Temporal changes and predictors of perioperative blood transfusion in elective spine surgery: a retrospective analysis of a large surgical database

Development of machine learning models to predict perioperative blood transfusion in hip surgery

BACKGROUND

Allogeneic Blood transfusion is common in hip surgery but is associated with increased morbidity. Accurate prediction of transfusion risk is necessary for minimizing blood product waste and preoperative decision-making. The study aimed to develop machine learning models for predicting perioperative blood transfusion in hip surgery and identify significant risk factors.

METHODS

Data of patients undergoing hip surgery between January 2013 and October 2021 in the Peking Union Medical College Hospital were collected to train and test predictive models. The primary outcome was perioperative red blood cell (RBC) transfusion within 72 h of surgery. Fourteen machine learning algorithms were established to predict blood transfusion risk incorporating patient demographic characteristics, preoperative laboratory tests, and surgical information. Discrimination, calibration, and decision curve analysis were used to evaluate machine learning models. SHapley Additive exPlanations (SHAP) was performed to interpret models.

RESULTS

In this study, 2431 hip surgeries were included. The Ridge Classifier performed the best with an AUC = 0.85 (95% CI, 0.81 to 0.88) and a Brier score = 0.21. Patient-related risk factors included lower preoperative hemoglobin, American Society of Anesthesiologists (ASA) Physical Status > 2, anemia, lower preoperative fibrinogen, and lower preoperative albumin. Surgery-related risk factors included longer operation time, total hip arthroplasty, and autotransfusion.

CONCLUSIONS

The machine learning model developed in this study achieved high predictive performance using available variables for perioperative blood transfusion in hip surgery. The predictors identified could be helpful for risk stratification, preoperative optimization, and outcomes improvement.

Development and Validation of a Risk-Prediction Nomogram for Preoperative Blood Type and Antibody Testing in Spinal Fusion Surgery

OBJECTIVE

With advancements in minimally invasive techniques, the use of spinal fusion surgery is rapidly increasing and transfusion rates are decreasing. Routine preoperative ABO/Rh blood type and antibody screening (T&S) laboratory tests may not be appropriate for all spinal fusion patients. Herein, we constructed a nomogram to assess patient transfusion risk based on various risk factors in patients undergoing spinal fusion surgery, so that preoperative T&S testing can be selectively scheduled in appropriate patients to reduce healthcare and patient costs.

METHODS

Patients who underwent spinal fusion surgery between 01/2020 and 03/2023 were retrospectively examined and classified into the training (n = 3533, 70%) and validation (n = 1515, 30%) datasets. LASSO and multivariable logistic regression were used to analyze risk factors for blood transfusion. Nomogram predictive model was built according to the independent predictors and mode predictive power was validated using consistency index (C-index), Hosmer-Lemeshow (HL) test, calibration curve analysis and area under the curve (AUC) for receiver operating characteristic (ROC) curve. Bootstrap resampling was used for internal validation. Decision curve analysis (DCA) was applied to evaluate the model's performance in the clinic.

RESULTS

Being female, age, BMI, admission route, critical patient, operative time, heart failure, end-stage renal disease or chronic kidney disease (ESRD or CKD), anemia, and coagulation defect were predictors of blood transfusion for spinal fusion. A prediction nomogram was developed according to a multivariate model with good discriminatory power (C-index = 0.887); Bootstrap resampling internal validation C-index was 0.883. Calibration curves showed strong matching between the predicted and actual probabilities of the training and validation sets. HL tests for the training and validation sets had p-values of 0.327 and 0.179, respectively, indicating good calibration. When applied to the training set, the following parameters were found: AUC: 0.895, 95% CI: 0.871-0.919, sensitivity 78.2%, specificity 86.7%, positive predictive value 29.4% and negative predictive value 98.2%. If the model were applied in the training set, 2911 T&S tests (82.4%) would be eliminated, equaling a RMB349,320 cost reduction. The AUC in the internal validation was: 0.879, 95% CI: 0.839-0.927, sensitivity 75.2%, specificity 88.8%, positive predictive value 34.3%, negative predictive value 97.9%, would eliminate 1276 T&S tests (84.2%), saving RMB 153,120. The DCA curve indicated good clinical application value.

CONCLUSION

The nomogram based on 10 independent factors can help healthcare professionals predict the risk of transfusion for patients undergoing spinal fusion surgery to target preoperative T&S testing to appropriate patients and reduce healthcare costs.

Patient blood management programmes: keeping the ball rolling

Patient blood management programmes have been endorsed by the World Health Organization and multiple medical societies. It seems important to review the progress and results of patient blood management programmes so necessary modifications or new initiatives can be added to achieve their major goals. In this issue of the British Journal of Anaesthesia, Meybohm and colleagues show that a nationwide patient blood management programme had an impact and was potentially cost-effective in centres that previously utilised large amounts of allogeneic blood transfusions. Before implementing a programme, each institution might need to identify the area(s) of deficiency with respect to established patient blood management methods, which will warrant specific focus in subsequent clinical practice reviews.