Clinicians and Cognitive Bias: A Case of Frontotemporal Dementia Misdiagnosed as Conversion Disorder

Every specialty of medicine is concerned with misdiagnosis. Recent developments in behavioral economics offer a helpful perspective on nonrational influences or cognitive biases that intrude on important decision making. The case presented here highlights how cognitive bias can affect both diagnosis and treatment of patients.

In hindsight, the slow, progressive course of the patient’s illness corresponded with what is known about the natural history of frontotemporal dementia associated with a C9orf72 genetic mutation. It is often associated with psychiatric symptoms and very subtle (or even absent) brain atrophy on MRI, which can make the clinical picture confusing (2). Several times during the 3-month hospital course, different approaches and diagnostic considerations might have led to an earlier correct diagnosis but would not have changed the outcome. These missed opportunities include 1) obtaining details of all prior assessments, including detailed reports, not just summaries, of prior neuropsychological testing; 2) pausing and reconsidering diagnoses when any abnormality was noted, including the mild CSF changes along with EEG changes; and 3) acknowledging earlier that the patient’s hospital course was not what one would see with conversion disorder, with or without a comorbid major depressive disorder. In particular, the clinical findings of the case did not present positive evidence of incompatibility between the patient’s symptoms and a recognized medical or neurological condition, which is a necessary criterion in DSM-5 for diagnosing conversion disorder. For example, we misattributed the significance of Ms. A’s speech peculiarities. Her speech while under our care was always sparse in production; had it been decidedly variable across testing occasions, it would have been stronger evidence in favor of a conversion diagnosis (3).

In 1965, Slater (4) estimated a misdiagnosis rate of 33% for “hysteria” (later renamed conversion disorder). His conclusion was to denounce the diagnosis itself, writing that “the diagnosis of ‘hysteria’ is a disguise for ignorance and a fertile source of clinical error. It is, in fact, not only a delusion but also a snare” (4). In contemporary research, the misdiagnosis rate is estimated to be much lower, approximately 4%, which is similar to rates for other psychiatric and neurological disorders (5). However, diagnostic uncertainty and risk of error remain, as evidenced by reports of cases of neurological or medical illness mistakenly called conversion disorder (6–13) and cases of conversion disorder mistakenly called a medical or neurological illness (14, 15), as well as combinations of psychiatric and neurological disorders in the same patient (16).

Many hypotheses have been proposed to explain conversion disorder misdiagnoses. Among them are the practical clinical challenges of applying the diagnostic criteria for conversion disorder (8, 17), the “subtle interplay” of biological and psychological models of neuropsychiatric illness (6), limited access to advanced imaging (11), current customs of neurological evaluation of functional neurological symptom disorders (9, 18, 19), and a limited neurobiological understanding of functional neurological disorders (7, 17). The contribution of cognitive biases leading to misdiagnosis of conversion disorder has been briefly noted in several reports (7, 11), although without substantial elaboration.

Cognitive Heuristics

In 1974, Kahneman and Tversky (20) described a series of experiments that demonstrated consistent, predictable patterns of how humans simplify complex cognitive tasks such as assessing probabilities and making predictions. They called these patterns “cognitive heuristics,” which are mental shortcuts that allow people to make decisions efficiently. These shortcuts are considered adaptive features of human cognition that are generally highly effective, although they can also lead to systematic errors in judgment, called “biases.” In the medical literature, the study of cognitive heuristics and resultant biases has been a focus of growing interest (21, 22). We review pertinent cognitive heuristics originally described by Kahneman and Tversky—representativeness, availability, adjustment and anchoring—and consider how they may help explain the diagnostic error in this case.

Representativeness

Representativeness refers to estimating probability based on similarity. Errors arise when the idea that “A looks like B” overwhelms other important but more complex considerations, such as estimates of prior probability, sample size, or statistical power. Kahneman and Tversky illustrate this concept by showing how stereotypical personality traits influence predictions about professions (20). For example, when presented with a stereotypical personality sketch of a shy and fastidious person drawn from a sample of more lawyers than librarians, most people will predict that the person is a librarian, in effect prioritizing the professional stereotype over the sample distribution and violating the rules of prior probability. However, when asked the same question in the absence of a personality sketch, we use prior probability estimates correctly. Kahneman and Tversky also observed that people have greater confidence in their decisions as the degree of similarity increases, even when the stereotype itself is known to be “unreliable or outdated” (20).

With regard to this case, representativeness appears to have contributed to diagnostic error because of the historical association between conversion disorder, traumatic stressors, and unexplained neuropsychiatric symptoms. As recently as the 4th edition of DSM, the diagnostic criteria for conversion disorder included a temporal relationship with a traumatic stressor (23). The fact that the patient’s unexplained symptoms appeared after her son’s suicide rendered them representative of an antiquated but common understanding of conversion disorder and likely also conferred a high degree of confidence in the diagnosis despite recent changes to the diagnostic criteria intended to address such errors. DSM-5 removed the criterion requiring association with a traumatic stressor, relegating it to the accompanying text because of concern that it had “not been shown to be either diagnostically reliable or predictive of outcome” (24, 25). DSM-5 also added the new criterion requiring positive evidence of symptoms incompatible with a known medical illness—not just the absence of an alternative diagnosis.

Availability

The availability heuristic refers to assessing the probability of an event on the basis of the ease with which the event can be brought to mind. This can be mediated by frequency (in medicine, known colloquially as “common things are common”) but also by other factors, including the salience of a concept or narrative. For example, Kahneman and Tversky (20) showed subjects lists of names with equal gender distribution but with different ratios of famous men to famous women. When the subjects were asked whether there were more male or female names on the list, the answers corresponded to the gender proportion of famous names: when more famous men made the list, it appeared that there were more male names; when more famous women made the list, it appeared that there were more female names.

In this case, the patient’s symptoms appeared to coincide with the tragic loss of a son to suicide. That particular historical detail was noted in nearly every evaluation during the course of the patient’s care, which stands in contrast to other milestones that were noted with less frequency and which arguably were just as important to the diagnostic evaluation, such as onset of depressive symptoms (5–8 years before Ms. A’s son died) and the point at which speech production became dramatically disturbed (1.5 years after her son died). In retrospect, it appears that the powerful cognitive “availability” of a child’s suicide may have overwhelmed the completeness of the differential diagnosis.

Adjustment and Anchoring

The central tenet of adjustment and anchoring is that the starting point influences the ultimate decision. Different starting points yield different end points, and errors are biased toward the starting point, not the endpoint (20). An example of a cognitive bias stemming from this heuristic is “premature closure,” which refers to when physicians insufficiently consider alternative diagnoses once a provisional diagnosis is made. In the medical literature on diagnostic error, premature closure is among the most common biases leading to misdiagnosis (26).

In this case, the treating physicians were tasked with not only arriving at the right diagnosis but also rejecting an incorrect one. Diagnostic anchoring contributed from the outset in this case, starting with the patient’s initial neuropsychological testing, where speech and memory deficits were misinterpreted as stemming from functional impairment associated with the patient’s son’s death. A pattern of insufficient adjustment ensued, and despite extensive further evaluations, the correct diagnosis remained elusive while the initial incorrect diagnosis remained in place.

Implications For Clinical Practice

To improve the diagnostic accuracy of “functional” neurological disorders, Friedman and LaFrance (19) propose that we elevate functional neurological illness to the same level as other neurological disorders, list them in the differential diagnosis, enumerate pertinent positives and negatives, and rule in the diagnosis with evidence, not just as a diagnosis of exclusion, as was also proposed in DSM-5. In the literature on the science of decision making, this approach is described as cognitive de-biasing. Kahneman and Tversky (20), however, suggest that de-biasing efforts are insufficient, and that we should also incorporate familiarity with cognitive heuristics and biases into our judgments. In other words, the question to ask as we consider our diagnostic decisions is not only “How do we avoid cognitive shortcuts that cause errors?” but also “What cognitive shortcuts might help us prevent errors?”

With those lessons in mind, we propose that recent changes to the diagnosis of conversion disorder be viewed as not just a technical list of diagnostic criteria, but also a “choice architecture” (27) for diagnostic decisions. In a case of suspected conversion disorder, consideration should start with identifying evidence of incompatibility between a specific symptom and recognized neurological or medical conditions (criterion B of the DSM-5 diagnostic criteria).

Other aspects of conversion disorder phenomenology can be helpful in further characterizing specific patterns of psychopathology but also confer an elevated risk of diagnostic error when used as a starting point, in part because of biases stemming from representativeness, anchoring, and availability.

Conclusions

The case presented here—a 47-year-old woman with a 2-year history of progressive, debilitating neuropsychiatric symptoms after her son’s suicide, initially misdiagnosed with conversion disorder but later found to have frontotemporal dementia—highlights the importance of identifying positive evidence of incompatibility between a symptom and a recognized neurological or medical illness when diagnosing conversion disorder. The study of cognitive heuristics may be used to help understand and prevent diagnostic errors by identifying and critically assessing mental shortcuts that can lead to misdiagnosis and delay appropriate treatment decisions.