A quantitative model to predict blood use in adult orthotopic liver transplantation

To identify preoperative predictors for the use of any blood component during and after orthotopic liver transplantation (OLT), we performed a retrospective analysis on 602 OLT patients who were randomly split into a training set (n=482) and a validation set (n=120). Hemoglobin and calculated MELD score were identified as independent predictors for blood use using bootstrap aggregation. A logistic regression model constructed using both variables showed comparable performance in the training and validation sets. Predictive scores can be obtained from a nomogram, and a score above -2.328 predicted transfusion of any blood component with a positive predictive value of 97% and 96% in the training and validation sets, respectively.