Kurt Kroenke, MD
What is the accuracy of unassisted diagnoses of depression by general practitioners (GPs)?
Included studies assessed the accuracy of unassisted GP diagnoses of depression (i.e., without help from severity scales, diagnostic instruments, education programs, or other organizational approaches). Outcomes included sensitivity, specificity, and likelihood ratios.
MEDLINE, EMBASE/Excerpta Medica, PsycINFO, and Scopus (all to April 2009); Science Direct; Ingenta Select; Ovid Full text; Wiley-Blackwell Interscience; and Web of Knowledge were searched for studies that had a sample size ≥ 50 and a reference standard of interview-based diagnoses made by psychiatric experts or researchers conducting structured or semistructured validated interviews. 41 studies (50 371 patients, mean age range 16 to 83 y) met the selection criteria.
Studies showed considerable heterogeneity. Overall, adjusted prevalence of depression was 20%; unadjusted rates ranged from 4% to 50%. Using a random-effects weighted proportion meta-analysis, the overall sensitivity for GP assessments was 47% (95% CI 42 to 53). Diagnoses of depression based on contemporaneous ratings (simple questionnaire) were more sensitive than case-note methods (53% vs 34%, P < 0.001), although such methods using cumulative recognition (clinical impressions over 3 to 12 mo) were more sensitive than those using cross-sectional methods (37% vs 30%, P < 0.001). Results for 19 studies that reported both rule-in and rule-out accuracy show low sensitivity but higher specificity (Table). Based on a 20% prevalence of depression, for every 100 unselected patients GPs would more likely overdiagnose (n = 15) than underdiagnose (n = 10) or correctly diagnose (n = 10) depression.
General practitioners can accurately identify only about 50% of patients who actually have depression but can accurately classify 81% of nondepressed patients.
Diagnostic accuracy of unassisted diagnoses of depression by general practitioners*
*Abbreviations and diagnostic terms defined in Glossary. Results based on random-effects weighted proportional meta-analysis.
Kurt Kroenke. Review: GPs accurately diagnose about 50% of patients with depression and accurately classify 81% of nondepressed patients. Ann Intern Med. 2010;152:JC4–13. doi: 10.7326/0003-4819-152-8-201004200-02013
Download citation file:
Published: Ann Intern Med. 2010;152(8):JC4-13.
Healthcare Delivery and Policy, Hospital Medicine, Prevention/Screening.
Results provided by:
Copyright © 2017 American College of Physicians. All Rights Reserved.
Print ISSN: 0003-4819 | Online ISSN: 1539-3704
Conditions of Use