all-cause death for HFpEF patients, we examined 1277 HF patients with LVEF ≥50% and BNP ≥100 pg/ml in the CHART-2 Study, a large-scale prospective cohort study for HF in Japan. We selected the optimal subset of covariates for the score with Cox proportional hazard models and random survival forests (RSF). Results: During the median 5.7-year follow-up, 576 deaths occurred. Cox models and RSF analyses consistently indicated age ≥75 years, albumin <3.7 g/dl, anemia, BMI <22 kg/m 2 , BNP ≥300 pg/ml (or NT-proBNP ≥1400 pg/ml), and BUN ≥25 mg/dl, as the important 6 prognostic variables. Incorporating these 6 variables, we developed a scoring system (3A3B score, with 2 points given to age ≥75 years and 1 point to the others based on the hazard ratios. The discrimination ability of the risk score was excellent (c-index 0.708). Regarding model goodness-of-fit, the overall gradient in 5-year risk was well captured by the score. The predictive accuracy of the 3A3B score was confirmed in the external validation cohorts from the TOPCAT trial (N = 835, c-index 0.652) and the ASIAN-HF registry (N = 170, c-index 0.741). Conclusions: We developed a simple risk score to predict long-term prognosis of HFpEF patients. The 3A3B score, comprising 6 commonly available parameters in daily practice, has potential utility in the risk stratification and management of HFpEF patients.
- Heart failure
- Left ventricular ejection fraction
- Risk score