There has been concern about whether standardized psychiatric interviews make valid diagnoses. Agreements between the Diagnostic Interview Schedule (DIS), as an example of a standardized interview, with independent assessments by a clinician are reasonably high in most studies, but the clinical assessment is itself of uncertain validity. Using predictive ability is an alternative way of judging validity. Data are presented to show that the DIS is almost as good at prediction as a clinician's assessment, but here too there are problems. Because prediction is probabilistic (i.e. the same disorder can have multiple outcomes, and different disorders can share outcomes), it is not possible to say how good prediction has to be to demonstrate perfect validity.
Across varied methods of validity assessment, some disorders are regularly found more validly diagnosed than others, suggesting that part of the source of invalidity lies in the diagnostic grammar of the systems whose criteria standardized interviews evaluate. Sources of invalidity inherent in the content and structure of a variety of diagnoses in DSM-III and its heir, DSM-III-R, are reviewed and illustrated, in part with results from the Epidemiological Catchment Area study.
The relationship between diagnostic criteria and standardized interviews is symbiotic. While attempts to adhere closely to existing diagnostic criteria contribute to the diagnostic accuracy of standardized interviews, the exercise of translating official diagnostic criteria into standardized questions highlights problems in the system's diagnostic grammar, enabling standardized interviews to contribute to improvements in diagnostic nosology.