Keywords: Neonatal opioid withdrawal syndrome, Dynamic linear mixed models, Morphine, Real world data
Background: Neonatal opioid withdrawal syndrome (NOWS) is a drug withdrawal syndrome that neonates exposed to opioids in-utero may experience after birth with atleast 75% of them requiring treatment, with morphine being most commonly used drug. Currently, the morphine dosing adjustments are often empiric or step-wise in nature; associated with longer hospital stays. The aim of the study is to use an objective, real world data-driven approach to optimize morphine dosing in neonates with NOWS to improve clinical outcomes. Methods: Longitudinal morphine dose and clinical response (Modified Finnegan Score (MFS)) data along with maternal and infant baseline factors were collected using a retrospective cohort design from the electronic medical records of infants with NOWS (N=189) admitted to the Neonatal Intensive Care Unit at the University of Maryland Medical Center (2013-2017). A dynamic linear mixed effects (DLME) model with independent covariance structure was used to develop the relationship between MFS and morphine dose adjusting for baseline risk factors. Simulation based model evaluation was performed by comparing the observed and predicted using clinically meaningful metrics such as time on treatment. Results: Maternal methadone use, poly-substance drugs, race, previous morphine dose were significant predictors of MFS. Significant positive autocorrelations of previous two MFS with current MFS was observed (?_1=0.72 and ?_2=0.25 respectively). On an average, for a 100mcg increase in morphine dose, the MFS decreased by 0.5 units. The model evaluation showed that observed and predicted time on treatment (median: 11.0 vs 9.8 days) was not significantly different (p=0.28). Conclusion: A model based framework was developed to describe the MFS–morphine dose relationship using real world data. Further improvements to the DLME model are underway to develop an adaptive, precision dosing strategy for infants with NOWS that could lead to better clinical outcomes.