Background. We aimed to evaluate the diagnostic accuracy of a highly structured diagnostic interview in relation to a semi-structured diagnostic procedure. We compared the World Health Organization Composite International Diagnostic Interview Short Form (CIDI-SF) in diagnosing major depressive episode (MDE) to consensus diagnoses based on the SCAN interview (Schedules for Clinical Assessment in Neuropsychiatry).

Method. Subjects comprised a follow-up sample of 239 20–24-year-old former high-school students who were administered the SCAN and immediately thereafter the CIDI-SF. Concordance was estimated for 12-month MDE, using different cut-points of the CIDI-SF and for any affective disorders.

Results. Correspondence between instruments was moderate for MDE (κ = 0.43, sensitivity 0.71, specificity 0.82), but better for any affective disorder (κ = 0.60, sensitivity 0.70, specificity 0.90). Most false negatives suffered from their depression as much as those correctly identified by the CIDI-SF. False negativity was mainly due to not endorsing the stem questions of the CIDI-SF. Of the false positives almost half had an affective disorder other than MDE.

Conclusions. The CIDI-SF seems to function best in identifying a broader category of affective disorders. It could be useful in large-scale community surveys where more extensive psychiatric interviews are not feasible.