Establishment And Application Of A Perioperative Red Blood Cell Transfusion Prediction Model

Purposes: The purpose of this study was to establish a predictive model for surgical patients to predict the risk of perioperative red blood cell transfusion in surgical patients by using clinical cases from public databases,and to identify patients with high perioperative red blood cell transfusion risk through the predictive model,providing information for surgeons’ clinical blood transfusion decisions.Methods: Clinical data of 97,443 surgical patients were extracted from public databases;75% of these patients were included in the derivation cohort cohort and 25% in the validation cohort by simple random sampling.Multivariate logistic regression analysis was used to identify risk factors for perioperative red blood cell transfusion,and logistic regression and nomogram were used to establish perioperative red blood cell transfusion prediction model of surgical patients.Receiver operating characteristic(ROC)curves,Youden index,and calibration curves were used to evaluate the discriminative ability of the models,and decision curves were used to evaluate model performance.Data from 400 surgical patients in Xiangya Hospital of Central South University were collected to validate the clinical application of the model.Result: 5,888 patients received more than 1 unit of red blood cells,and the transfusion rate is 6.04%.The final model screening out age,race,American Society of Anesthesiologists score,grade of kideny diseases,typeof anesthesia,priority of surgery,surgery risk,and a new 18-level variable defined by mean red blood cell volume,red cell distribution width,and anemia incorporated into the final model.The area under the curve(AUC)of the ROC curve was 0.870,showing the high specificity and sensitivity of this model in distinguishing patients with high transfusion risk.The Youden index identified an optimal cut-off value of 0.163 for the risk of perioperative red blood cell transfusion for patients in this study,corresponding to a sensitivity of 0.821 and a specificity of 0.744.Decision curves show that there is a net benefit to using this model to discriminate in surgical patients with a perioperative red blood cell transfusion risk of5-60%.Conclusion: A general model for the risk of red blood cell transfusion in perioperative patients was established,which can provide information for surgical blood decisions in clinical practice.