Abstract 2309: Evaluation of discrimination and calibration of the Heart Failure Survival Score and Seattle Heart Failure Model in ambulatory patients evaltuated for cardiac transplant.
Introduction: The Heart Failure Survival Score (HFSS) is a validated measure that predicts event-free survival in ambulatory heart failure patients undergoing transplant (TXP) evaluation. The Seattle Heart Failure Model (SHFM) was validated in several RCT databases, but has not been evaluated in patients specifically being considered for TXP. We compared the performance of the HFSS and SHFM in a broad group of ambulatory patients undergoing evaluation for TXP.
Methods: There were 498 patients from Center A (CtrA; index visits 1993–2001) and 666 patients from Center B (CtrB; 1990–2006). The primary outcome was UNOS 1 TXP, LVAD placement or death prior to TXP or device, as evaluated at 1 year. Model discrimination was assessed by the area under (AUC) the receiver-operator curve and was compared between models at each center. Calibration was assessed by plotting observed vs. predicted survival by quintiles of predicted survival.
Results: HFSS and its corresponding predicted 1 yr survival (1YRPRED) were 8.01+/−1.02 and 81% at CtrA and 8.25+/−1.28 and 85% at CtrB, p<0.001. SHFM and its corresponding 1YRPRED were 0.61+/−1.14 and 88% at CtrA and 0.98+/−1.00 and 85% at CtrB, p<0.001. Kaplan-Meier freedom from the primary outcome at 1 year was 81+/−2% (CtrA) vs. 86+/−1% (CtrB), log rank p=0.07. The AUCs for HFSS and SHFM were similar at CtrA (HFSS 0.75+/−0.03, SHFM 0.77+/−0.03; p=0.33) and CtrB (HFSS 0.70+/−0.03, SHFM 0.73+/−0.03; p=0.28). Calibration was good, with modest underestimation (HFSS) and overestimation (SHFM) of survival.
Conclusions: The HFSS and SHFM perform similarly well at 1 year with good discrimination and calibration in ambulatory patients undergoing TXP evaluation.