Abstract 12782: Predicting One-Year Survival in LVAD Patients Using Different Predictive Models

Introduction: Left ventricular assist device (LVAD) support improves quality of life and survival in patients with end-stage heart failure (HF), but predicting which patients benefit most remains a challenge. The Penn Columbia Risk Score (PCR) and pretransplant risk stratification score (RSS) have predicted 1-year mortality in LVAD and transplant patients, respectively. We sought to compare the PCR and RSS in our single center’s LVAD population and evaluate additional predictors of survival at 1 year. Methods: Perioperative data were collected from 202 patients who received an LVAD (HVAD, Heartmate II, or Heartmate III) from 2014-2020. The logistic regression model assessed goodness-of-fit of the PCR and RSS in our population. The PCR model is based on age, creatinine, total bilirubin (tbili), body mass index, right ventricular dysfunction, aortic insufficiency while the RSS score is based on age, prior sternotomy, HF etiology, eGFR, tbili, and mechanical support. Logistic regression models were used to identify additional predictors of 1-year mortality. Results: In our 202 patients, the RSS was more predictive of 1-year mortality (R 2 =0.14) than PCR (R 2 =0.02). In addition to age, tbili, inotrope score, pulmonary artery pulsatility index (PAPi), right ventricular dysfunction, mechanical support (defined as intra-aortic balloon pump, ECMO, or temporary LVAD), and albumin were each associated with 1-year mortality (p <0.05). A multivariate model including age, tbili, PAPi, right ventricular dysfunction, and albumin, had higher correlation with 1-year mortality than the RSS (R 2 =0.21). The addition of inotrope score and mechanical support to this model did not significantly change the association (R 2 =0.20). Discussion: In our patient population, the PCR and RSS were poorly predictive of 1-year survival showing almost no association. Our novel model showed higher association, while still being suboptimal. Further research is needed to refine which clinical characteristics will best predict success and survival in LVAD patients.