The effect of sample selection on the distinction between alcohol abuse and dependence

Martin Steppan, Daniela Piontek, Ludwig Kraus


Steppan, M., Piontek, D., & Kraus, L. (2014). The effect of sample selection on the distinction between alcohol abuse and dependence. The International Journal Of Alcohol And Drug Research, 3(2), 159-168. doi:

Aim: The effect of sample selection on the dimensionality of DSM-IV alcohol and dependence (AUD) criteria was tested applying different methods.

Sample: Data from the 2006 German Epidemiological Survey of Substance Abuse (ESA) were used. A mixed-mode design was used (self-administered questionnaires and telephone interviews), and 7,912 individuals, aged 18 to 64 years, participated. The response rate was 45%. Alcohol abuse and dependence were assessed according to DSM-IV, based on the Munich Composite International Diagnostic Interview (M-CIDI). Inter-item correlations, Confirmatory Factor Analysis (CFA), and Latent Class Analysis (LCA) were applied to the total sample (unrestricted sample, URS) and a subsample of individuals with at least one endorsed criterion (restricted sample, RS). Latent Class Factor Analysis (LCFA) was performed using the RS, including covariates (age, sex, education).

Findings: The mean inter-item correlation was higher in the URS than in the RS. When individuals without criterion endorsement were excluded, factor analyses resulted in more dimensions. In the RS, LCA yielded an interaction between abuse, dependence and class membership. The LCFA identified two dimensions and five classes corresponding to abuse and dependence.

Conclusions: Sample selection has a critical effect on dimensionality analyses. When individuals who do not endorse a single criterion are excluded, the bi-axial factor structure of the DSM-IV (abuse and dependence) can be supported. However, there is also evidence that a further diagnostic category should be included or that the threshold for dependence should be lowered.


alcohol use disorders; DSM-IV; dimensionality analysis; factor analysis; latent class analysis; latent class factor analysis

Full Text:




  • There are currently no refbacks.