Predicting 1-Year Mortality in Ambulatory Heart Failure Patients: Empiric Models Outperform Physician Intuitive Estimates

Abstract

<h3>Purpose</h3> This multicenter Canadian prospective cohort study evaluated the accuracy of physician predictions in comparison to model predictions in estimating 1-year mortality in ambulatory patients with heart failure (HF). <h3>Methods</h3> We included consented consecutive ambulatory HF patients (left ventricular ejection fraction (LVEF) <40%) followed at 11 HF clinics in 5 provinces in Canada (2018-2020). In a closed-question survey, HF cardiologists and family doctors, using their clinical judgment, estimated patient 1-year mortality in absolute terms. We collected clinical information to describe the population and calculate predicted mortality using three predictive models: the Seattle HF Model (SHFM), the MAGGIC score and the HF Meta-score. We documented patients' mortality over one year. Physicians were unaware of model predictions. We calculated mortality using Cox proportional hazards and assessed model and physician predictive accuracy using discrimination and calibration. <h3>Results</h3> We included 1,563 ambulatory HF patients, mean age of 63±14 years, 24% female, 70% NTHA class I-II, mean LVEF of 28±7%, and median brain natriuretic peptide of 350 pg/ml (25^th-75^th percentile 130-700). The population 1-year mortality was 7% (95%CI 6%-8%). Models' discrimination (c-statistic 0.80 for MAGGIC and HF meta-score, 0.84 for SHFM) proved superior to physicians (0.77 for family doctors and 0.78 for HF cardiologists). Models' calibration was adequate for patients at lower risk, but overestimated risk in those whose 1-year mortality was > 20%. In contrast, family doctors and HF cardiologists overestimated risk throughout the risk spectrum. In low-risk patients (1-year mortality ∼1%), physicians predicted a risk 7%-10%; in medium-risk patients (1-year mortality ∼5%), physicians predicted 15%-20%; and in high-risk patients (mortality >15%), physicians predicted >25%. <h3>Conclusion</h3> Predictive models showed higher accuracy in predicting 1-year mortality in ambulatory HF population than family doctors and HF cardiologists, both of whom consistently overestimated risk, potentially exposing patients early to more aggressive care. Incorporating predictive models in family and HF cardiology practices may improve patient management and resource use.