Comparison of observed survival after ventricular assist device placement versus predicted survival without assist device using the Seattle heart failure model.

Determining optimal timing for implant of ventricular assist device (VAD) in end-stage heart failure remains a challenge and may be aided by a risk assessment tool. For a cohort of 80 consecutive VAD implants at a single center, observed 1 year survival post-VAD was compared with the estimated survival had these patients not received a VAD, using the Seattle Heart Failure Model (SHFM). The SHFM was adjusted with a hazard ratio of 1.17 for inotrope use (Cochrane Meta-analysis of phosphodiesterase inhibitors) and a hazard ratio of 2.92 for balloon pump, ventilator, or renal replacement therapy (Comparative Outcome and Clinical Profiles in Transplantation [COCPIT] Model). Values immediately before surgery were used to calculate the SHFM score. Point estimates of 1 year survival were compared using Z scores. Mean age was 53 ± 14 (± standard deviation [SD]) years with mean left ventricular ejection fraction of 17 ± 6%. At the time of VAD implant, 92% were on inotropes, 53% had balloon pump, and 15% were intubated. For the entire cohort, 1 year survival without VAD predicted by the SHFM was 47% versus observed survival after VAD of 60% (p = 0.06). The model was most helpful in patients electively implanted with a left ventricular assist device (LVAD). In this group predicted 1 year survival on medical management was 49% versus an observed survival of 82% after LVAD placement (p < 0.05). The model was least helpful in patients undergoing placement of biventricular assist devices (BiVAD), where the model paradoxically predicted better survival with ongoing medical management. This indicated that the model was unable to forecast outcome in patients with higher severity of illness, for example, in cases warranting BiVAD placement. Observed 1 year survival was better with VAD versus that predicted with medical management. Tools such as the SHFM may aid in determining appropriate timing for VAD by providing an estimated survival with ongoing medical management. The model is best applied to more stable patients being considered for elective VAD implantation.