The availability of narratives i.e. experience-based, organized stories about the world, is an important tool allowing us to navigate our everyday tasks. During clinical work in particular, selecting one among many narratives is a very common task and one fraught with risk for physicians and patients alike. Consider the following, extremely common, scenario in my practice: a patient walks in a physician’s office with an elevated reading of his or hers Blood Pressure (BP) asking for an evaluation. The physician’s task at this point is two-fold:
- first, to confirm the presence of elevated BP (giving the patient the diagnosis of a hypertensive condition)
- second, to decide whether the patient has one of the many known disease states causing high BP (this is known as Secondary Hypertension) or finding no known cause, give the patient the diagnosis of Essential/Idiopathic (“don’t know what is going on”) Hypertension
After one has determined a hypertensive condition to be present, the physician will turn to the second problem of diagnosing which form of hypertension the patient has. To do so, the physician will mentally turn to a list of disease states handed down to him or her in the form of descriptions in medical texts and journal articles. These descriptions are nothing more than extremely long statements of a rather rudimentary form: “if a patient has the condition X, then the patient will manifest such and such symptoms and accompanying abnormalities in laboratory tests”. Hence medical textbook chapters are nothing more than narratives of what one expects to see if one is confronted with a patient who is known to have condition X. However the task of medical diagnosis in the clinic is rather different: to determine whether the particular patient has condition X, instead of Y,Z etc knowing that he has elevated BP along with any number of manifestations!
As an inferential task, medical diagnosis seems to be weighing two different sources of uncertainty:
- the set of possible narratives of disease states that can lead to elevated BP (the chief complaint for which the patient sought medical attention)
- the compatibility between the particular patient’s manifestations and the narrative for each possible disease state that is potentially associated with the chief complaint of elevated BP
Note that weighting of uncertainty from both sources is always conditional on a premise assumed to be true and indisputable, i.e. the presence of elevated BP to select the set of narratives for further consideration (Source 1) and the presence of a specific disease state (Source 2). A further historical observation is that our current approach of documenting patient encounters,dating all the way back to Pierre-Charles Alexandre Louis in the 18th century is also a conditional one. It starts from the chief complaint, followed by enumeration of complaints and observable manifestations separating the data from the list of hypotheses (aka differential diagnosis) from which one diagnosis is to be selected.
The role of familiarity with the narratives consisting our medical knowledge in the process of evaluating particular patients as complimentary to the data we obtain from each individual patient has been noted by Sir William Osler:
- “To study the phenomena of disease without books is to sail an uncharted sea, while to study books without patients is not to go to sea at all”
- “Medicine should begin with the patient, continue with the patient, and end with the patient“
So how can we conceptualize this conditional evaluation of evidence in terms of numbers ? One should ask for each condition e.g. out of all possible diagnoses, , the degree of compatibility between the patient’s symptoms and the narrative of particular condition: . These in turn should be weighted by the a priori belief that each condition is the cause of the elevated BP (the chief complaint) in the population of hypertensive patients for which nothing else is known: . The final diagnosis is the one with the highest composite numerical score: .
Failure to complete this calculation even semi-quantitatively can lead to embarrassment and possibly harm at different phases in a physician’s career:
- by not accounting for the a priori plausibility, relying instead on simple matching of patient cases to textbook narratives will lead one to over-interpret common experiences. Many medical students go through this phase when they start reading medical textbooks and before they have seen any actual patients: they think they have the disease in the last chapter of the textbook they read. My personal embarrassing version of this scenario occurred halfway through medical school; at that point I diagnosed myself with discoid lupus on the basis of being tired, losing weight and hair whereas exam fatigue/burnout and male baldness accounted for these symptoms.
- by not being familiar with the most recent updates or discrediting “book-smartness” clinicians will fail to consider alternative diagnoses. This is particularly dangerous form of deviation from the rules, because it happens at a time in which clinicians have an active medical license that allows them to evaluate and treat actual patients. At the very least, patients will not get the optimum out the encounter wasting thus their time and financial resources and decreasing their satisfaction with the medical system. The worst outcomes though occur when physicians blow off, delay to diagnose or misdiagnose patients leading to delayed/ineffectual/harmful treatment of patients.
This account of the task of medical diagnosis as a choice among narratives can in fact be made much more formal and put in sound quantitative basis. Such a formalization, will also allow us to make some connections with the evaluation of medical evidence (evidence based medicine), in the literature deserves its own post though 🙂