Kieran Walsh: Finding your way back from the wrong diagnosis

A 40 year old man has a cough—but his GP cannot find out the cause. Eventually the patient is referred to the local hospital where he is diagnosed with asthma. The doctor who sees him starts inhalers. But the inhalers don’t work and so he goes back for another appointment. The next doctor that he sees increases the dose of the inhalers. But this doesn’t work either so eventually the patient starts on oral steroids. These don’t help either. Finally the patient has some further tests and is diagnosed with gastro-esophageal reflux. He receives treatment for this, stops his inhalers, and gets better.

This is a fictional scenario, but stories like this happen all the time in real life. We can look at the story from a number of perspectives—but one way of looking at it is from the perspective of cognitive bias. In this case the cognitive bias is likely one of anchoring or diagnosis momentum. Once a patient receives a diagnostic label, it is difficult to remove and replace it. In this scenario the doctor thought the patient had asthma and more doctors continued down this diagnostic and treatment pathway until they ran out of road. They eventually got to the correct diagnosis but clearly could have got there faster. This story ended fairly happily with a delayed diagnosis of a benign and treatable condition. But the patient could just as easily have had lung cancer and then the lesson learned would have had a bitter aftertaste.

So what can we do to stop this happening? Maybe if we educate doctors about cognitive biases, then they would be more likely to recognise their own cognitive biases and less likely to make mistakes. Right? Wrong. The evidence shows that education to reduce cognitive biases has little or no effect on diagnostic errors. This paper by Norman et al has just been published and so opens up thinking in this area for the rest of 2017. So if education to stop cognitive biases won’t help, what will? This is a complex problem and the answer to it is complex, nuanced and dependent on circumstances.

Let’s look at the story again—this time from the perspective of knowledge based strategies. Let’s say that the doctor stopped when the patient didn’t respond to the inhalers. Let’s say that they stopped and analysed the information that they had and thought about what else the diagnosis could be and maybe identified new knowledge that would help them. And then they thought about the differential diagnosis of asthma and came up with the correct diagnosis. The evidence shows that this strategy is more likely to work. So knowledge deficits are important.

How we help with knowledge is a whole different story. Healthcare professionals need knowledge that is current, reliable, and evidence based and at the same time will help them quickly answer the specific clinical question that they have. But too often they don’t get this type of knowledge. Rather they get too much knowledge that is unwieldy, that answers other people’s questions rather than those of their patients and that is sometimes not evidence based.

However the area is even more complex than this—there are a range of nuances and caveats. Healthcare professionals are short of time and sometimes don’t have time to stop and think and look things up. This must be addressed. But in the meantime knowledge must be available at the point of care and in a format that healthcare professionals can use. This in turn depends on circumstances—but is likely to mean on a mobile or app or whatever device is being used.

The literature is currently pointing back to knowledge as a vital resource that can reduce error and improve the quality of care. Providers of knowledge can help by changing how they deliver knowledge and ensuring that they work to high standards and, as importantly, to the right standards. In this regard, currency, evidence base and applicability are all important.

Kieran Walsh is clinical director of BMJ Learning—the education service of the BMJ Group. He is responsible for the editorial direction of BMJ Online Learning, BMJ Masterclasses, and BMJ onExamination. He has written two books—the first on cost and value in medical education and the second a dictionary of medical education quotations. He has worked in the past as a hospital doctor – specialising in care of the elderly medicine and neurology.


  1. NormanGR, Monteiro SD, Sherbino J, Ilgen JS, Schmidt HG, Mamede S. The Causes of Errors in Clinical Reasoning: Cognitive Biases, Knowledge Deficits, and Dual Process Thinking. Acad Med.2017 Jan;92(1):23-30.

Competing interests: KW works for BMJ, which produces BMJ Best Practice – an evidence-based clinical decision support tool.

  • David Kvaratskhelia

    “Healthcare professionals are short of time and sometimes don’t have time to stop and think and look things up. This must be addressed. But in the meantime knowledge must be available at the point of care and in a format that healthcare professionals can use.” – The core of the cases, even for imaginary ones. Thank you, Kieran

  • David Levine

    Great blog Keiran! Correcting diagnostic error first requires recognition of the error but, at least with hospital shift patterns, trainees often don’t find out whether their diagnostic efforts were correct or not and this makes improving difficult. Easy access to stored knowledge is indeed vital, as is eliciting the initial knowledge from patients but choosing the right knowledge and being able to use it depend, critically, on diagnostic reasoning skill and I think this is rather different to knowledge.

    Worryingly, at least until 2015 (and possibly to date), only a handful of UK Medical Schools offered formal training in diagnostic reasoning in their curricula although excellent courses are run elsewhere. As with any skill, frequent, deliberate practice with feedback of results is essential to develop expertise.

    Avoiding errors in the first place is preferable and although the limited evidence about teaching on cognitive error has been disappointing perhaps better results might be obtained
    using more memorable and accessible prompts. You and I might find the neuroscience fascinating but it can seem a bit dry for novices. I can drive a car reasonably well and safely without knowing much about physics or mechanical engineering but I rely on a few key rules (eg always looking in the mirrors and to the
    right, in the UK, before overtaking) to avoid serious error. Similarly, in Medicine, we want to minimise diagnostic error from the start and perhaps a ‘simple’ rules-based approach might be more successful for novices.

    Such heuristics have to be memorable and can sometimes be based on popular literary quotes or other succinct maxims to embed important precepts. Some of these long predate the
    neuroscience but have stood the test of time and can still be applicable to day when used in the right contexts. 1.2

    Anchoring and premature closure probably account for more
    diagnostic mistakes than most other cognitive errors and can be circumvented by the discipline of always asking the simple question ‘could this be anything else?’ before a single diagnostic
    possibility is accepted. Reports that doctors are often lack time to think should cause alarm. Ironically, the least experienced doctors are the most likely to have their diagnostic skills degraded by being pushed into working too fast and by frequent interruptions.


    1) Levine D. Revalidating Sherlock Holmes for a Role in Medical Education. Clinical Medicine 2012; 12, No 2: 146–9

    2) Levine D, Bleakley A. Maximising Medicine through
    Aphorisms Medical Education 2012; 46:153–162

  • Jennifer Liddell

    I understand the reasoning behind the need to reconsider initial diagnoses, however I take issue with the case described. As any GP will tell you a persistent cough is an extremely common presentation. It is not clear in this case what investigations have been done or why he was referred to a specialist (in our area probably an 8 month wait). The more likely scenario would have been the GP considering all the causes, treatingthe most likely one and reviewing. This is more likely where there is continuity of care. GPs are generally used to uncertainty and constantly having to reconsider causes. I would have believed the case much more had it been an -ologist in an unusual presentation, with the GP then putting new symptoms down to the original obscure diagnosis.