(318) Estimating methadone dose at time of discharge among pregnant people hospitalized for management of OUD

The ability to predict the final dose of methadone accurately and precisely among pregnant people with opioid use disorder (OUD) holds the possibility of improving quality of care. Thus, we evaluated the characteristics of a predictive model aimed at estimating methadone dose at time of discharge among pregnant people admitted for OUD.

Study Design:

This is an analysis of a single-site retrospective cohort of pregnant people admitted for initiation of methadone at an urban, tertiary care center between 2013 and 2022. The primary outcome was methadone dose at time of discharge from initial hospitalization for management of OUD, evaluated both categorically (top quartile of methadone dose) and continuously. Both logistic and negative binomial regression models with bootstrapping were constructed, respectively. Covariates were selected using backwards elimination with p< 0.10 for retention. A receiver operating curve (ROC) was created for the logistic regression model. Derivation and validation cohorts were created. Model goodness-of-fit was determined by use of calibration belts for categorical outcomes and Akaike Information Criteria (AIC) for continuous outcomes. Statistical significance was set at p < 0.05.

Results:

117 people were included. The final model included self-reported dollar amount of heroin consumed and Clinical Opioid Withdrawal Score (COWS) on admission. For top quartile of methadone dose, the model demonstrated a fair classification capacity (area under ROC = 0.75, 95% CI: 0.65-0.86) with evidence of good internal and external validity upon evaluation of the calibration belts (p >0.05, Figure). For gestational latency as a continuous outcome, the model demonstrated similar AIC values in the derivation and validation cohorts (Table).

Conclusion:

Using both self-reported dollar amount of heroin consumed and COWS on admission, we can estimate the final dose of methadone at time of discharge among pregnant people with OUD with fair classification and validity. These data can be used in creation of clinical management protocols aimed at improving quality of care for people with OUD.