Copyright © 2020. Published by Elsevier Inc.PURPOSE: Shortage of organs for transplantation and improvements in LVADs make the use of this technology common as bridge to transplant (BTT). Compared to traditional statistical methods, machine learning (ML) techniques provides improvement in predictive modeling, identifying dimensionality and non-linear relationships between variables. Thus, we investigated specific risk factors that predispose to poor outcomes in pts supported with LVAD as BTT using ML and logistic regression (LR). METHODS: We included all pts that had heart transplant between 2006 and 2016. The primary outcome was the composite of 1-year mortality and re-transplant. We utilized ML method and LR to find the most predictive variables associated with the primary outcome. We excluded post-transplant variables. Receiver operating characteristic (ROC) curve was constructed to investigate the discriminatory capacity of the model. RESULTS: Of 18,612 pts (52±12 years, 24.58% female), 7,700 (41.12%) were on LVAD support. The discriminatory capacity predicting the primary outcome using the same variables modeled with ML or LR methods was similar in pts with LVAD or without the device (AUCs 0.61 and 0.63, respectively) (Figure A and B). Using ML and LR, the top 5 variables that were associated with poor outcomes in pts supported with LVAD were the recipient total bilirubin, creatinine, predicted right ventricular (RV) mass, and total albumin, as well as ischemic time during transplant. For pts without LVAD, the top 5 variables that were identified using ML and LR, were recipient total bilirubin, creatinine, and ventilator use, as well as ischemic time and distance of the donor. CONCLUSION: Both ML and LR methods identified total bilirubin, creatinine, and ischemic time among the strongest risk predictors of poor outcomes after transplant in pts with and without an LVAD. Notably predicted RV mass of the recipient was an important variable for pts with LVAD as a BTT.