The application of cluster analysis to the development of a typology of narcotic addicts

Abstract

A review of the extensive body of literature on narcotic addiction reveals several major flaws. Addiction and related concepts are poorly defined and almost purely theoretical; there is little empirical data reported. The principal weakness is the conceptualization of addiction as a unitary phenomenon; that is, addicts are considered to be a homogeneous group. However, descriptions of addicts and addiction processes show little agreement and the absence of empirical data makes resolution of conflicts in the literature difficult. More recently, the literature has begun to acknowledge a need for an addiction typology, thus showing some recognition of the fact that all addicts are not alike. Most attempts to produce typologies have nevertheless remained theoretical; the resulting types consequently have little meaning. An important exception has been the work of Berzins, Ross, English and Haley (1972) who made use of the cluster analysis of MMPI profiles to identify two replicable types in several samples of addicts civilly committed for treatment. Cluster methodology is widely criticized for imposing order artificially on data such that the clusters generated bear little resemblance to the natural structural relations (if any) in the data. Furthermore, cluster methods vary considerably and the methodological differences between techniques render them differentially sensitive to aspects of the data being clustered. Consequently, in order for a typology resulting from cluster analysis to be accepted with confidence, its relative independence from the method used to produce it must be demonstrated. The present study applied two divergent clustering methods to the MMPI profiles of 209 male addicts civilly committed to the NIMH Clinical Research Center, Lexington, Kentucky. Both sets of cluster arrays produced by this procedure closely resembled the types identified by Berzins et al. (1972). Cases assigned to corresponding clusters in each program were used to form a single set of cluster types maximally independent of the methodological biases of each of the programs. The resulting typology consisted of two homogeneous types and a heterogeneous, unclassified residual. Composite Type I subjects (33%) were characterized by MMPI profile elevations on scales 2, 4, 8, and 7. Composite Type II subjects (13%) were characterized by elevations on scales 4 and 9. Stepwise discriminant analysis produced a major discriminant differentiating the types on the basis of 'sick' versus 'healthy' self-presentation. The composite types were compared on selected social-demographic and personological variables. This exploration of the nonpsychometric characteristics of the type distinction represented an initial assessment of the predictive utility of cluster method addiction typologies. Chi square analysis of the comparisons showed no significant differences in the racial composition or age distributions of the groups, nor were they discriminable on the basis of urban versus small town-rural origins. However, the 'healthy' self presenters were more likely to have had better jobs (p<.O1) and better education (p<.05); they were older when they were first arrested (p<.05) and when they were first addicted (p<.05); they were more likely to have used heroin less frequently than codeine or synthetic prescription drugs (p<.05); but they were much more likely to have sold heroin before they began using it (p<.01). The sick versus healthy self-presentation discriminant was interpreted to support cultural relativistic views of street addiction. However, an integration of the results of the present study with current theories about street addicts indicated that, contrary to the literature, the street addict cannot be considered a homogeneous type. The two composite types of the present study clearly represent markedly divergent motivations for the drug use and drug-sub culture involvement of street addicts. It was concluded that the present investigation generally supports the utility of cluster analysis for the production of a meaningful typology of addiction which is satisfactorily invariant to method. Such an empirically based typology appears to have valuable potential for lending operational clarity to otherwise uselessly inclusive theoretical concepts, for delimiting specific areas of applicability for extant addiction theory, and for bringing some order and meaning to the generally chaotic addiction literature.