Utilizing Machine Learning to Predict Perioperative Blood Transfusion in Pediatric Craniosynostosis Patients...

Area of Science:

• Pediatric Surgery
• Neurosurgery
• Medical Informatics

Background:

• Cranial vault remodeling (CVR) for craniosynostosis presents a high risk of perioperative blood transfusion.
• Understanding predictors of transfusion is crucial for mitigating associated complications.

Purpose of the Study:

• To identify patient- and procedure-specific factors associated with perioperative blood transfusion in pediatric craniosynostosis repair.
• To evaluate the predictive performance of machine learning models for transfusion risk.

Main Methods:

• Retrospective cohort study utilizing the ACS NSQIP-Pediatric database (2012-2023).
• Included patients aged 2 and under undergoing CVR for craniosynostosis.
• Employed supervised machine learning algorithms (logistic regression, random forest, etc.) to predict transfusion.

Main Results:

• Over 53% of 10,732 patients received a perioperative transfusion.
• Increased transfusion risk was linked to longer anesthesia duration and younger age.
• Higher preoperative hematocrit, minimally invasive surgery, and combination approaches were associated with lower transfusion rates.

Conclusions:

• Age, comorbidities, surgical approach, and ethnicity are key factors influencing transfusion risk in craniosynostosis repair.
• Early intervention and targeted strategies are needed to reduce transfusion necessity in high-risk pediatric populations.