Development and evaluation of a dynamic nomogram model for intraoperative blood transfusion decision-making.

Objective: By gathering data on patients with intraoperative blood transfusion and investigating the factors influencing intraoperative blood transfusion in patients, we aimed to construct a dynamic nomogram decision-making model capable of continuously predicting the probability of intraoperative blood transfusion in patients. This was done to explore a new mode of individualized and precise blood transfusion and to guide doctors to make timely and reasonable blood transfusion decisions and save blood resources. Methods: Data of surgical patients in our hospital from 2019 to 2023 were collected. Among them, 705 patients who had blood transfusions were the experimental group, and 705 patients without intraoperative blood transfusions were randomly selected as the control group. Preoperative and intraoperative indicators of 1,410 patients were collected. 80% of the data set was used as the training set and 20% as the test set. In the training set, independent risk factors affecting intraoperative blood transfusion in patients were obtained through Lasso regression and binary logistic regression analysis, and the regression model was established and validated. Results: Through Lasso regression with cross-validation and binary logistic regression analysis, the independent risk factors affecting patients' intraoperative blood transfusion decision-making were determined as ASAs (III) (OR = 3.009, 95% CI = 1.311-6.909), surgical grading (IV) (OR = 3.772, 95% CI = 1.112-12.789), EBL (OR = 1.003, 95% CI = 1.002-1.004), preHGB (OR = 0.932, 95% CI = 0.919-0.946), LVEF (OR = 1.063, 95% CI = 1.028-1.098), Temp (OR = 57.14, 95% CI = 9.740-35.204), preAPTT (OR = 1.147, 95% CI = 1.079-1.220), and preDD (OR = 1.127, 95% CI = 1.048-1.212). The area under the curve (AUC) of the receiver operating characteristic curve (ROC) of the training set was 0.983, Conclusion: Variables were screened by Lasso regression, the model was developed by binary logistic regression, and a dynamic nomogram model for intraoperative transfusion decision-making was successfully fitted using R software. This model had good goodness of fit, discrimination, and calibration. The model can dynamically and instantaneously predict the probability of blood transfusion and its 95% confidence interval under the current patient indicators by selecting the patient's independent risk factors in the drop-down mode during the operation. It can assist doctors in making a reasonable blood transfusion decision quickly and save blood resources.