Longitudinal Multimarker Risk Models for Very Elderly Patients With Heart Failure and Preserved Ejection Fraction

The goal of this observational study is to develop longitudinal multimarker risk models for decision support during the clinical follow-up of very elderly patients with heart failure and preserved ejection fraction (HFpEF). The main questions it aims to answer are: * Can advanced risk prediction models accurately estimate the prognosis of very elderly patients with HFpEF over a 1-year follow-up after a hospitalization for acute heart failure? * Do novel biomarkers, in addition to routine clinical variables and elderly-specific predictors, contribute to improved risk prediction for these patients? To this end, very elderly patients (aged 80 or older) who have HFpEF and were admitted for acute heart failure will be included. Clinical and biological data will be collected during their hospitalization and also during follow-up visits 30 and 90 days after discharge. There is no comparison group in this observational study.

Background: Very elderly patients with heart failure and preserved ejection fraction (HFpEF) are under-represented in risk prediction models, and the role of prognostic biomarkers in this population is unclear due to the presence of cumulative comorbidity burden. Risk prediction is a useful tool to support decision making across the clinical follow-up of very elderly HFpEF patients. Aim: To develop longitudinal prognostic models based on readily available clinical and biological variables, novel biomarkers and elderly-specific predictors to estimate prognosis over 1-year follow-up after a HF hospitalization in very elderly patients with HFpEF. Design: Observational, single-centre, prospective cohort study of very elderly patients (≥80 years old) with HFpEF consecutively admitted for acute HF. Main outcome: Composite of 1-year all-cause mortality and/or HF-hospitalization. Sample size: 184 patients. Follow-up time: 1 year. Predictors: Routine clinical variables (sociodemographic, medical history, physical examination, vital signs, laboratory tests, imaging, concomitant medication, quality of life and elderly-specific factors) and novel biomarkers will be longitudinally collected during index hospitalization, 30-day and 90-day post-discharge visits. Statistical analysis: Kaplan-meier survival analysis. Logistic and Cox proportional-hazards regression models, time-to-event models for repeated events, linear-mixed effects, joint models, LASSO and machine learning techniques will be used for model development.