An investigation of the feasibility of predicting outcome indices in the treatment of heroid addiction.

An exploratory study employing multiple linear regression techniques was conducted to investigate the feasibility of predicting outcome indices in the treatment of heroin addiction. Demographic and biographic intake variables were obtained for heroin addicts prior to treatment in one of two different types of residential treatment programs. Length of treatment, further categorized into "split-stay" categories, was chosen as the criterion because of its importance from the standpoint of program management. Subsequent stepwise regression procedures resulted in 11 variables of sufficient predictability to account for approximately 30% of the variability in the split-stay criterion. The employment of the derived equation resulted in correct classification in approximately 75% of the cases. This was beyond what could have been predicted by chance alone or from the knowledge of the observed split-stay ratio. A clear demonstration of the feasibility of the model was obtained; however, as to its utility, implementation in the actual treatment setting would be required. Implications for the use of this type of prediction model were discussed.