Abstract 4000: Application of the Seattle Heart Failure Model to Patients Presented to a Heart Transplant Committee
Objective: To assess the predictive value of the Seattle Heart Failure Model (SHFM) when applied to patients with advanced HF presented to a heart transplant committee at a US institution.
Methods: We evaluated model discrimination and calibration in 357 consecutive patients presented to Cleveland Clinic heart transplant committee for the first time between 2004 –2007. Those with prior organ transplant, VAD, or in need for multi-organ transplantation were excluded. Eighty five percent had one or no missing SHFM variables. Missing data was imputed. Outcomes assessed were mortality (censored at VAD or UNOS 1 transplantation), and a composite of mortality/VAD/urgent transplantation.
Results: During a median follow-up of 26 months, 93 died, 57 received a VAD, and 205 were transplanted. Discrimination was better for mortality (c-index 0.71) than for the combined outcome (0.65). Calibration was adequate for mortality (Figure⇓), but there was systematic overestimation of event rates for the combined outcome. Amongst those deemed not transplant candidates discrimination was similar for the two outcomes (c-index 0.66 and 0.65) with adequate calibration. Amongst those listed as UNOS 2, discrimination varied widely (c-index 0.77 and 0.62) with systematic underestimation of event rates for death, and systematic overestimation for the combined outcome.
Conclusion: In patients with advanced HF presented to a heart transplant committee, SHFM performed better for mortality than for mortality/VAD/urgent transplantation. This difference was most prominent for those listed as UNOS status 2, suggesting that interpretation of SHFM prediction for this subset should be done with caution.