AI-guided precision parenteral nutrition for neonatal intensive care units

One in ten neonates are admitted to neonatal intensive care units, yet the application of AI in guiding their care remains underexplored. Total parenteral nutrition is a life-saving treatment for preterm neonates, but implementing it in its current form is subjective, error-prone, and resource-consuming. This study introduces TPN2.0, a data-driven approach that optimizes and standardizes TPN using information collected routinely in electronic health records.

Trained on a decade of TPN prescriptions from over 9,000 patients across two hospitals, TPN2.0 matched expert dosing (Pearson R = 0.94), and in a blinded review physicians rated its formulas above current best practice. Orders that diverged most from the model's recommendation carried about threefold the odds of necrotizing enterocolitis, suggesting the model captures dosing patterns that matter for outcomes.