(P158) ESTIMATING PROGNOSIS AMONG ADVANCED HEART FAILURE PATIENTS REFERRED FOR ADVANCED HEART FAILURE THERAPIES AT THE UNIVERSITY HEALTH NETWORK

Background: Patients with advanced heart failure (HF) are often considered for advanced therapies, including ventricular assist device or heart transplant, that improve quality of life and survival. This study evaluates (1) prognostic accuracy of 2 risk scores among advanced HF patients undergoing assessment for advanced HF therapies and (2) relation between medical decisions at the end of the assessment and patient prognosis.

METHODS AND RESULTS: This retrospective cohort study included 1,178 adult advanced HF patients who underwent assessment for advanced HF therapies at the University Health Network. Patients were included if they underwent assessment as an out-patient or as an in-patient but were not on mechanical circulatory support or using intravenous inotropes. At the end of the assessment, the medical team classified patients as a) candidates for advanced HF therapies, b) too well, or c) too sick. In patients who did not undergo advanced HF therapies, we assessed the discrimination (c-statistic) and calibration (observed vs predicted risk) of two risk scores (Seattle Heart Failure Model [SHFM] and MAGGIC score) using Cox proportional hazard models. Among all patients, we reported the proportion of patients in each of the three decision groups across model-predicted risk deciles.

Of the 1,178 patients (median age 55 years, 72% male, 32% ischemic cardiomyopathy), 397 (34%) were deemed candidates. Among the non-candidates (405 too well and 376 too sick), discrimination of the MAGGIC score and SHFM was limited (c-statistic 0.66 and 0.64, respectively). Model calibration by the MAGGIC score was excellent among patients in risk deciles 1-6 (predicted mortality risk 5-17%) however both over- and under-estimated 1-year mortality risk by ~10% among patients in risk deciles 7-10 (predicted risk 19-38%). The SHFM under-estimated risk of 1-year mortality by ~5% among patients in risk deciles 1-7 (predicted risk 2-18%) and over-estimated risk by ~5-15% among patients in risk deciles 8-10 (predicted risk 25-54%). Approximately one third of patients were deemed candidates across all risk categories (low to high risk), while the proportions of patients considered too well decreased and those considered too sick increased with increasing risk deciles (Figure 1).

Conclusion: The prognostic accuracy of the MAGGIC score and SHFM remain limited among ambulatory advanced HF patients. Physicians offered advanced HF therapies to patients across all risk deciles. A clinically significant proportion of patients were deemed too well when their predicted risk was high and too sick when their predicted risk was low.