Skip to main content

Identification of empirical subtypes among treatment-seeking alcoholics using latent class analysis

Wednesday, October 26, 2011 — Poster Session III

10:00 a.m. – Noon

Natcher Conference Center




  • L Kwako
  • M Schwandt
  • V Ramchandani
  • D George
  • D Hommer
  • M Heilig


The purpose of the present study was to empirically identify meaningful subtypes based on clinical characteristics in a sample of treatment-seeking alcoholics. The sample included 411 alcohol dependent participants who were admitted to the NIH Clinical Center. We used latent class analysis, including 37 indicators and nine covariates, to identify the clusters among participants. Model fit was assessed by Bayesian Information Criterion. Our analyses identified a three cluster model as providing the most parsimonious and descriptive structure. The clusters were significantly differentiated (at the level of p < 0.001) by 15 indicators. The clusters identified included: (1) Mild (64% of total), characterized by lower alcohol dependence severity, fewer psychiatric problems, lower rates of sexual abuse and psychiatric treatment, lower neuroticism, and lower rates of PTSD and anxiety disorders than the other clusters; (2) Anxious (25%), typified by greater number of alcohol treatments, higher rates of sexual and emotional abuse, psychiatric treatment, and PTSD than the other two clusters, along with lower agreeableness; and (3) Functional (11%), which had lower rates of employment problems than the other two clusters, and higher rates of anxiety disorders, neuroticism, and outpatient psychiatric treatment than the Mild cluster, but lower than the Anxious cluster.

back to top