Predicting Disclosure Among People With Opioid Use Disorder: A Machine Learning Approach

Abstract

Disclosure can be an important step in treatment and recovery for people with opioid use disorder. In this longitudinal study, we used a Machine Learning-based Lasso logistic regression model to develop an accurate prediction model for disclosure among people with opioid use disorder. Participants included n = 145 individuals in recovery from opioid use disorder who planned to disclose to k = 190 recipients. The Lasso logistic regression model predicted disclosure with 72% accuracy. Feature importance analysis identified approach goals, avoidance goals, and relationship length as the most important predictors of disclosure. Approach goals led to more disclosure and avoidance goals led to less disclosure. Moreover, longer relationship duration between the disclosure and recipient was associated with a higher likelihood of disclosure. Our prediction model can help researchers and healthcare providers create interventions to support the disclosure process among people with opioid use disorder and improve treatment outcomes.