Predictive validity of the Addiction Severity Index's composite scores in the assessment of 2-year outcomes in a methadone maintenance population.

The current research tested the predictive validity of 6 of the 7 Composite Scores (CSs) of the Addiction Severity Index (ASI) in 310 methadone maintenance patients assessed at treatment entry using the ASI and other measures, and followed for 2 years. Logistic regression was used to estimate the sensitivity and specificity of the CSs at intake in predicting their respective validity criterion measures at follow up. Except for the Medical CS, each of the other 5 CS measures significantly predicted its validity criterion measure. The CSs measuring drug use, alcohol abuse, psychopathology, and legal problems had high specificity (88% [corrected] for the Drug CS, 91% for the Alcohol CS, 96% for the Legal CS, and 100% for the Psychological CS). The CSs measuring employment problems had high sensitivity (76%). The results support the predictive validity of most of the ASI CSs as measures of specific problems as well as the validity of the multidimensional construct on which the ASI is based.