• Posted March 27, 2025

AI Improves IV Nutrition For Preemies

Artificial intelligence (AI) can help improve how premature babies are fed, giving them a better chance at normal growth and development, a new study says.

Currently, preemies in a neonatal intensive care unit are fed by IV, receiving a drip-drop handmade blend of nutrients that doctors call total parenteral nutrition, or TPN.

This is the only way to feed newborns whose digestive systems haven’t matured enough to properly absorb nutrients, researchers said.

“Right now, we come up with a TPN prescription for each baby, individually, every day,” senior researcher Nima Aghaeepour, an associate professor of pediatrics at Stanford University, said in a news release. “We make it from scratch and provide it to them.”

Unfortunately, the process is error-prone, and it’s tough for docs to know if they’ve gotten the formula right, researchers said.

There’s no blood test to measure whether a preemie has received enough daily calories, and preemies don’t necessarily cry when they’re hungry or become calm and content when they’re full.

“Total parenteral nutrition is the single largest source of medical error in neonatal intensive care units, both in the United States and globally,” Aghaeepour said.

To try to solve this problem, researchers trained an AI program on nearly 80,000 past prescriptions for preemie IV nutrition, linked to data on how the tiny patients fared.

The AI uses information in a preemie’s electronic medical chart to predict which nutrients they need and in what quantities, researchers said.

Tests of the AI show that it could perform better than a team of human doctors in getting babies the nutrition and medication they need, researchers reported March 25 in the journal Nature Medicine.

About 10% of babies are born prematurely, at least three weeks prior to their due date. Babies born more than eight weeks early generally are not ready to absorb nutrients through their intestines and require IV feeding, researchers said.

Based on the AI’s results, researchers honed down the countless variations in potential TPN prescriptions to 15 standard formulas.

“It turns out that with 15 distinct formulas for IV nutrition, what you are recommending is pretty similar to what the physicians, pharmacists and dietitians would have done anyway,” Aghaeepour said. “But then these 15 AI-based formulas can be used to significantly improve speed and safety.”

Researchers then tested whether the AI could use the preemie’s medical data to predict which of the 15 formulas a baby might need, and whether it could adjust the recommendations daily as the newborns grow and their medical condition changes.

In one test, researchers asked 10 neonatologists to review clinical information for past patients, alongside the IV nutrition prescriptions the preemies had actually received and the prescriptions that the AI would recommend.

Researchers asked the doctors, who were not told which prescription was human- or AI-generated, to pick the one that best matched the preemie’s condition.

The neonatologists consistently preferred the AI-generated prescriptions to the real prescriptions, researchers reported.

In another test, researchers scanned the medical records of past patients to find instances where the preemie’s TPN prescription was different from what the AI recommended. The team then assessed how the AI prescription might have improved the newborn’s outlook.

Results show that patients had a significantly higher risk of death, sepsis and bowel disease if their actual nutrition was significantly different from what the AI would have prescribed.

For example, babies had a more than three times higher risk of necrotizing enterocolitis, a serious GI problem affecting preemies, if their human-created TPN differed from what the AI would have recommended, the study says.

The researchers’ next step will be to run a clinical trial in which preemies fed the usual way will be compared to others receiving AI-recommended nutrition.

The research team noted that the AI’s recommendations will still need to be run past doctors and pharmacists, to make sure the program isn’t overlooking anything.

“The AI recommendation is based on whatever information has been added to a patient’s electronic medical record, so if something is missing from the record, the recommendation won’t be accurate,” researcher Shabnam Gaskari, executive director and chief pharmacy officer at Stanford Medicine Children’s Health, said. “We need a clinician to look at it and review.”

But if the system works, a preemie can promptly be provided one of the 15 standard nutrient formulas, which would be waiting on a hospital shelf, researchers said.

Using standard formulas also would make IV nutrition more accessible and less expensive, researchers noted.

Currently, crafting nutrition prescriptions requires the combined input of six experts, researchers said. 

These prescriptions include a blend of proteins, fats, carbohydrates, vitamins, minerals and electrolytes, as well as any medicines a preemie might need.

“This reflects our hope for how AI will enhance medicine: What it’s going to do is make doctors better and make top-notch care more accessible,” researcher Dr. David Stevenson, a neonatologist and professor of pediatrics at Stanford, said in a news release.

“Hopefully, it will also give our physicians more time to do the things computers can’t do, such as spending time with babies and their families, listening to them, and providing comfort and reassurance,” he added.

More information

Children’s Health has more on feeding premature babies.

SOURCE: Stanford University, news release, March 25, 2025