“What, diagnosis?” Between Bayes and biases

I conducted a personal survey, completely informal, asking groups of doctors during training courses to give me in writing, in a strictly anonymous fashion, a brief operative description of diagnosis. The majority were unfocused and confusing regarding the process and nobody mentioned any probability-based criteria. Objectively, diagnosis is a difficult concept and rich with implications: this is the heart of medical activity and few doctors are able to define it correctly. There exist well-founded reasons to ask ourselves what diagnosis means and how many are the implications and the many environments in which the concept might be declined. We have to accept that even at the end of a complete and exhaustive diagnostic workup, it might not be possible to reach a diagnosis and it is much wiser to admit it rather than giving the patient a label that will abandon them only with extreme difficulty. Whether or not we like it, we behave, even unconsciously, like convinced Bayesians and we should be aware of the fact that the predictive value of your test will vary on how high or low is the presence of the illness. With a very low prevalence, even a very sensitive test might produce an unacceptably high number of false positives. Naturally, it is not necessary to supply a definition of diagnosis to be a good doctor. What counts is that we are aware that the diagnostic process is based on probability and not certainty, we are prone to biases, diagnostic tests can have false positives and false negatives, and having reached a certain threshold of probability and trust in diagnosis we should decide, with our patients, what to do or not do.