Procedural validity of the computerized version of the Composite International Diagnostic Interview (CIDI-Auto) in the anxiety disorders

Abstract

The procedural validity of the computerized version of the Composite International Diagnostic Interview (CIDI-Auto) was examined against the consensus diagnoses of two clinicians for six anxiety disorders (agoraphobia, panic disorder (±agoraphobia), social phobia, simple phobia, obsessive compulsive disorder (OCD), generalized anxiety disorder (GAD) and major depressive episode (MDE)). Clinicians had available to them all data obtained over a 2- to 10-month period. Subjects were 98 patients accepted for treatment at an Anxiety Disorders Clinic, thus, all subjects had at least one of the diagnoses being examined. While the CIDI-Auto detected 88·2% of the clinician diagnoses, it identified twice as many diagnoses as did the clinicians. The sensitivity of the CIDI-Auto was above 0·85 except for GAD, which had a sensitivity of 0·29. The specificity of the CIDI-Auto was lower (range: 0·47–0·99). The agreement between the CIDI-Auto and the clinician diagnoses, as measured by intraclass kappas, ranged from poor (k = 0·02; GAD) to excellent (k = 0·81; OCD), with a fair level of agreement overall (k = 0·40). Canonical correlation analysis suggested that the discrepancies between the CIDI-Auto and clinicians were not due to different diagnostic distinctions being made. It is suggested that the CIDI-Auto may have a lower threshold for diagnosing anxiety disorders than do experienced clinicians. It is concluded that, in a sample where all subjects have at least one anxiety disorder diagnosis, the CIDI-Auto has acceptable validity.