Background: The most widely used survey measures in psychiatry, the Composite International Diagnostic Interview (CIDI) and the Clinical Interview Schedule-Revised (CIS-R) have generated estimates of psychiatric morbidity that show considerable variation. Doubts have been raised regarding the validity of these structured lay interviewer assessments. There have been no direct comparisons of the performances of these instruments against a common, established criterion.
Method: A total of 105 unselected primary care attendees were each interviewed with CIDI, CIS-R and SCAN in a single sitting with random order of administration. SCAN was administered by a SCAN trained psychiatrist, and CIDI and CIS-R by a public health doctor. Concordance was estimated for all ICD-10 neurotic disorders. We assessed the overall discriminability of the CIS-R morbidity scale using a receiver operating characteristic (ROC) analysis.
Results: The concordance for CIDI for ICD-10 diagnoses was moderate to excellent (kappa= 0.58-0.97). Concordance for CIS-R ranged between poor and moderate (kappa = 0.10-0.65). The area under the ROC curve for the CIS-R morbidity scale with respect to any ICD-10 disorder [0.87 (95% CI 0.79-0-95)] indicated good overall discriminability, but poor sensitivity (44%) and high specificity (97%) at the usual CIS-R cut-point of 11/12.
Conclusion: Among primary care attendees the CIDI is a highly valid assessment of common mental disorders, and the CIS-R is moderately valid. Previous studies may have underestimated validity. Against the criteria of all ICD-10 diagnoses (including less severe depressive and anxiety disorders) a much lower CIS-R cut-point is required than that which is usually advocated.