AI-guided precision parenteral nutrition for neonatal intensive care units

Thanaphong Phongpreecha; Marc Ghanem; Jonathan D Reiss; Tomiko T Oskotsky; Samson J Mataraso; Davide De Francesco; S Momsen Reincke; Camilo Espinosa; Philip Chung; Taryn Ng; Jean M Costello; Jennifer A Sequoia; Sheila Razdan; Feng Xie; Eloise Berson; Yeasul Kim; David Seong; May Y Szeto; Faith Myers; Hannah Gu; John Feister; Courtney P Verscaj; Laura A Rose; Lucas W Y Sin; Boris Oskotsky; Jacquelyn Roger; Chi-Hung Shu; Sayane Shome; Liu K Yang; Yuqi Tan; Steven Levitte; Ronald J Wong; Brice Gaudillière; Martin S Angst; Thomas J Montine; John A Kerner; Roberta L Keller; Gary M Shaw; Karl G Sylvester; Janene Fuerch; Valerie Chock; Shabnam Gaskari; David K Stevenson; Marina Sirota; Lawrence S Prince; Nima Aghaeepour

doi:10.1038/s41591-025-03601-1

AI-guided precision parenteral nutrition for neonatal intensive care units

Nat Med. 2025 Jun;31(6):1882-1894. doi: 10.1038/s41591-025-03601-1. Epub 2025 Mar 25.

Authors

Thanaphong Phongpreecha^#^{1

2

3

4}, Marc Ghanem^#^{1

2

3}, Jonathan D Reiss^#², Tomiko T Oskotsky^#^{5

6}, Samson J Mataraso^{1

2

3}, Davide De Francesco^{1

2

3}, S Momsen Reincke^{1

2

3}, Camilo Espinosa^{1

2

3}, Philip Chung¹, Taryn Ng⁷, Jean M Costello⁵, Jennifer A Sequoia², Sheila Razdan^{2

8}, Feng Xie^{1

2

3}, Eloise Berson^{1

3

4}, Yeasul Kim^{1

2

3}, David Seong^{1

2

3}, May Y Szeto², Faith Myers², Hannah Gu², John Feister², Courtney P Verscaj², Laura A Rose², Lucas W Y Sin¹, Boris Oskotsky⁵, Jacquelyn Roger⁵, Chi-Hung Shu^{1

2

3}, Sayane Shome^{1

2

3}, Liu K Yang^{1

2

3}, Yuqi Tan^{4

9}, Steven Levitte¹⁰, Ronald J Wong², Brice Gaudillière¹, Martin S Angst¹, Thomas J Montine⁴, John A Kerner², Roberta L Keller⁶, Gary M Shaw², Karl G Sylvester², Janene Fuerch², Valerie Chock², Shabnam Gaskari⁷, David K Stevenson², Marina Sirota^{5

6}, Lawrence S Prince², Nima Aghaeepour^{11

12

13}

Affiliations

¹ Department of Anesthesiology, Pain and Perioperative Medicine, Stanford University, Stanford, CA, USA.
² Department of Pediatrics, Stanford University, Stanford, CA, USA.
³ Department of Biomedical Data Science, Stanford University, Stanford, CA, USA.
⁴ Department of Pathology, Stanford University, Stanford, CA, USA.
⁵ Bakar Computational Health Sciences Institute, University of California, San Francisco, CA, USA.
⁶ Department of Pediatrics, University of California, San Francisco, CA, USA.
⁷ Department of Pharmacy, Lucile Packard Children's Hospital, Palo Alto, CA, USA.
⁸ Department of Pediatrics, Keck School of Medicine of USC, Division of Neonatal and Infant Critical Care Unit, Children's Hospital Los Angeles, Los Angeles, CA, USA.
⁹ Department of Microbiology and Immunology, Stanford University, Stanford, CA, USA.
¹⁰ Division of Pediatric Gastroenterology, Hepatology and Nutrition, Stanford University, Stanford, CA, USA.
¹¹ Department of Anesthesiology, Pain and Perioperative Medicine, Stanford University, Stanford, CA, USA. naghaeep@stanford.edu.
¹² Department of Pediatrics, Stanford University, Stanford, CA, USA. naghaeep@stanford.edu.
¹³ Department of Biomedical Data Science, Stanford University, Stanford, CA, USA. naghaeep@stanford.edu.

^# Contributed equally.

Abstract

One in ten neonates are admitted to neonatal intensive care units, highlighting the need for precise interventions. However, the application of artificial intelligence (AI) in guiding neonatal care remains underexplored. Total parenteral nutrition (TPN) is a life-saving treatment for preterm neonates; however, implementation of the therapy in its current form is subjective, error-prone and resource-consuming. Here, we developed TPN2.0-a data-driven approach that optimizes and standardizes TPN using information collected routinely in electronic health records. We assembled a decade of TPN compositions (79,790 orders; 5,913 patients) at Stanford to train TPN2.0. In addition to internal validation, we also validated our model in an external cohort (63,273 orders; 3,417 patients) from a second hospital. Our algorithm identified 15 TPN formulas that can enable a precision-medicine approach (Pearson's R = 0.94 compared to experts), increasing safety and potentially reducing cost. A blinded study (n = 192) revealed that physicians rated TPN2.0 higher than current best practice. In patients with high disagreement between the actual prescriptions and TPN2.0, standard prescriptions were associated with increased morbidities (for example, odds ratio = 3.33; P value = 0.0007 for necrotizing enterocolitis), while TPN2.0 recommendations were linked to reduced risk. Finally, we demonstrated that TPN2.0 employing a transformer architecture enabled guideline-adhering, physician-in-the-loop recommendations that allow collaboration between the care team and AI.

MeSH terms

Algorithms
Artificial Intelligence*
Electronic Health Records
Female
Humans
Infant, Newborn
Infant, Premature
Intensive Care Units, Neonatal*
Male
Parenteral Nutrition* / methods
Parenteral Nutrition, Total* / methods
Precision Medicine* / methods

Abstract

MeSH terms

Grants and funding