Respiratory risk score for the prediction of 3-month mortality and prolonged ventilation after liver transplantation.

Survival of critically ill patients is significantly affected by prolonged ventilation. The goal of this study was the development of a respiratory risk score (RRS) for the prediction of 3-month mortality and prolonged ventilation after liver transplantation (LT). Two hundred fifty-four consecutive LT patients from a single center were retrospectively randomized into a training group for model design and a validation group. A receiver operating characteristic (ROC) curve analysis was used to test sensitivity and specificity. The accuracy of the predictions was assessed with the Brier score, and the model calibration was assessed with the Hosmer-Lemeshow test. Cutoff values were determined with the best Youden index. The RRS was calculated in the first 24 hours as follows: (laboratory Model for End-Stage Liver Disease score > 30 = 2.36 points) + (fresh frozen plasma > 13.5 U = 2.70 points) + (partial pressure of arterial oxygen/fraction of inspired oxygen ratio < 200 mm Hg = 2.23 points) + (packed red blood cells > 10.5 U = 3.50 points) + (preoperative mechanical ventilation = 3.87 points) +  (preoperative dialysis = 2.83 points) + (donor steatosis hepatis > 40% = 2.95 points). The RSS demonstrated high predictive accuracy, good model calibration, and c statistics > 0.7 in the training and validation groups. The RSS was able to predict 3-month mortality [cutoff = 6.64, area under the (ROC) curve (AUROC) = 0.794] and prolonged ventilation (cutoff = 3.69, AUROC = 0.798) with sensitivities of 69% and 81%, specificities of 83% and 73%, and overall model correctness of 76% and 77%, respectively. In conclusion, this study provides the first prognostic model for the prediction of 3-month mortality and prolonged ventilation after LT with high sensitivity and specificity and good model accuracy. The application of the RRS to an external cohort would be desirable for its further validation and introduction as a clinical tool for intensive care resource planning and prognostic decision making.