Interviewer severity ratings and composite scores of the ASI: a further look

The psychometric characteristics of the Addiction Severity Index (ASI) interviewer severity ratings (ISRs) and composite scores (CSs) were examined for a newly trained group of interviewers. The interrater reliabilities of seven raters for 9 methadone maintenance (MM) pilot subjects were determined. These were found to be excellent for the CSs, but only moderate for the ISRs. Regression analyses were performed on the data of 407 methadone maintenance patients entering all of the variables in each area as the independent variables and the ISR as the dependent variable. These analyses indicated that, on average, 55-60% of the variance in ISRs was explained. However, while 63% of the variance was explained for the legal scale only 38% of the variance of the drug scale was explained. The subject's rating of either the need for treatment or the seriousness of the problem accounted for the most variance, with the exception of the drug and alcohol areas. The internal consistency of the composite scores was then examined for this sample using Cronbach's standardized alpha statistic. These were found to be generally satisfactory ranging from 0.62 for the drug scale to 0.87 for the alcohol and psychiatric scales. A mean interitem correlation of 0.11 for the drug scale was obtained suggesting relatively low item homogeneity. Finally, correlations between the CSs and ISRs were calculated for each scale. Moderate to high relationships were found (0.53-0.78) with the exception of the correlation for the employment area which was only 0.08. The findings are discussed in terms of the nature of and limitations of the ISRs and CSs.