What Would You Do? Cognitive Bias in ECMO Decisions

Important note: every case in this episode is completely fictitious. The scenarios are teaching constructs, invented to illustrate patterns of clinical reasoning. No case describes a real patient, and any resemblance to any individual is entirely coincidental. In ECMO, the hard part is almost never the cannulation. Putting cannulas in and troubleshooting hypoxia are learnable. What's hard is the decision — and there's no textbook for it. In this episode Mike and Calum work through four invented scenarios. In each one, Mike takes Calum to the point where a decision has to be made, asks what he'd do, and then reveals what happened. Every scenario has more than one entirely defensible answer, and in every one a cognitive bias is quietly doing the deciding. We start with a young man who needs ECMO for trauma-related lung injury — and also has a traumatic brain injury. The reflex is that you can't anticoagulate intracranial contusions, but that reflex rests on an assumption modern circuits no longer require, and it crowds out one of the best prognostic groups we ever see. That's base rate neglect. Then a patient at day forty, where a phrase appears in the notes almost daily: "he's been on for forty days." The number has become the argument. Duration isn't a diagnosis — and asking a different question ("what specifically is stopping him weaning?") turns up something entirely fixable. That's anchoring, with availability bias underneath it. The third scenario is the subtlest, because nobody does anything wrong. A young woman is treated for severe pneumonia, correctly. Then she starts bleeding, and it's blamed on the anticoagulation — also a completely satisfying explanation. Two plausible answers in a row, each of which stops anyone completing the diagnosis. That's search satisficing, and it ends with a thirty-second rule you can use tomorrow. The last scenario ends badly, deliberately. A single word in a CT report — "fibrosis" — carries a certainty the imaging doesn't support. A time-limited trial of steroids is agreed, with a review date set in advance. It doesn't work, and the patient dies. And the decision was still reasonable. Judging it by the result would be outcome bias — which leads to the most uncomfortable idea in the episode: when you decline, you almost never find out you were wrong, so the only errors you can see are the ones where you acted. We finish with the full list — availability, conjunction, overconfidence, representativeness, diagnostic momentum, commission bias, the IKEA effect — and the one to end on, the GI Joe fallacy: the tendency to think that knowing about cognitive bias is enough to overcome it. It isn't. So we close with four things that genuinely help, none of which are clever. Chapters (00:00) Cold open — the hard part isn't the cannula (00:50) Why these cases are fictional, and why the reasoning still isn't (01:40) The referee problem: making calls you can't verify (02:20) Case one: trauma lung injury, and a head injury (04:50) Base rate neglect (05:30) Case two: "he's been on for forty days" (07:40) Anchoring to a number rather than a trajectory (08:30) Case three: the bleeding everyone blames on the circuit (11:00) Search satisficing — and a thirty-second rule (11:50) Case four: one word in a CT report (13:40) Outcome bias, and the asymmetry of declining (14:40) The full list of biases (15:30) The GI Joe fallacy, and what actually helps (16:10) Wrap-up Key takeaways The hard part of ECMO is the decision, not the cannula — and there's no textbook for it Base rate neglect: one alarming feature can crowd out a favourable underlying picture A contraindication that feels absolute may be a modifiable risk — ECMO does not obligatorily mean full anticoagulation Duration is not a diagnosis; ask what specifically is preventing weaning A satisfying diagnosis stops the search — and two satisfying explanations in a row are worse In a bleeding ECMO patient, ask whether it's the circuit bleeding or the disease bleeding. Dip the urine A single word in a report can carry more certainty than the evidence behind it A reasonable decision can produce a bad outcome; don't judge the decision by the result When you decline, you rarely learn you were wrong — which should make everyone humbler about saying no Time-limited trials turn an impossible decision into a manageable one Knowing about cognitive bias does not protect you from it References / further reading Croskerry P. From mindless to mindful practice — cognitive bias and clinical decision making. NEJM 2013 Croskerry P. The importance of cognitive errors in diagnosis and strategies to minimise them. Acad Med 2003 Saposnik G et al. Cognitive biases associated with medical decisions: a systematic review. BMC Med Inform Decis Mak 2016 Baron J, Hershey JC. Outcome bias in decision evaluation. J Pers Soc Psychol 1988 Chang DW et al. Evaluation of time-limited trials among critically ill patients. JAMA Intern Med 2021 Kahneman D. Thinking, Fast and Slow. 2011 Extracorporeal Life Support Organization (ELSO) General Guidelines for Adult ECMO This podcast is for medical education for healthcare professionals. It is not clinical advice. All cases are entirely fictitious and were created for teaching purposes only; they do not describe real patients. Always follow your own centre's guidelines and current local policy.