A systematic look at bipolar diagnostic accuracing found that if you stick to the DSM-IV diagnostic rules, more than half of all patients now being considered “bipolar” are “over-diagnosed”.Zimmerman But Dr. Zimmerman also found a 30% underdiagnosis rate, and an 86% accuracy rate for bipolar diagnosis overall. So the story is not as bad is it sounds. Yet according to other experts, we’re asking the wrong question, with all this over- versus under-diagnosis talk. We should be asking “how bipolar are you?” and using a different system to characterize the answer. Then there wouldn’t be “overdiagnosis”. There would be ineffective diagnosis.
A recent research study found that over half a group of patients who had been told they had bipolar disorder did not meet official standards for that diagnosis. This result was announced widely, including in the Los Angeles Times (May 7, 2008):
Bipolar Disorder Overdiagnosed?
Study Shows Many People Who Are Told They Have the Disorder Don’t Meet Standard Criteria
That’s odd. Until recently, most research seemed to show that bipolar disorder was underdiagnosed, not over-diagnosed.
Here’s what Zimmerman and colleagues found (and published). Seven-hundred psychiatric outpatients were asked if they’d ever been told they had bipolar disorder. All 700 also met with a trained interviewer who used a well-respected tool called the Structured Clinical Interview for Diagnosis (SCID), which applies the official rules for diagnosis in a systematic way. The SCID was used as the standard for determining whether a patient really had bipolar disorder, or not. The results that got all the press are shown below.
As you can see, 82 patients who’d been told they had bipolar disorder did not actually have it, according to the SCID. For 63 patients, the SCID diagnosis confirmed what the patient had been told. Conclusion: over half of the patients had been “overdiagnosed” with bipolar disorder.
Also found underdiagnosis
Not reported in the press was the next finding, though this was fairly presented in the research team’s paper:
Here we see the percentage of patients who were accurately detected as having bipolar disorder, 63 out of 90. Another 27 patients had bipolar disorder, according to the SCID. But they had not been recognized as bipolar. Thirty percent of the bipolar patients had been missed: “underdiagnosis.”
Overall, reasonably accurate?
Finally, one can look at the folks who don’t have bipolar disorder as an entire group, to get an idea of what percentage of these folks were incorrectly diagnosed as having bipolar disorder. For that, look at the numbers in red below:
Of all the patients who didn’t have bipolar disorder according to the SCID, 86% were correctly diagnosed. This is far from perfect, but it is in the range of some lab tests used in other areas of medicine. Not so bad.
Over- or under-diagnosis? Wrong question
The real problem is, we’re asking the wrong question. The term “overdiagnosis” itself strongly suggests that bipolar disorder is a thing, that you either have or do not have. Granted, that is how our DSM diagnosis system works, but it was never intended to imply that bipolar disorder is a thing. It is an agreement among members of a committee. The reality is more complex. For example, a recent genetics research team found 226 genes with some connection to bipolar disorders.Nurnberger Imagine all the tiny variations you could create by mixing up 266 different genes (its 35,000 permutations. And that’s if each gene only had one variation; some have several).
By contrast, at Harvard’s mood clinic, they ask a different question: “how bipolar are you? how much bipolarity do you have?” (as shown in this interview with their director, back in 2004). I think that’s the direction we ought be pursuing. How many bipolar factors do you need to have before antidepressants are so unlikely to work, or so likely to cause significant problems, that you should start with a different treatment instead? Diagnosis should move toward the way they’ve been doing it at Harvard for over 10 years, using their Bipolarity Index.