The alcohol use disorder and associated disabilities interview schedule (AUDADIS), was designed for use in the general population, and was previously shown to have good reliability in a sample of household residents. However, measurement problems are different in clinical samples. Thus, a test-retest study was conducted of the AUDADIS in a clinical sample of 296 substance-using patients from substance- and psychiatrically-identified treatment settings. Reliability for current drug-specific AUDADIS dependence diagnoses was good to excellent for high-prevalence as well as low-prevalence drug categories. Reliability for abuse diagnoses was not as good, although this was due to the hierarchical nature of the abuse diagnosis itself, rather than its defining criteria. Demographic and other factors were investigated for their potential effects on the reliability of alcohol and cocaine diagnoses; low severity was the only consistent predictor of unreliability for both of these categories. Reliability of consumption variables was generally good, with a few notable exceptions. Results suggest that the AUDADIS can be used in research comparing treated to community samples of individuals with alcohol and drug diagnoses.