A comparison of the predictive validity of four sets of baseline ASI summary indices

Abstract

This study compared the long-term predictive validity of original and new baseline Addiction Severity Index summary scores in methadone patients. The indices included the original Interviewer Severity Ratings (ISRs) and the new Clinical Indices (CIs), which use both lifetime and recent problem information, and the original Composite Scores (CSs) and Evaluation Indices (EIs), based on recent problems only. Outcomes were medical hospitalization, employment, alcohol intoxication, drug hospitalization, and psychiatric hospitalization in Months 7-24 poststudy entry and criminal charges in Months 0-24. Hierarchical logistic regression analyses were used in which 1 index was entered first and the other in the 2nd step. The reverse order of entry was used in a 2nd analysis. A final analysis set compared the best predictor from each of the 2 prior analysis sets. The CIs were superior to the other indices in predicting 3 of 6 outcomes (psychiatric hospitalization, drug hospitalization, and criminal charges); the EI was the best predictor of alcohol intoxication, and the CS the best predictor of unemployment.