Multivariable statistical models to predict red cell transfusion in elective surgery.

Area of Science:

• Anesthesiology
• Transfusion Medicine
• Health Informatics

Background:

• Predicting red cell transfusion is crucial for optimizing patient blood management (PBM) strategies.
• Identifying patients likely to require transfusion allows for targeted interventions in elective surgery.

Purpose of the Study:

• To develop a simple statistical model for predicting red cell transfusion in elective surgery.
• To utilize routinely available pre-admission data for transfusion prediction.

Main Methods:

• A multicentre cohort of 42,546 patients was analyzed.
• Potential predictors included age, sex, pre-admission hemoglobin, surgical procedure, and comorbidities.
• Multivariable logistic regression and 10-fold cross-validation were employed to build and evaluate the predictive model.

Main Results:

• 7.0% of patients received a red cell transfusion.
• A model using admission hemoglobin and surgical procedure achieved an Area Under the Curve (AUC) of 0.862.
• The model demonstrated high accuracy (0.934) and specificity (0.986) at a probability cut-off of 0.4.

Conclusions:

• A parsimonious model effectively predicts red cell transfusion using readily available pre-admission data.
• This predictive model can identify high-risk patients for proactive PBM interventions.
• The findings support the integration of simple predictive tools into pre-operative care pathways.