Diagnostic concordance of substance use disorders in DSM-III, DSM-IV and ICD-10.

Diagnostic concordance of DSM-III, DSM-IV and ICD-10 was tested in a heterogeneous unrestricted sample of 370 clinical cases drawn from a regional consortium. Agreement for abuse/harmful use, dependence, and the collapsed category of 'any diagnosis' was studied across eight drug classes. A probabilistic approach to the cross-classifications based on configural frequency analysis was applied, permitting the computation of four indices of agreement. In contrast to earlier studies, ICD-10 appeared to be the most inclusive system, and often diagnosed cases that were undiagnosed by both DSMs. Generally satisfactory coherence between the ICD-10 harmful use category and the DSM category of abuse was found, but this agreement was often due to a preponderance of negative or undiagnosed cases; disagreement was common on which cases in particular warrant a mild diagnosis. In general, the greatest diagnostic concordance was observed for sedative/hypnotics, opiates and alcohol, the poorest for amphetamines, cocaine and PCP. The analytic approach produced an array of cross-system relationships that are more complex and conditional than those previously reported, and scientists and clinicians are cautioned to study particular drugs, diagnostic levels and measures of concordance before applying cross-system results to their own data or design needs.