OBJECTIVES:

Medication for Opioid Use Disorder Treatment (MOUD) has poor adherence and retention. Premature discontinuation from MOUD has significant negative health and societal consequences. Our objective is to develop machine learning and causal models to investigate how variations in treatment type influence treatment discontinuation. The findings of this research help guide evidence-based strategies for improving treatment retention.

METHODS:

We used the Substance Abuse and Mental Health Services Administration's (SAMHSA) Treatment Episode Data Set in Discharges (TEDS-D) from 2015-17. The data was filtered by the patient’s opioid use history, demographics, treatment modality, and any other relevant information. Machine learning models (Decision Trees, Random Forest and XGBoost) were developed to predict treatment discontinuation. Outcome Adaptive Elastic Net and Nearest Neighbor Matching were used to evaluate the Causal relationship between treatment discontinuation and treatment modality that combines MOUD and self-help group participation.

RESULTS:

The machine-learning models revealed a strong association between treatment discontinuation and participation in self-help groups, frequency of substance use, length of stay in the treatment, and states with a predictive accuracy of 82.34% for patients with OUD. Our causal hypothesis tests show an average treatment effect on treated (ATT) of 0.079. This finding can inform the development of more effective strategies to reduce the associated treatment discontinuation rate and combat the OUD epidemic.

CONCLUSIONS:

Our study demonstrates the importance of involvement in self-help groups and MOUD in OUD treatment. We found that participation in MOUD and self-help groups caused a lower risk of treatment discontinuation compared to MOUD alone. This suggests that policymakers should consider the addition of self-help groups to OUD treatment to improve the rate of adherence and retention. Further research is needed to investigate how MOUD and self-help groups can be tailored to fit individual needs and help people across various demographics.