Solving the Mystery of Medical Mistakes
Why do medical errors occur? A better understanding of the brain mechanisms involved in diagnosis can help answer the question – and avoid future issues
At a Glance
- A new functional magnetic resonance imaging study shows that diagnosing disease engages the same brain mechanisms used to identify and name everyday objects
- Highly diagnostic information at the beginning of assessment may predispose to premature diagnostic closure, an important cause of error
- A dramatic shift in brain activity between making and speaking decisions suggests that doctors need to hear themselves, either silently or aloud, to become aware of their diagnostic decisions
- The above holds true for simple diagnoses; the next step is to investigate the brain’s activity during more complex decision-making processes
Diagnostic error. It’s something we’re all aware of, but hesitant to discuss (1)(2). Whether we talk about it or not, though, there’s one obvious question to address: how can we reduce the rate at which it occurs? Perhaps the key lies in understanding why we might make mistakes during diagnosis. What takes place in the brain between seeing a patient or sample and reaching a conclusion? Until now, we haven’t known – but new functional MRI (fMRI) methods have helped increase our understanding of the neural processes behind diagnostic decision-making (3)(4). We spoke to Marcio Melo, lead author of the study, to find out more…
Anatomy of a diagnosis
Schematically, the traditional view of the diagnostic process is:
- Diagnostic cues evoke diagnostic hypotheses.
- The investigative process (including history taking, physical examination, and diagnostic testing) confirms or disproves each hypothesis in the differential diagnosis through an interactive process, until a final conclusion is reached.
- Based on that conclusion and any additional relevant information, a treatment is chosen.
But the decision-making process is not as simple as it looks, according to Melo. There are actually many, many choices made in the course of a single medical assessment. After the patient presents with the initial complaint, each step in the diagnostic investigation demands a decision: what question should I ask next? What exam should I order? With the degree of uncertainty we can expect from a diagnostic assessment, especially in its early stages, the physician needs to decide at each point in the process which of the many possible paths to choose.
Even when we factor in all of these branching points, we haven’t fully encompassed the complexity of the process. It’s certainly not as linear in real life as it looks on paper. For instance, the more complicated the disease, the more likely it is that a doctor will have to take a few steps back along the diagnostic pathway to find the right route. And, of course, we often find ourselves jumping ahead as well as back – the study’s results indicate that treatment alternatives are often evoked during the diagnostic assessment, before a final diagnosis has even been reached. It’s clear that the real process does not follow the standard description.
But as powerful as fMRI is, that level of complexity would have made it difficult to observe what was actually going on in the brain during the process of making a diagnosis. So, to test their hypotheses using fMRI, Melo and his colleagues needed participants to follow a simplified diagnostic process based on only key information – allowing them to investigate the brain mechanisms underpinning that process. But Melo believes it is worth keeping in mind that, even in everyday medical practice, things are much more complicated than they appear here…
What did we see?
First, that the way the brain identifies and names objects in everyday life is very similar to the way it tackles a tough diagnosis (see Figure 1). Second, that the more conclusive information the brain has at the start of the diagnostic process, the less attention it’s likely to pay to the rest of the process, meaning that vital information may be overlooked – an important cause of diagnostic errors. And third, that there’s a distinct shift in brain activity between reaching a conclusion and speaking it aloud (see Figure 2) – suggesting that we may, in fact, become more aware of our own conclusions simply by verbalizing them.
Why does a complex process like disease diagnosis resemble something so seemingly simple as naming objects? Identifying and naming things in everyday life is, in fact, a highly complex neurocognitive process; its triviality in our lives is what gives it the illusory appearance of simplicity. In domains where identification and naming demand a high level of expertise – like, for instance, botany – it becomes clearer that it poses quite a challenge for the brain. Conversely, disease diagnosis doesn’t always seem complicated, either; many illnesses are easily identified by laypeople who have familiarity with them. After a child has had a few cases of tonsillitis, most parents have already guessed the problem by the time they’ve made a doctor’s appointment.
And so, the initial hypothesis was that diagnosing diseases was similar to identification and naming in everyday life. Other people had already proposed that diagnosis is a categorization process, but the team’s study on radiological diagnosis was the first experimental demonstration that the brain systems involved in diagnosing and naming were similar. Radiology seemed to them the obvious place to start because, at the time, the neural basis of naming things using visual stimuli had already been extensively investigated with fMRI.
Talk it out
While observing the diagnostic process from start to finish, Melo’s team encountered one completely unexpected finding – a large-scale switch of brain activity between the end of the decisional period and the beginning of the vocalization of the responses. It turns out that mentally making the diagnosis involves quite a different pattern of brain activity to actually speaking the diagnosis out loud. The activity shift takes place in several brain structures related to both awareness and auditory monitoring; careful analyses led them to postulate that this is the neural correlate of the doctors’ becoming aware of the diagnoses they were making. If that conclusion is correct, then it’s something that inevitably occurs every time a physician makes a diagnosis, right or wrong. From a broader perspective, this is how we become aware of our own thoughts. It’s a counterintuitive conclusion – we’ve been taught to expect that we first think, and only then speak our thoughts out loud, but it seems that perhaps it’s the speaking that actually materializes the thought.
It has been proposed that being reflective (for instance, by talking to oneself) could help decrease diagnostic error by increasing physicians’ awareness of their thought processes and conclusions. The supporting evidence for this approach is inconclusive, though, so more research is needed to determine whether or not the benefits are worth the time investment. After all, it’s difficult to pause and reflect when your consultations have an average duration of 10–15 minutes, as is common in primary care.
A better way to diagnose
So how can all of this knowledge improve diagnostic accuracy? Or, put another way, what should pathologists and other diagnostic professionals learn from Melo’s work? The researchers observed that highly specific information at the beginning of the diagnostic process – reducing uncertainty about the final diagnosis – decreases activity in brain regions involved in attention. This reduction of attentional monitoring may cause doctors to end their diagnostic investigations too soon, and could even lead to missing significant abnormalities. A similar process, called “satisfaction of search,” takes place in the visual domain when a diagnostician finds one prominent abnormality, which increases the probability of missing other lesions. But how can we overcome such biases? Melo’s suggestion is to present a list of differential diagnoses at the beginning of the diagnostic assessment instead of later in the process – thereby increasing uncertainty regarding the final conclusion and potentially preventing premature diagnostic closure. This can be implemented by employing user-friendly computerized diagnostic support systems if you have access to them. “I firmly believe,” says Melo, “that tools like these can help guide the diagnostic process while reducing errors and biases, and will become increasingly helpful in improving medical accuracy.”
Until now, the team’s studies have focused on situations in which the diagnosis is straightforward. Their next step will be to investigate the brain mechanisms involved in conditions that require more complex diagnostic processes. The investigation and understanding of the neural processes physicians and healthcare professionals employ in making diagnoses and prescribing treatments are in their earliest stages. “I’d like to see them examined in much more detail,” says Melo, “but we’ll need more resources devoted to these kinds of studies. I think that, as our understanding of disease advances and our brain visualization technologies improve, studies like ours will become increasingly important in helping us better comprehend the inner workings of the diagnostic process – and mitigate medical error.”
Marcio Melo is a psychiatrist affiliated with the Laboratory of Medical Investigations, Faculty of Medicine of the University of São Paulo, Brazil.
- N Miller, “We Need to Talk: Pathologists, Patients and Diagnostic Errors – Part I”, The Pathologist, 20, 18–29 (2016). Available at: bit.ly/2991Tbz.
- N Miller, “It’s Our Turn to Talk: Pathologists, Patients and Diagnostic Errors – Part II”, The Pathologist, 21, 18–33 (2016). Available at: bit.ly/2sbX4q1.
- M Melo et al., “How doctors generate diagnostic hypotheses: a study of radiological diagnosis with functional magnetic resonance imaging”, PLoS One, 6, e28752 (2011). PMID: 22194902.
- M Melo et al., “How doctors diagnose diseases and prescribe treatments: an fMRI study of diagnostic salience”, Sci Rep, 7, 1304 (2017). PMID: 28465538.
While obtaining degrees in biology from the University of Alberta and biochemistry from Penn State College of Medicine, I worked as a freelance science and medical writer. I was able to hone my skills in research, presentation and scientific writing by assembling grants and journal articles, speaking at international conferences, and consulting on topics ranging from medical education to comic book science. As much as I’ve enjoyed designing new bacteria and plausible superheroes, though, I’m more pleased than ever to be at Texere, using my writing and editing skills to create great content for a professional audience.