Is underdiagnosis the main pitfall when diagnosing bipolar disorder? YesBMJ 2010; 340 doi: http://dx.doi.org/10.1136/bmj.c854 (Published 22 February 2010) Cite this as: BMJ 2010;340:c854
- 1Cardiff University School of Medicine, Cardiff
- 2Tufts University School of Medicine, Boston, Massachusetts, USA
- Correspondence to: D J Smith
Bipolar disorder is a complex condition, and patients can present with the entire range of psychiatric symptoms.1 Its underdiagnosis has always been, and continues to be, a major problem.
Ever since Emil Kraepelin defined manic depressive insanity as recurrent mood episodes (either mania or depression, but not necessarily requiring mania) in 1898,2 the condition has been underdiagnosed. For almost a century it was commonly misdiagnosed as schizophrenia.3 In 1980, the American Diagnostic and Statistical Manual (DSM-III) narrowed the definition of schizophrenia and divided manic depressive insanity into two groups: broadly defined major depressive disorder and narrowly defined bipolar disorder.4 Reasons for underdiagnosis since then include lack of insight into mania by patients, lack of systematic assessment of mania by clinicians, stigma, and the aggressive marketing of antidepressants. The broadening of the definition of bipolar disorder to include hypomania in 1994 and marketing of new drugs have begun to address a century of relative neglect; yet, perhaps predictably, objections about overdiagnosis have been raised.
Let’s start by agreeing on some definitions. Overdiagnosis implies that a disorder is frequently diagnosed when absent, as well as when present. Underdiagnosis, conversely, implies that a disorder is often not diagnosed when present and also infrequently diagnosed when absent. Underdiagnosis and overdiagnosis are claims of validity (based on a definitive diagnosis from pathological findings or, in psychiatry, a formal research diagnostic interview5) not reliability (whether clinicians agree on what they have diagnosed). We should therefore start with certain knowledge about diagnosis (from a research interview, not clinicians’ diagnoses) and assess the accuracy of earlier diagnoses, not the other way around. Stated simply, poor reliability means only that clinicians could not agree on the diagnosis. It says little about true rates of misdiagnosis.
In Zimmerman and colleagues’ widely cited study,6 43% (63/145) of patients who answered “yes” to a questionnaire item (“Have you been diagnosed with bipolar or manic depressive disorder by a healthcare professional?”) subsequently had bipolar disorder diagnosed by formal research interview. This suggests only that the earlier diagnoses were unreliable (which is unsurprising given that these diagnoses were made by different health professionals using different diagnostic criteria over long periods of time). This kind of disagreement among clinicians and researchers does not prove overdiagnosis because it does not begin with findings on validity.
Validity and reliability
We carried out the required validity analysis based on the research interview data from that same study. According to the formal research interview, 90 patients had bipolar disorder and 610 did not.6 Thirty per cent (27) of patients with true bipolar disorder had not been previously diagnosed whereas only 13% (82) of those without bipolar disorder had previously had the condition diagnosed. In effect, bipolar disorder had been missed more than twice as frequently as it was mistakenly diagnosed (relative risk 2.23, 95% confidence interval 1.53 to 3.25). Unfortunately, low reliability in this study was reported as evidence of overdiagnosis when in fact the (validity) data suggest underdiagnosis.6
We have found no study reporting overdiagnosis that started with definitive diagnoses and compared these with earlier clinical diagnoses. Studies that have reported overdiagnosis in substance abuse7 8 or adolescents9 actually describe only low reliability because, as with the Zimmerman study, they start with a relatively unreliable measure of diagnosis (routine clinical practice) and then compare this with the formal research assessment.
In contrast, four studies that used the validity approach all identified underdiagnosis (40-67% of patients with bipolar disorder were previously misdiagnosed).10 11 12 13 These include detailed case series studies of adults with mood disorders10 11 and substance abuse12 and in children.13 Although sample sizes in these studies were smaller (pooled n=271 versus n=700 in the Zimmerman study), they were not less rigorous in design because all misdiagnosis studies must retrospectively assess previous diagnoses. Taken together, these studies (as well as the Zimmerman study when fully and correctly analysed) are strong evidence of underdiagnosis of bipolar disorder.
Low diagnostic reliability is not unique to bipolar disorder because all psychiatric diagnoses have, at best, only moderate reliability14 and complex psychiatric disorders may be difficult to diagnose without necessarily implying overdiagnosis. Diagnostic disagreement is common in other disorders that are known to be underdiagnosed—for example, asthma,15 dementia,16 chronic cholecystitis,17 and coeliac disease.18 These are all, like bipolar disorder, unreliably diagnosed.
In summary, although clinicians’ diagnoses of bipolar disorder can be unreliable, this does not contradict the reality that if patients truly have bipolar disorder they are more likely to be missed than correctly diagnosed. Underdiagnosis, not overdiagnosis, remains the major problem for bipolar disorder within contemporary clinical practice.
Cite this as: BMJ 2010;340:c854
Competing interests: All authors have completed the unified competing interest form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare (1) no support from any company for the submitted work; (2) no relationships with companies that might have an interest in the submitted work in the previous 3 years; (3) their spouses, partners, or children have no financial relationships that may be relevant to the submitted work; and (4) no non-financial interests that may be relevant to the submitted work.