learning models can analyze trends in hemoglobin levels, oxygen requirements, vital signs, and clinical parameters.Research shows that random forest models achieved accuracies and AUCs of 0.87 in predicting neonatal outcomes.Similar architectures are being adapted for transfusion thresholds. 1 Deep neural networks (DNNs) have been validated for predicting outcomes in fragile NICU patients by analyzing ventilator support and intravascular volume expansionparameters directly linked to the need for red blood cell transfusions. 2 IntroductIonArtificial intelligence (AI) is increasingly transforming health care by improving clinical decision-making, optimizing resource utilization, and enhancing patient safety.In transfusion medicine, AI offers promising solutions to improve the efficiency, safety, and personalization of blood transfusions in neonates and pediatric patients.These populations present unique challenges due to small blood volumes, developmental physiology, and increased vulnerability to transfusion-related complications. Clinical Decision Support for Transfusion ThresholdsOne of the most important roles of AI is the development of clinical decision support systems (CDSS) that assist clinicians in determining the optimal timing and necessity of transfusion.Machine learning models can analyze trends in hemoglobin levels, oxygen requirements, vital signs, and clinical parameters.AI can generate personalized transfusion thresholds, especially for preterm neonates or critically ill children.Predictive models may reduce unnecessary transfusions and improve adherence to evidence-based guidelines.Such systems are particularly useful for managing conditions such as anemia of prematurity, where transfusion practices vary widely across institutions.Machine
Arora et al. (Fri,) studied this question.