Adjusted preoperative variables to predict perioperative red blood cell transfusion in coronary artery bypass grafting

Background The variables for predicting blood transfusion perioperatively are not completely clear in coronary artery bypass grafting (CABG) patients. Objectives To construct a comprehensive model to predict perioperative RBC transfusion in patients undergoing isolated CABG using adjusted preoperative variables. Methods Perioperative data of 1253 patients who underwent isolated CABG by the same surgical team were collected from April 2018 to March 2019. Logistic regression analyses were used to establish equations to construct two models for predicting intraoperative and postoperative RBC transfusions, respectively. All significant variables included in the two models were combined to form a comprehensive model to predict perioperative RBC transfusion. Area under the receiver operating characteristic (ROC) curve (AUC) was used to evaluate the discriminatory power of the models. Results The total RBC transfusion rate for CABG patients during hospitalization was 29.05%. The rate of intraoperative and postoperative RBC transfusions was 6.9% and 26.7%, respectively. Eight variables in a total of 30 risk factors constituted the intraoperative prediction model, 12 variables constituted the postoperative prediction model, and 13 variables for the combined model. The AUC of the three models were 0.87, 0.82, and 0.83, respectively, demonstrating moderate discriminatory power for RBC transfusion during the intraoperative, postoperative, and perioperative periods. Conclusion The comprehensive model combined with all variables of predicting intraoperative and postoperative RBC transfusion is feasible for predicting perioperative RBC transfusion.