Evidence-Based Answers

Evidence Central™ is an integrated web and mobile solution that helps clinicians quickly answer etiology, diagnosis, treatment, and prognosis questions using the latest evidence-based research.


Evidence Central for Mobile Devices

Evidence Central iOS iPhone iPad Android

Evidence Central from Unbound Medicine, available for iOS® and Android™, is optimized for each platform and features superior navigation, so answers are easy to find at the bedside or anywhere they’re needed. Learn More

Word of the Day

PCP diagnosis of depression: not very accurate

Clinical Question:
How reliable is unassisted clinical diagnosis of depression in primary care?

Bottom Line:
Primary care physicians are not very accurate in the unassisted clinical diagnosis of depression. (LOE = 2a)

Mitchell AJ, Vaze A, Rao S. Clinical diagnosis of depression in primary care: a meta-analysis. Lancet 2009;374(9690):609-619.  [PMID:19640579]

Study Design:
Meta-analysis (other)

Outpatient (primary care)

These authors searched multiple databases looking for studies of unassisted primary care physician diagnosis of depression: no severity scales, no diagnostic instruments, no education programs, and no organizational approaches to diagnosis. The "gold standard" diagnosis was determined by psychiatric interview using DSM criteria. The authors don't describe using paired, independent assessment of study eligibility or a search for unpublished studies. The authors excluded studies with fewer than 50 patients. (This exclusion is becoming more popular and I don't approve. Throwing away data just doesn't seem right to me. And one of the goals of meta-analysis is to look at the totality of data to develop more robust estimates of treatment effects, rates of events, and diagnostic accuracy. Although a large study may swamp the contribution of smaller studies, it typically takes an unusally enormous study to do this.) These authors included 118 studies with more than 50,000 patients. Overall, slightly less than 20% of the patients were depressed, and the combined sensitivity and specificity for clinician diagnosed depression was 50.1% (95% CI, 41.3% - 59.0%) and 81.3% (74.5% - 87.3%), respectively. These translate to a positive likelihood ratio of 2.67 (1.6 - 6.4) and a negative likelihood ratio of 0.61 (0.21 - 0.79). In other words, the unassisted clinical diagnosis isn't very accurate.


Site Licenses

Site licenseSite Licenses are available for schools, universities, hospitals, government agencies, and companies. For more information, contact us.