Bronchopulmonary dysplasia (BPD) is a common and chronic respiratory condition that occurs in premature newborns with underdeveloped lungs. BPD can affect growth and/or neurodevelopment and is sometimes fatal. Early, personalized treatments could help many of these children; however, it's not always clear which babies are at risk for BPD.
But now, a study published in the Journal of Pediatrics has demonstrated a new computational method that could better predict which infants will develop BPD, giving clinicians better tools to individualize care.
"The ability to identify children who will develop severe BPD would help us target those kids earlier," said Divya Chhabra, associate professor of Pediatric Pulmonology at UC Davis Health and first author on the study. "Eventually, we would like to incorporate this into the electronic health record to provide more immediate insights at the point of care."
Building a dynamic diagnostic model
Most infants who develop BPD are born premature, making them quite fragile. In addition, a number of factors can play a role in BPD development, making it difficult for physicians to know who is most at risk.
While the Neonatal Research Network, a large collaboration of clinical sites, has developed an online BPD calculator, the study authors and others believe this tool could be improved. The calculator estimates risk by processing a variety of data about the baby, including birth weight and respiratory support. However, it does so based on static (single timepoint) data from each infant at different ages, rather than gathering continuous patient information over time.
Chhabra and colleagues at University of Rochester Medicine, where she worked before joining UC Davis Health in September 2025, built a database of sick infants to study BPD and other health issues.
"We collected their vitals, birth weights, gestational ages, the medications they were given, how much oxygen they were on and other data from their charts," Chhabra said. "We used that data to develop a more dynamic version of the BPD calculator. Instead of using infant data from one timepoint, we used information from a series of points."
The research team devised three computational models to predict BPD, with each more sophisticated version improving results. The third model used a machine learning technique called long short-term memory, which showed much stronger predictive capabilities to guide care.
"The more data we added to the model, the better it got," Chhabra said. "In the future, we hope these predictions are available to us when we are rounding in the Neonatal Intensive Care Unit (NICU). As a result, our approach to each patient's condition would change. Also, sharing this information is good for families. Having a baby in the NICU is scary and better data can help reduce people's fears."
One of the most important, and surprisingly simple, findings was that the baby's first temperature reading correlated closely with their risk of developing BPD. This highlighted the importance of keeping infants warm during and right after delivery.
"These babies are born into a cold environment, and they don't have the ability to maintain their own temperatures, depending on the neonatal teams to keep them warm," Chhabra said.
The authors hope this new BPD analysis tool will eventually be added to electronic health records, which could give physicians critical advice to enhance BPD care. In addition, Chhabra is looking forward to creating a comprehensive infant database for UC Davis Health, similar to the one she helped develop at Rochester.
"My goal would be to build another deidentified database at the UC Davis NICU," Chhabra said. "This would provide a number of new research opportunities, since it's a completely different population. By collecting and analyzing this data, we can do so much to support more precise care."
