Graft Recovery Estimation at Transplantation (GREAT) Algorithm Identifies the Most Important Risk Factors for Primary Graft Dysfunction for Heart Transplantation

Purpose Primary Graft Dysfunction (PGD) is the most prevalent cause of one-month mortality following heart transplant(1). With the increase need to use marginal donor, because of the limited pool of donor available, an algorithm to predict PGD following heart transplantation could help identify marginal donors that are still suitable for transplantation. Methods 64 patients who underwent heart transplantation procedure without the use of organ care system were analyzed. PGD was defined as described before (2) and was present in 21. Donor, procedural and Recipient factors were run in an machine learning algorithm with lasso regression based on the coordinate descent technique provided by the glmnet algorithm. The predictive algorithm was generated for the outcome Primary Graft Dysfunction by continuous cross-referencing and re-sampling. Depending on the importance and distribution of each predictor in the model was assigned a score. The sum score was tested in a ROC-analysis with optimal cut-off calculated Results The following variables were found to be high risk to develop PGD (in descendant order): Cardiopulmonary bypass time, units of red blood cells given, cold ischemia time, cardioplegia amount, Time from brain death to recovery, Troponin T level in donor, diabetes in recipient and post-op creatinine level were left for the final algorithm. Scores were assigned in proportion to the weight given to them in the algorithm.). A ROC- analyze for the prediction of PGD showed a AUC (95% conf interval) 0.93(0.86-0.99). The optimal cut off was determined at 10 (0-29) Conclusion GREAT score can intra -operatively be used to prepare the transplantation team for who needs post-procedural support.