Background. Exclusion criteria are present in almost all diagnostic categories in both DSM-IV and ICD-10. These exclusion criteria state that one diagnosis is not made if it is ‘due to’ another disorder. However, there is little empirical evidence demonstrating that the hierarchy imposed by the exclusion criteria is meaningful. The current study examines associations between ten common mental disorder pairs to determine whether they are higher for disorder pairs that are classified in DSM-IV and/or ICD-10 as having a hierarchical relationship, than for those that are not hierarchically related.

Method. Data were analysed from the Australian National Survey of Mental Health and Wellbeing, a large (N = 10641) epidemiological survey of mental disorders. Bivariate odds ratios between disorder pairs were calculated from logistic regression analyses. Multivariate odds ratios were also calculated from separate logistic regression analyses in which the sample was restricted to likely positive cases, and co-morbid mental disorders and neuroticism scores were both controlled.

Results. The odds ratios between disorder pairs related according to the exclusion criteria were higher than those of unrelated disorders, for both DSM-IV and ICD-10. When constraints were placed on the data to control for the effects of co-morbid disorders and neuroticism the same pattern was evident in DSM-IV, but not in ICD-10.

Conclusion. The patterns of association between disorder pairs found in the exclusion criteria for DSM-IV appear to more closely mirror the patterns of association found in epidemiological data. While this does not guarantee that the exclusion criteria are valid it is argued that identifying significant associations should be the first step in establishing meaningful exclusion criteria.