AI Model Can Optimize, Standardize Total Parenteral Nutrition Formulas

TUESDAY, April 1, 2025 (HealthDay News) — An artificial intelligence (AI) approach identified total parenteral nutrition (TPN) formulas that increase safety and potentially reduce costs for neonates admitted to neonatal intensive care units, according to a study published online March 25 in Nature Medicine.

Thanaphong Phongpreecha, from Stanford University in California, and colleagues developed TPN2.0, a data-driven approach that optimizes and standardizes TPN using information obtained from electronic health records. To train TPN2.0, TPN compositions from a decade were assembled (79,790 orders; 5,913 patients). The model was also validated in an external cohort (63,273 orders; 3,417 patients) from a second hospital.

The researchers identified 15 TPN formulas that can enable a precision-medicine approach using the algorithm (Pearson”s R = 0.94 compared to experts), which increased safety and potentially reduced costs. In a blinded study of 192 neonates, TPN2.0 was rated higher than current best practice by physicians. In patients with high disagreement between actual prescriptions and TPN2.0, increased morbidities were seen in association with standard prescriptions (e.g., odds ratio 3.33 for necrotizing enterocolitis), while a reduced risk was seen with TPN2.0 recommendations. Guideline-adhering, physician-in-the-loop recommendations were enabled by TPN2.0 employing a transformer architecture, allowing collaboration between the care team and AI.

“Taken together, TPN2.0 demonstrates the potential of AI to go beyond predictive diagnosis and guide a key therapeutic decision for newborns in the most vulnerable time of their lives,” the authors write.

Several authors disclosed ties to the biopharmaceutical industry; the methods described in this study are covered in a U.S. provisional patent.

Abstract/Full Text