a1 Department of Health Care Policy, Harvard Medical School, Boston, MA, USA
a2 Department of Psychiatry, University Hospitals Case Medical Center, Case Western Reserve University, Cleveland, OH, USA
a3 Clinical Services, EPI-Q, Inc., Oakbrook Terrace, IL, USA
a4 School of Medicine, University of Washington, Seattle, WA, USA
a5 Lindner Center of HOPE and Department of Psychiatry, University of Cincinnati College of Medicine, Cincinnati, OH, USA
a6 Depression Clinical and Research Program and the Bipolar Clinic and Research Program, Massachusetts General Hospital and Harvard Medical School, Boston, MA, USA
a7 Columbia University School of Social Work, New York, NY, USA
a8 Department of Psychiatry, University of California San Diego, San Diego, CA, USA
a9 Departments of Psychiatry, University of Pennsylvania School of Medicine, Philadelphia Veterans Affairs Medical Center, and the University of Pittsburgh Medical Center, Philadelphia and Pittsburgh, PA, USA
a10 Institute of Clinical Psychology and Psychotherapy, Technische Universität Dresden, Dresden, Germany
Background Lack of coordination between screening studies for common mental disorders in primary care and community epidemiological samples impedes progress in clinical epidemiology. Short screening scales based on the World Health Organization (WHO) Composite International Diagnostic Interview (CIDI), the diagnostic interview used in community epidemiological surveys throughout the world, were developed to address this problem.
Method Expert reviews and cognitive interviews generated CIDI screening scale (CIDI-SC) item pools for 30-day DSM-IV-TR major depressive episode (MDE), generalized anxiety disorder (GAD), panic disorder (PD) and bipolar disorder (BPD). These items were administered to 3058 unselected patients in 29 US primary care offices. Blinded SCID clinical reinterviews were administered to 206 of these patients, oversampling screened positives.
Results Stepwise regression selected optimal screening items to predict clinical diagnoses. Excellent concordance [area under the receiver operating characteristic curve (AUC)] was found between continuous CIDI-SC and DSM-IV/SCID diagnoses of 30-day MDE (0.93), GAD (0.88), PD (0.90) and BPD (0.97), with only 9–38 questions needed to administer all scales. CIDI-SC versus SCID prevalence differences are insignificant at the optimal CIDI-SC diagnostic thresholds (χ2 1 = 0.0–2.9, p = 0.09–0.94). Individual-level diagnostic concordance at these thresholds is substantial (AUC 0.81–0.86, sensitivity 68.0–80.2%, specificity 90.1–98.8%). Likelihood ratio positive (LR+) exceeds 10 and LR− is 0.1 or less at informative thresholds for all diagnoses.
Conclusions CIDI-SC operating characteristics are equivalent (MDE, GAD) or superior (PD, BPD) to those of the best alternative screening scales. CIDI-SC results can be compared directly to general population CIDI survey results or used to target and streamline second-stage CIDIs.
(Received June 21 2012)
(Accepted September 05 2012)
(Online publication October 18 2012)