Estimating measurement error in alcohol dependence symptomatology: findings from a multisite study

Inter-rater test-retest reliability of alcohol diagnoses and symptom ratings, made by means of a modified version of the Structured Clinical Interview for DSM-III-R (SCID), were evaluated in the context of a multisite clinical trial of alcoholism treatment. Reliability coefficients for the subject's 'worst period' and the 'current period' were compared. The results show that with proper training, reliable diagnostic classification can be achieved across multiple sites. Symptom reliabilities for the current period were higher than the worst period. Respondents with higher discrepancies between test and retest tended to drink more on each drinking occasion, reported more medical detoxifications, changed residence more often, had lower occupational status and were rated by the interviewer as less attentive, motivated and intelligent. Methods to monitor and improve the reliability of diagnostic interviews in addiction research, particularly in multisite studies, are discussed in relation to the findings.