A. Arteriosclerotic renal artery stenosis would fit the picture A.
Hence, there is reason to suspect that A is true. (after Peirce 1998, 231)
In other words, clinicians begin by knowing nothing definitive about the origin of the signs and symptoms and search by conjecturing hypotheses that, if true, would explain the symptoms (Stanley and Campos 2013, 301-302). For example, a physician may auscultate a dyspneic patient’s chest for crackles. She reasons that if the patient had pulmonary congestion, then crackles would be a predictable consequence. Therefore, she conjectures that the patient may have pulmonary congestion. The specific choice of the type of congestion is based, of course, on the relevant evidence available at the time: e.g. cardiac murmur, breathlessness, previous history of cardiac disease, etc. Similarly, uncontrollable hypertension, in spite of the usual cause, idiopathic, would suggest a different etiology.
Importantly, however, Peirce emphasizes these hypotheses are only asserted “problematically or conjecturally.” The mere fact that the hypothesis explains the symptoms is not yet a reason to accept it. Rather, the conjectured hypotheses are proposals to be subjected to further testing. For Peirce, as Kapitan (1997, 477) points out, the purpose of abduction is both to generate hypothesis but also to “recommend a course of action” by selecting hypotheses for further examination. This course of action may be likened to an internal question and answer format (auto-elenchus.): considering these observations and tests, what are the likely diagnoses.
For any given set of symptoms, physicians may often conjecture many possible diagnoses; not all of these can be pursued at once. The second type of reasoning involves clinician having to prioritize one or perhaps a few hypotheses for further testing. In other words, the clinician has to rank the hypotheses in terms of their pursuitworthiness (McKaughan, 2008, Nyrup 2015). While it is often important to consider the most likely hypotheses, Peirce emphasized that others factors are also relevant, including how easy it would be to confirm or eliminate a hypothesis by further testing, the strategic “value” of the hypothesis and what we might learn from it, even if the hypothesis turns out to be false (e.g. Peirce 1932-58, §7.220) and how much is at stake if we do not follow-up on a suspicion. Even false hypotheses may lead to a diagnosis, e.g. the hypertension imaging may not show renal artery stenosis, but the adjacent adrenal gland may be enlarged suggesting Pheochromocytoma!
In a clinical context, the value of a hypothesis consists in its consequences for further treatment. Aspects of this Peircean approach are reflected in the received wisdom of the literature on clinical decision-making. For instance, Richardson & Wilson (2015) highlight two relevant considerations (or strategies) in addition to the probability of hypotheses: first, that clinicians should sometimes prioritize hypotheses that, if true, would support an effective treatment and, second, that they should prioritize hypotheses which would be very harmful if left untreated. Similarly, in the analysis underlying the so-called ‘threshold approach’, Pauker & Kassirer (1980) consider four possible consequences of testing a diagnosis: (i) treatment is applied and the diagnosis is true; (ii) treatment is applied, but the diagnosis is false; (iii) treatment is withheld, but the diagnosis is true; and (iv) treatment is withheld and the diagnosis is false. Consequently, whether to apply the test depends both on the accuracy of the test (false positive, false negatives etc.), on the potential harms of the treatment and the test and on how serious the disease would be if left untreated.
In addition to these sensible strategies, one ought to remember that more strategic considerations can also play an important role. By strategic, we mean considerations, which may not in themselves take us closer to the truth, but which will help our further investigations in other ways (Hintikka 1998, Paavola 2004) as in the clinical example noted above. For instance, it can sometimes be valuable to rule out certain plausible hypotheses, e.g. because this would allow us to ignore otherwise possible complications related to further treatments or tests, or simply because it makes our diagnostic space more manageable. We are not considering wild hypotheses because we have a patient and circumstances before us in the clinic. Thus, if a hypothesis can be easily and reliably tested without harm to the patient, this can be a good reason to prioritize it, even if it seems unlikely or a priori improbable. Examples of this include serologic testing for Lyme disease, Chagas cardiomyopathy, serum protein electrophoresis, fat aspirate for amyloidosis, and ferritin levels \soutand cardiac imaging for myocardial iron.
While we emphasized above that in the opening stages of inquiry, considerations besides likeliness of being true play an important role, i.e. conjectures, medical diagnosis usually demands that we find, as economically and quickly as possible, the explanation for the patient’s condition. This is the third type of reasoning, where physicians conclude that a hypothesis is either ruled out, and can therefore be ignored in further inquiry, or that it is likely enough to merit further treatment or exploration. Experienced clinicians should have a diversified kit of strategies rather than relying on any one approach every time.11See also Campos & Stanley (2015) on the diversity of logical strategies employed in clinical reasoning. Thus, while clinicians may sometimes use something akin to Bayesian reasoning or frequentist statistical reasoning, they sometimes have to rely on less formalizable clinical judgments, based on their experience. We illustrate this in the clinical case below.
  1. Clinical Case: Chest Pain
Having introduced our general account of the logic of clinical diagnosis, we proceed to examine a detailed case. In the example we distinguish between the description of the clinical details (in italics) and our logical commentary.
Scene : At home: 09:00. 54 year-old man walking down the stairs to breakfast. He suffers sudden substernal chest pain radiating to his left shoulder and back. No previous history of chest pain. No previous chronic disease. His spouse immediately calls the local emergency number.