Longitudinal Measurement Invariance of Constructs Derived from the Addiction Cycle

Objective: The Alcohol Addiction Research Domain Criteria (AARDoC) is an organizational framework for assessing heterogeneity in addictive disorders organized across the addiction cycle domains of incentive salience, negative emotionality, and executive functioning, and may have benefits for precision medicine. Recent work found pre-treatment self-report items mapped onto the addiction cycle domains and predicted one- and three-year alcohol use disorder (AUD) treatment outcomes. Given the potential utility of the addiction cycle domains for predicting relevant treatment outcomes, this study sought to evaluate the longitudinal measurement invariance of the domains. Method: We conducted a secondary analysis of individuals with AUD (n=1,383, 30.9% female, 76.8% non-Hispanic White, 11.2% Hispanic) who participated in the COMBINE study. Eleven items assessed at pre- and post-treatment were included in exploratory structural equation modeling (ESEM) and longitudinal invariance analyses. Results: The pre- and post-treatment ESEM models had factor loadings consistent with the three addiction cycle domains and fit the data well. The ESEM factor structure was invariant from pre- to post-treatment and metric invariance (factor loadings) was largely supported, but analyses failed to support scalar invariance (item-level thresholds) of the addiction cycle domains. Conclusions: A three-factor structure representing addiction cycle domains can be modeled using brief self-report measures pre- and post-treatment. Individuals demonstrated a downward shift in level of item endorsement, indicating improvement with treatment. Although this 11-item measure might be useful at baseline for informing treatment decisions, results indicate the need to exercise caution in comparing the addiction cycle domains pre- to post-treatment within persons.