Which instruments to support diagnosis of depression have sufficient accuracy? A systematic review

Abstract

Background: Instruments are frequently used in case finding, diagnosis and severity grading of major depression, but the evidence supporting their utility is weak. Aim: To systematically review the specificity and sensitivity of instruments used to diagnose and grade the severity of depression. Methods: MEDLINE, PsycInfo, Embase and the Cochrane Library databases were searched until April 2014. Fifty studies fulfilled the inclusion criteria. Risk of bias was assessed with QUADAS. The average sensitivity and specificity of each instrument was estimated with hierarchical summary receiver operating characteristics analyses and the confidence in the estimates was evaluated using GRADE. Minimum acceptable sensitivity/specificity, with structured interview as the reference, was 80%/80% for structured interviews and 80%/70% for case-finding instruments. The minimum acceptable standard for severity measures was a correlation of 0.7 with DSM-IV classification. Results: Twenty instruments were investigated. The average sensitivity/specificity was 85%/92% for the Structured Clinical Interview for DSM-IV-Axis-I Disorders (SCID-I), 95%/84% for the Mini International Neuropsychiatric Interview (MINI), < 70%/85% for the Primary Care Evaluation of Mental Disorders (PRIME-MD), 88%/78% for the Patient Health Questionnaire-9 (PHQ-9) with a cut-off score of 10, 69%/95% for PHQ-9 as a diagnostic algorithm and 70%/83% for the Hospital Anxiety and Depression Scale (HADS) with a cut-off score of 7. The confidence in the estimates for the other instruments was very low. Conclusions: Only the SCID-I, MINI and PHQ-9 with a cut-off score of 10 fulfilled the minimum criteria for sensitivity and specificity. The use of the PRIME-MD and HADS is not supported by current evidence.

• Depressive disorder
• Diagnosis
• Evidence
• Interview
• Meta-analysis
• Questionnaires
• Psychiatric status rating scales
• Psychological standards
• Standards

Acknowledgments

We thank Cecilia Svanborg, M.D., Ph.D. and Professor Maria Tillfors, who assessed the structured interviews and BDI-II, Professor Måns Rosén, who gave valuable comments throughout the manuscript process, Derya Akcan and Agneta Brolund, who performed the literature search, and Lars Berglund at Uppsala Clinical Research Center, who performed the HSROC calculations.

Disclosure of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.