Personality dimensions related to premature termination from an inpatient drug abuse treatment program

Comparison of 135 program completers and 65 dropouts from an inpatient drug abuse detox/rehab. unit on 29 variables (MMPI validity and clinical scales, 13 Wiggins Content Items, number of Grayson Critical Items selected, MacAndrews Alcoholism Scale, and Age) revealed few group differences with univariate analyses. Multivariate and stepwise discriminant function analyses resulted in selected variables that achieved 64-87% stay/leave classification accuracy, depending on method of application. However, when these functions were cross-validated on independent samples within the same program at 3-year intervals, the classification accuracy generally was reduced by 50% or more. Results suggest that (a) there are few individual MMPI indices that predict patients at risk for program attrition; and (b) "good" multiple regression equations may not generalize to independent samples across time. Conceptualizing the dropout process ontogenetically will facilitate further experimental refinement and clinical understanding.