A Multi-State Model to Predict Heart Failure Hospitalizations and All-Cause Mortality in Outpatients With Heart Failure

Background: Among outpatients with heart failure (HF), early identification of those at high risk for HF hospitalization and/or death may help direct disease management services or advanced HF therapies. Currently, however, there are no models in this population to predict both heart failure hospitalization and all-cause mortality as individual outcomes as well as a composite outcome. Thus, we developed a model to predict both HF hospitalization and mortality, accounting for the semi-competing nature of the two outcomes. Methods and Results: A multi-state model to predict HF hospitalization and all-cause mortality was derived using data from the Heart Failure Endpoint evaluation of Angiotensin II Antagonist Losartan (HEAAL) study cohort, a multicenter, randomized trial of 3,834 symptomatic patients with reduced left ventricular ejection fraction. The following predictors were pre-specified for model inclusion: age, gender, New York Heart Association class III vs II, left ventricular ejection fraction, serum creatinine, serum sodium, systolic blood pressure, weight, history of diabetes mellitus, ischemic heart disease, atrial fibrillation, peripheral vascular disease or prior stroke. In this model, all patients were in the initial state of prevalent HF and were at risk for a HF hospitalization (transition 1, n5944) or death without a preceding HF hospitalization (transition 2, n5757). In addition, those who were hospitalized for HF were also at risk for death after a HF hospitalization (transition 3, n5528). To demonstrate model use, patients were grouped by quartile of predicted risk and the predicted probabilities of the patient with the median risk in each quartile were plotted over 7 years of follow-up (Figure 1). At one year of follow up, patient A (the patient with the median risk from the lowest risk quartile) has a 2% predicted