Background and Objectives: Craniosynostosis repair is associated with a high perioperative transfusion rate, but preoperative prediction models remain limited. This exploratory study aimed to develop and internally validate clinically prespecified preoperative models for predicting intraoperative red blood cell transfusion in pediatric craniosynostosis surgery and to evaluate whether adding fused suture extent improved model performance. Materials and Methods: This retrospective single-center prediction model study included children who underwent craniosynostosis repair between 2014 and February 2026. Patients undergoing repeat procedures or concurrent surgery for other craniofacial anomalies were excluded. The outcome was any intraoperative red blood cell transfusion. Candidate predictors were prespecified as age, weight, American Society of Anesthesiologists Physical Status (ASA-PS), preoperative hemoglobin, preoperative platelet, and fused suture extent. Five paired baseline/full ridge-penalized logistic regression models were developed, with fused suture extent added only to the full models. Performance was evaluated using apparent and bootstrap optimism-corrected area under the receiver operating characteristic curve (AUC) and Brier score. Results: Twenty-one patients were included, and nine (42.9%) received intraoperative transfusion. Across all five comparisons, inclusion of fused suture extent improved optimism-corrected discrimination and reduced prediction error. Corrected AUC increased from 0.470 to 0.674, from 0.475 to 0.738, from 0.552 to 0.667, from 0.516 to 0.704, and from 0.466 to 0.694 across the five model pairs. The best-performing model included weight, preoperative hemoglobin, ASA-PS, and fused suture extent, with an optimism-corrected AUC of 0.738 and an optimism-corrected Brier score of 0.242. Conclusions: Inclusion of fused suture extent improved preoperative prediction of intraoperative transfusion and may support perioperative blood management planning in pediatric craniosynostosis surgery. However, external validation using larger independent cohorts is necessary prior to clinical implementation.
Byun et al. (Thu,) studied this question.