A natural classification of alcoholics by means of statistical grouping methods

The results of previous studies using statistical grouping methods for subtyping of alcoholics did not converge to a coherent natural classification, probably among other things as a consequence of improper selection of methods. As an alternative an analytic strategy is formulated and tested. First, a cluster procedure tests a categorical model of the data. If no compact isolated groups are detected non-metric multi-dimensional scaling is used to unravel the complex relations in the data by reconstructing a low dimensional spatial solution.

Cross-sectional study.

In and outpatient treatment programs of the Amsterdam health region in The Netherlands.

A consecutive sample of 277 men and women voluntarily seeking treatment for their alcohol problems. No exclusion criteria were applied.

A sample of 102 symptoms reflecting various aspects of alcoholism measured by structured self-report questionnaire (93 symptoms) or derived from a semi-structured interview taken by treatment staff (nine symptoms).

The cluster procedure failed to show a fitting categorical model. Non-metric multi-dimensional scaling produced a three-dimensional spatial solution. The first dimension reflects the alcohol dependence syndrome. The second bipolar dimension reveals a male-dominated anti-social alcoholism, and a female-dominated pattern of isolated home drinking. The third bipolar dimension represents chronic alcoholism, and young quarrelsome people from troubled families.

The three-dimensional solution obtained has a high face validity and incorporates a number of aspects of previous classifications. The results illustrate the strength of the analytic strategy in unravelling complex symptomatology. Limitations of the classification obtained are stressed and directions for subsequent validation research are given.