Taking Baby Steps with AI, Starting in the NICU

Tharika Arunkumar - Pediatrics Senior Editor

Link: https://www.nature.com/articles/s41598-023-32570-z

Typically, newborns stay in the hospital with their mother until they are discharged. Unfortunately, if there are complications, newborns are admitted to the Neonatal Intensive Care Unit (NICU) for critical care. This often happens to babies that are born premature (before their due date) as they haven’t had sufficient time to develop within the uterus. When babies are hospitalized there is an increased risk of malnutrition, unfortunately the likelihood that preterm babies while in the NICU will experience malnutrition is around 80 percent. To address this issue, clinicians have traditionally monitored each NICU patient’s weight gain and loss in order to identify abnormal patterns of growth. However, with growing provider shortages, it is becoming increasingly challenging to monitor each infant’s weight closely enough to prevent malnutrition. As a result, this study explored the use of machine learning to identify malnutrition in any baby, allowing clinicians to then take action after the problem is identified.

This study involved 512 ICU patients aged 0-28 days. Clinicians collected routine data on the patients - blood pressure, temperature, urine output, and birth weight - as well as more advanced tests when needed. The data was collected daily and analyzed using the open-source software R. After analyzing the data, the team created various graphs and classification models to predict weight gain and discharge weight for NICU patients. The results of the study showed that machine learning was able to predict the presence of weight gain at discharge and discharge weight for NICU patients. It may serve as a decision reference for neonatologists, dietitians, or nurses to adjust the level of care a NICU patient is receiving. Ultimately, these tools may help clinicians discover malnutrition more efficiently in NICU patients and improve their overall care.

Resources:

Yalçın, Nadir, et al. “Development and Validation of Machine Learning-Based Clinical Decision Support Tool for Identifying Malnutrition in NICU Patients.” Nature News, Nature Publishing Group, 30 Mar. 2023, https://www.nature.com/articles/s41598-023-32570-z.