Department of Anesthesiology, Sun Yat-sen Memorial Hospital, Sun Yat-sen University, Guangzhou, China. Electronic address: email@example.com. Department of Surgical Intensive Care Unit, Sun Yat-sen Memorial Hospital, Sun Yat-sen University, Guangzhou, China. Department of Anesthesiology, Sun Yat-sen Memorial Hospital, Sun Yat-sen University, Guangzhou, China. Department of Physiology, Zhongshan Medical School, Sun Yat-sen University, Guangzhou, China. Department of Anesthesiology, Sun Yat-sen Memorial Hospital, Sun Yat-sen University, Guangzhou, China. Electronic address: firstname.lastname@example.org. Department of Anesthesiology, Sun Yat-sen Memorial Hospital, Sun Yat-sen University, Guangzhou, China. Electronic address: email@example.com.
European journal of surgical oncology : the journal of the European Society of Surgical Oncology and the British Association of Surgical Oncology. 2022
BACKGROUND A predictive model that can identify patients who are at increased risk of intraoperative blood transfusion could guide preoperative transfusion risk counseling, optimize health care resources, and reduce medical costs. Although previous studies have identified some predictors for particular populations, there is currently no existing model that uses preoperative variables to accurately predict blood transfusion during surgery, which could
help anesthesiologists optimize intraoperative anesthetic management. METHODS We collected data from 582 patients who underwent elective liver resection at a university-affiliated tertiary hospital between January 1, 2018, and December 31, 2020. The data set was then randomly divided into a training set (n = 410) and a validation set (n = 172) at a 7:3 ratio. The least absolute shrinkage and selection operating regression model was used to select the optimal feature, and multivariate logistic regression analysis was applied to construct the transfusion risk model. The concordance index (C-index) and the area under the receiver operating characteristic (ROC) curve (AUC) were used to evaluate the discrimination ability, and the calibration ability was assessed with calibration curves. In addition, we used decision curve analysis (DCA) to estimate the clinical application value. For external validation, the test set data were employed. RESULTS The final model had 8 predictor variables for intraoperative blood transfusion, which included the following: preoperative hemoglobin level, preoperative prothrombin time >14 s, preoperative total bilirubin >21 μmol/L, respiratory diseases, cirrhosis, maximum lesion diameter >5 cm, macrovascular invasion, and previous abdominal surgery. The model showed a C-index of 0.834 (95% confidence interval, 0.789-0.879) for the training set and 0.831 (95% confidence interval, 0.766-0.896) for the validation set. The AUCs were 0.834 and 0.831 for the training and validation sets, respectively. The calibration curve showed that our model had good consistency between the predictions and observations. The DCA demonstrated that the transfusion nomogram was reliable for clinical applications when an intervention was decided at the possible threshold across 1%-99% for the training set. CONCLUSION We developed a predictive model with excellent accuracy and discrimination ability that can help identify those patients at higher odds of intraoperative blood transfusion. This tool may help guide preoperative counseling regarding transfusion risk, optimize health care resources, reduce medical costs, and optimize anesthetic management during surgery.