Alcohol Use Disorder (AUD) presents in a continuum of severity, with distinct profiles exhibiting unique drinking behavior, motivations, consequences, and neurobiological underpinnings (Citation1). This heterogeneity leads to diverse treatment outcomes, posing a challenge to developing effective, personalized intervention strategies. Recognizing this heterogeneity, contemporary research in substance use has adopted person-centered clustering methodologies to identify homogeneous subgroups with differential responses to targeted interventions (Citation2). The conceptualization of AUD as a heterogeneous disorder with distinct profiles has clinical implications. Previous studies using latent class analysis and other clustering methods have identified multiple profiles with variations in symptom severity, comorbid psychopathology, and neurobiological dysfunction (Citation3,Citation4). These findings suggest that a one-size-fits-all treatment approach may be inadequate, and identifying and characterizing profiles could facilitate precision medicine approaches to AUD treatment.

Data-driven clustering approaches have become powerful tools for identifying heterogeneity across diverse medical and psychiatric conditions (Citation5). Unlike hypothesis-driven categorical approaches, clustering algorithms use multivariate patient-level data to discover naturally occurring groupings without preconceived classification schemes, uncovering clinically actionable patient subtypes with distinct prognostic profiles. This study used k-means cluster analysis, an unsupervisedmachine learning algorithm that mathematically groups patients into distinct profiles based on their underlying similarities across baseline biopsychosocial variables to identify clinically meaningful profiles of AUD.

READ THE FULL ARTICLE