Detalle Publicación

Development and performance evaluation of a clinical predictive model to estimate the risk of red blood cell requirements in brain tumor surgery

ISSN: 0898-4921
Volumen: 35
Número: 1
Páginas: 74 - 79
Fecha de publicación: 2023
Background: The identification of factors associated with perioperative red blood cell (RBC) transfusion provides an opportunity to optimize the patient and surgical plan, and to guide perioperative crossmatch and RBC orders. We examined the association among potential bleeding risk factors and RBC requirements to develop a novel predictive model for RBC transfusion in patients undergoing brain tumor surgery. Methods: This retrospective study included 696 adults who underwent brain tumor surgery between 2008 and 2018. Multivariable logistic regression with backward stepwise selection for predictor selection was used during modeling. Model performance was evaluated using area under the receiver operating characteristic curve, and calibration was evaluated with Hosmer-Lemeshow goodness-of-fit ¿2-estimate. Results: Preoperative hemoglobin level was inversely associated with the probability of RBC transfusion (odds ratio [OR]: 0.50; 95% confidence interval [CI]: 0.39-0.63; P<0.001). The need for RBC transfusion was also greater in patients who had a previous craniotomy (OR: 2.71; 95% CI: 1.32-5.57; P=0.007) and in those with larger brain tumor volume (OR: 1.01; 95% CI: 1.00-1.02; P=0.009). The relationship between number of planned craniotomy sites and RBC transfusion was not statistically significant (OR: 2.11; 95% CI: 0.61-7.32; P=0.238). A predictive model for RBC requirements was built using these 4 variables. The area under the receiver operating characteristic curve was 0.79 (95% CI: 0.70-0.87; P<0.001) showing acceptable calibration for predicting RBC transfusion requirements. Conclusions: RBC requirements in patients undergoing brain tumor surgery can be estimated with acceptable accuracy using a predictive model based on readily available preoperative clinical variables. This predictive model could help to optimize both individual patients and surgical plans, and to guide perioperative crossmatch orders.