- How Predictive Analytics, Patient IDs Can Improve the Care Experience
- 3 Ways Predictive Analytics is Advancing the Healthcare Industry
“We estimate that around 100,000 opioid-exposed infants are born each year and many are observed excessively,” said Stephen Patrick, MD, MPH, MS, a neonatologist at Monroe Carell Jr. Children’s Hospital at Vanderbilt and associate professor of Pediatrics and Health Policy.
There is a growing need for advanced tools to help providers manage opioid-exposed infants after they’re born: Researchers stated that in the last twenty years, the number of mothers diagnosed with opioid use disorder increased fourfold, while the rate of newborns with NAS grew sevenfold.
By 2016, one infant was diagnosed with NAS every 15 minutes on average in the US, resulting in more than $500 million in hospital costs.
“Our research group has found that opioid use in pregnancy and NAS has grown substantially over the past two decades,” said Patrick.
“We need to guide some of the initial steps that we do for infants — how we observe them, based upon their risk, as opposed to treating all infants the same. Although the AAP recommends that infants exposed to opioids be observed in the hospital for four to seven days after they’re born, we wondered if it might be possible to discharge infants to home earlier with the right support.”
Researchers designed a predictive analytics model for NAS based on a set of 30 demographic and prenatal exposure covariates collected during pregnancy.
The team came up with a variable based on several different factors, including the infant’s gestational age, whether infants are pre-term, the types of opioids they’ve been exposed to, and if they have been exposed to cigarettes and other drugs.
“We approached it with ‘what does the literature say? What does our prior work say? What does our clinical experience say in terms of which factors may increase risk of drug withdrawal?’ Then we tested those factors to see which things did, in fact, increase the risk of withdrawal,” said William Cooper, MD, Cornelius Vanderbilt Chair in Pediatrics and senior author on the study.
The team developed two predictive analytics models for NAS, a general population model and the other for use with a subset of subjects with Hepatitis C virus infection or opioid exposure in the last 30 days – the higher-risk model.
The results showed that both models discriminated well, with an AUC of 0.89, and were well-calibrated for low-risk infants. These findings show that predictive analytics models could help hospitals accurately determine which infants are at high risk for NAS, preventing low-risk newborns from having to be kept in the neonatal intensive care unit (NICU) after birth.
“Immediately after birth, we often find it difficult to identify which infants with opioid exposure are safe to be discharged. Having a tool like this will help us determine the best type of care to provide in these situations,” said Cooper.
Going forward, the research team will continue to monitor how the tool performs using data collected at Vanderbilt. An interactive web tool is also currently available on the Center for Child Health Policy’s website.
“That’s the next step, before we think about how to apply more widely,” Patrick said.
By Jessica Kent