Short-term and lifetime prediction of heart failure hospitalisation and cardiovascular mortality in patients with heart failure with preserved ejection fraction: the LIFE-Preserved model

Background Heart failure with preserved ejection fraction (HFpEF) represents a heterogenous group of patients with varying prognosis. Early and accurate prediction of heart failure (HF) hospitalisation and cardiovascular (CV) mortality risks could enable identification of patients at highest risk, support shared decision-making, and increase motivation for drug treatment and lifestyle changes. Purpose To develop and validate the LIFE-Preserved model for the prediction of short-term and lifetime risk for HF hospitalisation or CV mortality in patients with HFpEF. Methods LIFE-Preserved was derived using data from 21,213 individuals aged between 40 and 90 years with established HF and a left ventricular ejection fraction ≥50% from the SwedeHF registry. Predictors were selected based on prior literature and availability in routine clinical settings and included: age, sex, diabetes mellitus, current smoking, any previous hospitalisation for HF, chronic obstructive pulmonary disease, atrial fibrillation, New York Heart Association class, ischemic heart disease, N-terminal pro-B-type natriuretic peptide, body mass index, heart rate, haemoglobin and estimated glomerular filtration rate. Sex-specific Cox proportional hazards models using age as time axis were derived to estimate the risk of HF hospitalisation or CV death (primary event) adjusted for the competing risk of non-CV death. Model performance was assessed using a predefined temporal validation subset of individuals in SwedeHF that did not contribute to model derivation (time-split with 75% derivation and 25% validation). Results The derivation sample in SwedeHF consisted of 8,222 females and 7,914 males with HFpEF, enrolled between 2000 and 2017, with a mean age at baseline of 78 ±9 years for females and 75 ±10 years for males. During a median follow-up of 2.7 years (interquartile range 0.7 - 5.2), in total 9,182 (57%) primary events and 6,041 (37%) non-CV deaths were observed. The validation sample consisted of 5,077 individuals, enrolled between 2017 and 2020, with 827 (16%) primary events during a median follow-up of 0.9 years (0.4 – 1.6). C-statistic in the time-split validation sample was 0.763 (95% confidence interval 0.745 – 0.781) and predicted risks were consistent with the observed incidence (Figure 1). Performance was similar in subgroups of sex and in patients enrolled from the in- and out-patient settings. The individual predicted risks of HF hospitalisations or CV death varied across the HFpEF population, with 2-year risks ranging from 0.02% to 73% depending on the individual’s characteristics (Figure 2). Conclusion The LIFE-Preserved model predicts both short-term and lifetime risk of HF hospitalisation or CV mortality in patients with HFpEF. It could enhance the early identification of high-risk HFpEF patients, aiding in personalised medicine and shared decision-making. Distribution of risks