Skip to main content

Simulation aiding critically ill newborns heralded as NC's best use of analytics

Tuesday, December 5, 2017
David Tanaka

A discrete event simulation model of the neonatal intensive care unit (NICU) at Duke Children's Hospital took top honors at the 2017 NC Tech Awards, winning the NC Technology Association's Beacon Award for best use of technology in analytics.

The simulation, pioneered by Duke neonatologists and a SAS R&D team, is a first in neonatal care. It is modeled on Duke's level IV NICU, specially equipped to treat critically ill newborns. From the just-born preemie weighing less than a pound to the months-old infant readying for discharge, Duke's neonatal clinicians treat more than 800 babies each year, of which about 40 do not survive.

One lost baby is one too many, so neonatologist David Tanaka, MD, turned to SAS to create a virtual replica of the Duke NICU, where he and his research team could test how changes to various factors affect care and outcomes.

"We were interested in the interaction of several key clinical factors associated with length of stay, mortality and cost," said Tanaka of the analytic simulation, developed by SAS pro bono. "The result was a discrete event simulation model that closely resembled the clinical outcomes of our training unit, validated using data held back from the original model, which also closely tracked actual unit outcomes."

The model uses a vast resource of clinical data to simulate the experience of patients, their conditions and staff responses in a computerized environment. It creates virtual babies experiencing care in a simulated NICU environment, which includes virtual beds staffed by virtual nurses.

"We ultimately entered data from the Neonatal Research Network into the simulation model and examined several hypothetical NICUs created with the worst and best outcomes," said Tanaka. "The resulting outcomes aligned with previous clinical reports from the NRN and demonstrated the importance of improving care to reduce costs and mortality despite increases in lengths of stay."

Based on these findings, the research will now take aim on what the simulation model determined to be the most serious clinical factor affecting neonatal outcomes: necrotizing enterocolitis in its pre-morbid stage. The life-threatening intestinal disease primarily affects premature infants. Using a combination of machine learning, text mining and image analysis, Tanaka's research team is developing a predictive model to identify the condition at its earliest onset to expedite medical interventions and improve outcomes.

"This process represents a milestone effort in the computational analysis of neonatal disease," said Tanaka of the effort. "When successful, the predictive model will provide clinicians a powerful new tool to significantly reduce neonatal mortality, morbidity and costs."

Read the full article