Overdiagnosis of Psychiatric Disorders Still Happens

I read an excellent article in Clinical Psychiatry News recently in the Hard Talk section. The title is “A prescription for de-diagnosing” by psychiatrists Nicholas Badre, MD and David Lehman, MD in the July 2022 issue (Vol 50, No. 7).

The bottom line is that too many psychiatric patients have too many psychiatric diagnoses. A lot of patients have conflicting diagnoses (both unipolar and bipolar affective disorder for example) and take many psychotropic medications which may be unnecessary and lead to side effects.

It takes time to get to know patients in order to ensure you’re not dropping diagnoses too quickly. Discussing them thoroughly in clinic or in the hospital is an excellent idea. And after getting to know patients as people, it makes sense to discuss reduction in polypharmacy, which can be quite a burden.

This reminds me of the Single Question in Delirium (SQiD), a test to diagnose delirium by simply asking a friend or family member of a patient whether their loved one seems to be more confused lately. It’s a pretty accurate test as it turns out.

This also reminds me of the difficulty in making an accurate diagnosis of bipolar disorder. I and a Chief Resident wrote an article for The Carlat Report in 2012 (TCPR, July / August 2012, Vol 10, Issue 8, “Is Bipolar Disorder Over-Diagnosed?”) which warned against overdiagnosis of bipolar disorder. Excerpts below:

Some argue that bipolar disorder is actually under-diagnosed. They have support from abundant literature showing that bipolar disorder tends to present more often with depression than mania or hypomania (Judd LL et al, Arch Gen Psychiatry 2002:59(6):530–537). As many as 10% of patients with unipolar depression ultimately are shown to have bipolar illness instead, according to some experts (Goodwin GM et al, Eur Neuropsychopharm 2008:18(7):535–549). (See this month’s Q&A with Claudia Baldassano for more on this.) In addition, a new emphasis on subthreshold mood symptoms and more rapid mood shifts has led some psychiatrists to promote the concept of a “bipolar spectrum disorder” (Youngstrom EA et al, Curr Psychiatry Rep 2010;12(6):479–489).

While it’s important to remain vigilant about a history of manic and hypomanic symptoms, we think the problem of over-diagnosis is probably greater. For instance, in a 2008 study, Zimmerman and colleagues performed a comprehensive diagnostic interview on 700 patients, nearly 21% of who self-reported a history of “bipolar disorder.” However, when using the gold-standard SCID (structured clinical interview), only 13% had the diagnosis; they also had more first-degree relatives with bipolar disorder than the others (Zimmerman M, Ruggero CJ et al, J Clin Psychiatry 2008:69(6):935–940). The authors hypothesized that over-diagnosis of bipolar disorder might be a consequence of efforts to improve recognition of it and avoid under-detection. In fact, the same authors studied 40 depressed patients previously diagnosed with bipolar disorder and found that, by the SCID, they had specific phobia, PTSD, drug abuse/dependence, or a personality disorder instead (Zimmerman M et al, Compr Psychiatry 2010;51(2):99–105).

Over-diagnosis can also occur when apparent mood episodes are defined as psychiatric when in fact, they have a different etiology altogether. Decreased need for sleep, disorganized or racing thoughts, increased activity and agitation, and delusional thinking, even when they occur together, can represent a sort of “final common pathway” for medical conditions and other syndromes. The manic phenotype can occur in patients with agitated delirium, brain tumors, corticosteroid treatment, and of course substance intoxication (Bunevicius A et al, CNS Spectr 2008;13(11):950–958; Brooks JO and Hoblyn JC, Am J Psychiatry 2005;162(11):2033–2038). These other phenotypes can be distinguished by recognition of key features such as the fluctuating nature of consciousness in delirium, neuroimaging findings, and positive urine drug screens.

Unfortunately, physicians may also be susceptible to diagnostic shortcuts. When faced with limited time for diagnostic interviews and the pressure to prescribe by patients and their families, well-meaning clinicians may give the diagnosis after a single brief interview. Not uncommonly, we find that it was diagnosed on the basis of mood fluctuation over minutes, temper tantrums, and fleeting insomnia. The rapidly expanding repertoire of medications approved for bipolar disorder, and their relative ease of use, may also contribute to over-diagnosis. Unfortunately, in some cases the treatment may be worse than the symptoms themselves (Iordache I and Low NC, J Psychiatry Neurosci 2010;35(3): E3–4).

I was accustomed to asking what I called the Single Question in Bipolar (SQiB). I frequently saw patients who said their psychiatrists had diagnosed them with bipolar disorder. I would ask them, “Can you tell me about your manic episodes?”

Often, they looked puzzled and replied, “What’s a manic episode?” I would describe the typical symptoms and they would deny ever having them.

The article by Drs. Badre and Lehman is a bit disappointing in that it doesn’t look as though we’ve improved our diagnostic acumen much in the last decade.

We need to try harder.

Informal Bedside Tests for Delirium

Most of this post is an updated redux from years ago about an informal bedside test for delirium called the oral trails test. I learned about it from my senior resident when I was a junior psychiatry resident in training at the VA Medical Center.

There was an elderly patient admitted to the psychiatry unit who was thought to be psychiatrically ill but who actually seemed confused to me and the senior resident. We consulted medicine in order to get him transferred to the general medicine unit but it was tough going. I think the medicine resident disagreed with our clinical impression that he was confused and didn’t think medical transfer was necessary.

Anyway, my senior resident showed me her version of the oral version of the mixed Trails A and B Test for executive function. There is a written form which is part of the Repeatable Battery for the Assessment of Neuropsychological Status (RBANS). First, she asked him to count from 1 to 7; then she asked him to recite the letters of the alphabet from A to J. She then asked him to try reciting those letters in alternating sequence. Of course, he failed miserably and was eventually transferred to internal medicine. The Trails actually is a paper and pencil test and it looks like a dot to dot game, like the example below:

Trails Test

My senior resident told me she learned the oral Trails test from her senior resident and couldn’t remember anything else about it. I used the test for years but a neuropsychologist criticized the practice, questioning the test’s validity, and rightly so.

Of course, I’d been doing it wrong. You’re supposed to have the patient count to 25, then recite the letters of the alphabet, then recite the numbers and letters in alternating sequence from 1 to 13 and A to M. More than two errors in pairings indicate cognitive impairment.

There are limitations to the verbal Trails and caution is advised in more recent literature, indicating that there are moderate education effects in older patients and that it may be wiser to use both paper and pencil and oral versions together.

Still the search goes on for quick and dirty ways to screen for cognitive impairment in the elderly because this and advanced age are two main risk factors for delirium.

Nowadays, I do the Mini-Cog (shown in the video below) or the Single Question in Delirium (SQiD) test, which just involves asking a family member if the patient seems confused lately.


Mrazik, M., Millis, S., & Drane, D. L. (2010). The oral trail making test: effects of age and concurrent validity. Archives of clinical neuropsychology: the official journal of the National Academy of Neuropsychologists, 25(3), 236–243. doi:10.1093/arclin/acq006

Ricker, J. H., & Axelrod, B. N. (1994). Analysis of an Oral Paradigm for the Trail Making Test. Assessment, 1(1), 47–51. https://doi.org/10.1177/1073191194001001007

Sands, M., Dantoc, B., Hartshorn, A., Ryan, C., & Lujic, S. (2010). Single Question in Delirium (SQiD): testing its efficacy against psychiatrist interview, the Confusion Assessment Method and the Memorial Delirium Assessment Scale. Palliative Medicine, 24(6), 561–565. https://doi.org/10.1177/0269216310371556

SQiD vs CAM Redux

This was a blog post I wrote back in 2011 on another blog, The Practical C-L Psychiatrist. SQiD is short for Single Question in Delirium and it’s a very short and effective screen for delirium, if you have a reliable informant. I also mention the Edinburgh Delirium Test Box (EDTB). It has been further developed into a smartphone app.

“The November Vol. 3 issue of the Annals of Delirium published a summary of an interesting study of a Single Question in delirium (SQiD) as a screen for delirium compared to the Confusion Assessment Method (CAM), the Memorial Delirium Assessment Scale (MDAS) and a psychiatrist interview[1].

The question “Do you think (name of patient) has been more confused lately?” was put to a friend or relative of 21 patients. Compared with psychiatric interview, the SQiD achieved a sensitivity and specificity of 80% (95% CI 28.3-99.49%) and 71% (41.90-91.61%) respectively. The CAM demonstrated a negative predictive value (NPV) of 80% (51.91-95.67%) and the SQiD showed an NPV of 91% (58.72-99.77%). The CAM in the study had only a 40% sensitivity used by minimally trained clinical users.

The negative predictive value of a test tells you how likely it is that you actually don’t have the condition or disease. It’s defined as the number of true negatives (people who test negative who are not affected) divided by the total number of patients who test negative and it varies with test sensitivity, test specificity, and disorder prevalence. The sensitivity of a test is how accurately it detects patients who are positive for the disorder (in this case delirium). If 100 patients are positive for the disorder, then a test that is 80% sensitive will detect 80 of those cases and miss 20 actual cases of the disorder. Specificity is defined as how accurately a test detects patients who do not have the disorder. In our delirium example, if 100 patients are free of the disorder, then a test that is 71% specific will correctly tell 71 of those people that they are not affected and will incorrectly tell 29 that they have the disorder when they don’t.

This seems to suggest that a single question screening question packs a fair punch compared to screening instruments and psychiatric interview for identifying delirium. The CAM takes a few minutes to complete and requires training to achieve optimal identification rates.

The authors suggest the SQiD deserves further study and their results seem to support the conclusion. The study is limited by small sample size, but intuitively the premise is appealing. This is one of the quickest tests for delirium applicable and can be applied by almost anyone.

Single question screening exams for depression are not unheard of so there is precedence for the SQiD. You just have to be careful about what you say in front of patients and families. “Go ahead and run the squid on Mr. Jones” could raise a few eyebrows.

This is possibly a low tech solution in a pinch when the CAM forms file is empty or the battery is low on the Edinburgh Delirium Test Box (EDTB)[2]. The EDTB is a more high-tech solution to testing for what neuropsychologists believe what one of the main abnormalities is in delirium—lack of sustained attention. It’s a computerized neuropsychological testing device.

And that face-off would be called SQiD versus Box.”


1.         Sands, M., et al., Single Question in Delirium (SQiD): testing its efficacy against psychiatrist interview, the Confusion Assessment Method and the Memorial Delirium Assessment Scale. Palliative Medicine, 2010. 24(6): p. 561-565.

2.         Brown, L.J.E., et al., Detecting deficits of sustained visual attention in delirium. Journal of Neurology, Neurosurgery & Psychiatry.

%d bloggers like this: