Upon arrival, a review of the inflight recordings and TTE change the diagnosis to dissecting thoracic aortic aneurysm. Patient is taken directly to surgery where ascending (type A) aortic dissection was repaired. He was discharged home after one week.
  1. Conclusions
Extant discussion of diagnosis selection tends to represent the reasoning probabilistically (e.g. Gøtzsche 2008, Ch. 4; Richardson & Wilson 2015): the physician is supposed to estimate the prior probability of the hypothesis and update this in light of information from incoming tests. This model may fit some aspects of our case study. For instance, when the physician was considering the implications of the negative tests, she may be taken to rely (implicitly) on something like a Bayesian estimation of the probability given test results. Given appropriate assumptions about prior probability, sensitivity, and specificity of the test, a positive test would have given a high posterior probability of the disease being there, with negative tests only moderately reducing it and thus failing to conclusively rule out the diagnosis. (Insert here the Bayesian graph parts or not?)
However, often the physician will not be able to give anything like a precise numerical estimate of their confidence on the hypothesis. Instead she will have to rely on less precise judgments about whether the clinical picture ‘fits’ the candidate hypothesis, judgments which have to rely on clinical acumen and experience. In these cases she will at best be able to state rough qualitative estimates, e.g. that a given diagnosis is “the most likely” or “unlikely, but still possible”. In these cases, it might be better to represent the physician’s reasoning analogous to a fitting together of clues in a crossword puzzle, rather than as vaguely approximating the mathematical machinery of the probability calculus (see Haack 2003).
Even when probabilistic considerations can be applied, they are not sufficient for diagnosis selection. As highlighted throughout, the physician must also bear in mind the seriousness of the potential diagnoses and test possibility of testing. Furthermore, if probabilistic calculations are isolated from careful observation of the particular facts of the clinical case and if assigned prior probabilities disregard clinical experience and insight, exclusive consideration of probabilities may lead to misdiagnosis. In this case, in order not to miss a very plausible diagnosis and to prevent further muscle damage, the cardiologist persists in the diagnosis of myocardial injury; therefore, she recommends thrombolytic therapy, even in the face of an atypical clinical picture.
Judicious attention to relevant facts that complicate this picture is also relevant to perspicuous diagnosis. In the review, the cardiologist returns to the bedside, listens for the holosystolic murmur reported at initial examination, notes the non-stress induced pain; the slightly widened aortic valve shadow on the CT, and the revises the diagnosis: it must be a dissection . Abduction introduces an hypothesis that provides more plausible explanation for the positive and negative findings. The consideration of all relevant facts eliminates several possible diagnoses from the diagnostic space—for example, the negative CT scan ruled out pulmonary embolism. Several possible diagnoses are conjectured. Consequences are deduced from them. For some of the diagnoses, the predicted consequences are not observed, and the diagnoses are discarded. Meanwhile, other diagnoses are consistent with the observed facts, even though the salient clue (soft holosystolic murmur) is soft, it and other clues, fit into an explanatory pattern. All of the observations, history, testing, and physical examination must be fitted together as clues in a crossword puzzle.
In general, the foregoing case illustrates: (1) the generation of plausible diagnoses by abduction, and (2) their selection on the basis of a combination of strategies. These strategies include the evaluation of diagnoses through Bayesian statistical criteria, where the experience and acumen of the clinicians provide informed prior probabilities. However, they also involve careful observation and the comparison of predicted signs and symptoms given a conjectured disease to the observed bedside facts. The review of clinical findings and testing is crucial to resolving atypical or unusual presentations. Experienced clinicians do not rely exclusively on prior and posterior probabilities. They also hold onto clinical history, the laboratory findings while considering other conditions in order to refine their impressions and form reasonable patterns that result in correct diagnoses.