Composite international diagnostic interview (CIDI-auto)

The CIDI is a structured clinical interview using criteria from the Diagnosis and Statistical Manual of Mental Disorders (4th edition) (DSM-IV) and International Classification of Diseases (ICD) to determine the diagnostic status of respondents [26]. The CIDI is the most widely used of the available measures to estimate population rates of DSM/ICD mental illnesses.

The CIDI has good test-retest reliability and good agreement with clinical diagnoses for a number of diagnostic categories [27]. Early validation used a criterion (gold standard) of two clinicians who agreed on a diagnosis then compared this with CIDI-Auto diagnoses [28]. The sensitivity and specificity of the CIDI-Auto varied by diagnostic category with major depression and panic disorder having good sensitivity and moderate specificity. Generalised anxiety disorder had poor sensitivity but moderate specificity. A larger scale primary health care sample has compared CIDI diagnoses with the WHO semi-structured SCAN interview [29]. This study finds good sensitivity and specificity (but poorer concordance) for depression, panic disorder and generalised anxiety disorder. A similar study using population data and the Structured Clinical Interview for DSM-IV as the criterion reports moderate sensitivity and high specificity for a wide number of anxiety disorders and major depressive disorder [30]. The modest level of reliability associated with CIDI-Auto diagnoses may reflect the nature of psychiatric diagnoses (the state of the science) more than the use of the CIDI. Thus studies of diagnostic reliability using both forensic samples [31, 32] and psychiatric patient samples [33] point to reliability estimates (test-retest) which are of the same order suggested when CIDI-Auto diagnoses are assessed against clinical interviews. Studies using specifically briefed and trained interviewers [28] provide good supportive evidence for the validity and reliability of the CIDI-Auto and modest levels of sensitivity and specificity suggested by other studies need to be interpreted in the context that clinical diagnoses in population settings are affected by similar limitations.