Intuitvely, patient diagnosis seems like a pardigmatically bayesian problem. I'm interested in the extent to which explicit (or implicit, if it can be established) bayesian reasoning informs patient diagnosis in clinical settings. Is there any documentation from medical associations which state that bayesian approcahes to diagnosis are best practices? Do medical textbooks recomend bayesian reasoning?


P.S, for context, I'm interested in studying medical practice as a potential example of succesful bayesian reasoning in complex and individualistic cases. I'm exploring analogies to issues in legal inference, where conceptually similar concerns about evidence arise.

  • 1
    Bayesian analysis is explicitly used in clinical genetics occasionally to assess risk for genetic disease. However, this question, as written, strikes me as extremely broad, as practices will very widely based on specialty and country(among other things). What would you expect an answer to provide? A systematic survey of tens of thousands of providers across countries and nations asking about their application of Bayesian approaches?
    – Ian Campbell
    Jul 28, 2022 at 15:08
  • Not necessarily a survey. I was hoping for instances of medical professional associations or medical training materials describing it as a best practice, or, failing that, some study (again, not necessarily a survey) of clinician behaviour that indicated it was frequently used. Jul 28, 2022 at 15:18
  • 1
    The implicit side of this is pretty basic diagnosis 101. en.wikipedia.org/wiki/Zebra_(medicine) is one example of this concept. I don't know much about medical education having not gone down that path myself, but I'd expect that students would encounter Bayesian ideas from the more general concept of prevalence rather than "priors". Of course as I'm sure you're familiar, Bayesians like to promote that the Bayesian approach in general approximates human reasoning (with of course many well-known psychological biases departing from this).
    – Bryan Krause
    Jul 28, 2022 at 15:52
  • I'd argue that every time a differential diagnosis is used, Bayesian approaches are more or less implicit (a la @BryanKrause 's zebra) though I see that the literature isn't strong on this. Of interest: cognitiveresearchjournal.springeropen.com/articles/10.1186/…
    – bob1
    Jul 28, 2022 at 23:27
  • Ok, I found this paper by Fuller and Flores which gives an account of prediction in medicine. It's pretty helpful. doi.org/10.1016/j.shpsc.2015.06.006 Aug 1, 2022 at 23:13


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.