Teaching Diagnostic Clinical Reasoning

Initial Diagnostic Approaches

Table 1 summarizes the four most common diagnostic approaches used by clinicians at various levels of training. In traditional medical school curricula, beginning students are taught to conduct an exhaustive review of a patient's medical history and physical examination before initiating clinical reasoning.

In our illustrative case, Mrs. B's past medical history included several emergency department visits during her early 20s for similar chest pain that defied diagnosis. Her younger sister was bothered by “anxiety attacks.” An exhaustive review of systems uncovered a feeling of unreality and terror during the episode. These data were in the medical student note, unread by the rest of the team.

Experienced clinicians quickly recognize patterns of symptoms and signs that mirror previously seen pictures of disease. For example, Mrs. B's attending physician knew that coronary artery disease is frequently underdiagnosed in women and thought that her pattern and risks fit reasonably well. No one on the team matched the pattern of her presentation with panic disorder, so it was not initially considered.

Multiple branching algorithms have been proposed for exploring and evaluating common clinical problems such as chest pain in the emergency department. Here, the clinician asks a series of yes/no questions where the answer determines the next step, usually based on the best available clinical evidence for populations of similarly situated patients. In the example above, Mrs. B was placed on “rule out MI” and “rule out pulmonary embolism” algorithms based on her presentation with substernal chest pain. Troponins were drawn, a ventilation-perfusion scan was ordered, and she was tentatively scheduled for an exercise tolerance test. No estimate of pretest probability for either diagnosis was in the chart. Her ventilation-perfusion scan was “low probability.” Her stress test showed minor nonspecific flattening of her T waves at an excellent rate-pressure product.

...