Evidence of the Construct Validity and Potential Application of a Multivariate Approach to Alcoholism Diagnosis

Abstract

To investigate the construct validity of a proposed 10-factor alcoholism diagnostic profile, a principal components analysis was performed on the responses of 1,535 alcoholic inpatients to a questionnaire designed to assess functioning in a number of life areas. A total of 16 components, explaining 50% of the total variance, were retained for rotation and interpretation. On this basis it is recommended that consideration be given to additional factor areas in the proposed profile. It is suggested that factor profiles from the proposed model might be of use in formulating predictions regarding treatment completion. To evaluate the likelihood of this suggestion, a stepwise discriminant analysis was performed using scale scores for the 16 components and the additional variable of age in an attempt to differentiate between treatment completers and premature discharges in the sample. The discriminant function which resulted contained 12 of the 17 variables available and served to correctly identify 62.5% of the premature discharges. Cautions are expressed regarding generalizations from the results of both analyses. It is concluded that evidence was obtained for both the construct validity of the factors in the proposed model and of the potential clinical use of the resultant profiles. Although substantial research and development is required, the proposed model seems to offer numerous advantages over the diagnostic approaches currently available.