237 Views

Neonatal-Perinatal Health Care Delivery: Epidemiology/Health Services Research

Neonatal-Perinatal Health Care Delivery 3: Epi/HSR Antenatal Exposures and Neonatal Outcomes

756 - Performance of different computable phenotypes for the identification of opioid-exposed infants in the electronic health record

Saturday, April 29, 2023

3:30 PM – 6:00 PM ET

Poster Number: 756
Publication Number: 756.246

Andrew D. Wiese, VUMC, BRENTWOOD, TN, United States; Julia Phillippi, Vanderbilt University School of Nursing, Nashville, TN, United States; Alexandra A. Muhar, Vanderbilt University Medical Center – Graduate Medical Education - - Nashville, TN, Nashville, TN, United States; Aleksandra Polic, Vanderbilt University School of Medicine, Nashville, TN, United States; Ge Liu, Vanderbilt University Medical Center, Nashville, TN, United States; Sarah F. Loch, Vanderbilt Center for Child Health Policy, Nashville, TN, United States; Henry H.. Ong, Vanderbilt University Medical Center, Nashville, TN, United States; david A. foutch, Vanderbilt University Medical Center, Oak Ridge, TN, United States; Wu-Chen Su, Vanderbilt University Medical Center, Nashville, TN, United States; Wei-Qi Wei, Vanderbilt University, nashville, TN, United States; Stephen W. Patrick, Monroe Carell Jr. Children's Hospital at Vanderbilt, Nashville, TN, United States

Presenting Author(s)

Andrew D. Wiese, MPH, PhD (he/him/his)

Assistant Professor
VUMC
BRENTWOOD, Tennessee, United States

Background: Computable phenotypes to identify opioid-exposed infants with and without a NOWS diagnosis across multiple institutions are lacking. Thus, short and long-term outcomes among infants with neonatal opioid withdrawal syndrome (NOWS) remain poorly described.

Objective: We sought to validate the performance of different infant and birthing parent electronic health record (EHR)-based computable phenotypes (CPs) for the identification of opioid-exposed infants.

Design/Methods: We developed CPs for the identification of opioid-exposed infants among a population of birthing parent-infant dyads from a single large, academic healthcare system with evidence of a live birth >= 33 weeks gestation without evidence of critical illness or critical congenital malformations (2010-2021). The CPs were based on combinations of 6 EHR-based indicators for opioid exposure including: (1) infant NOWS/opioid-exposure diagnostic code, (2) positive infant opioid toxicology, (3) birthing parent opioid use disorder diagnosis in pregnancy, (4) birthing parent opioid prescription, (5) opioids listed in medication reconciliation, and (6) positive birthing parent urine opioid toxicology. We determined the positive predictive value (PPV) and 95% confidence intervals (CI) for the different EHR-based CPs compared to medical record review among a random sample of dyads identified using the broadest CP (evidence of >=1 EHR-based indicators for opioid exposure). For this random sample, two trained reviewers used a standardized extraction form to manually abstract information from the EHR to identify confirmed opioid-exposed infants.

Results: Among 41,047 dyads meeting inclusion criteria, we identified 1,558 infants (3.80%) meeting the broadest CP for opioid-exposure and 32 (0.08%) meeting the strictest CP [Table 1]. Among the sample of dyads randomly selected for review (n=280), the PPV for the broadest CP was 95.0% (CI: 91.8-97.0) with varying PPVs for the CPs defined based on a combination of infant and birthing-parent indicators (PPV range: 95.0-100.0) [Table 2]. The stricter CPs requiring EHR-based evidence of opioid exposure from the birthing parent and infant records had the highest PPVs, but also had the highest number of confirmed opioid-exposed infants incorrectly identified as unexposed [Table 2].

Conclusion(s):

Opioid-exposed infants can be accurately identified from the EHR. Our computable phenotypes could be used to conduct future epidemiological research examining outcomes among opioid-exposed infants.