( Case 37, Case 38– Case 39, Case 56)
Before a diagnosis is accepted as a basis for action, namely prognostication or therapy (i.e., a working diagnosis), it must be subjected to a detailed validity assessment. The processes used in this assessment are described here as diagnostic verification. This process makes a final check of a patient’s findings against known patterns of the disease or suspected diseases. When a given patient’s manifestations are identical to those of a known clinical entity, further search for a diagnosis usually is not required, and action can be taken.
When some clinical feature differs from the known pattern of a clinical entity, however, a judgment must be made: Is the feature merely a variation of that entity, or is the finding sufficiently unusual that it raises doubts about the current diagnostic hypothesis? This dilemma is particularly likely when the tentative diagnostic hypothesis is a disorder that the physician encounters infrequently. In such instances, detailed review of disease manifestations in source materials may help to determine whether a particular clinical finding is consistent with the current hypothesis.
Criteria of Validity
( Case 12 , Case 17, Case 38, Case 39– Case 40, Case 43)
Given the nature of evidence, one must always maintain some element of doubt even with biochemical, histologic, or radiologic confirmation of a diagnosis. Several criteria are useful in helping to decide whether a diagnostic hypothesis qualifies as a working diagnosis. One “test” assesses diagnostic adequacy.[ 19] A diagnostic hypothesis is considered adequate when it encompasses all surviving hypotheses and when it accounts for all the patient’s findings, whether abnormal or normal. Another “test” or criterion is coherency. A diagnosis is coherent when a patient’s findings are consistent with the altered pathophysiology of the hypothesized disease state.[ 19] As noted in the preceding chapter, it is in this phase of the diagnostic process that detailed causal models are most helpful. Parsimony is another criterion, although no specific guidelines for its assessment exist.
A parsimonious diagnosis represents the simplest possible explanation of all the findings.[ 19] Generally one tries to find a single diagnostic hypothesis that explains all findings, but in some instances several diagnoses must be invoked to explain all of a patient’s findings and to achieve confidence that clinical features are physiologically and causally consistent with the hypothesized entity. Falsification is also used during hypothesis verification. Falsification, an approach commonly used to discredit scientific hypotheses, is the process by which certain features that are unequivocally inconsistent with the tentative hypothesis discredit this hypothesis and thus call for new hypotheses or revision of competing hypotheses. This process consists in a systematic consideration of alternative, feasible hypotheses as a last step before accepting a given disease entity as the working diagnosis.[ 19]
The credibility of a diagnosis is also a function of its likelihood. After all information is collected, a diagnosis with the highest probability is the most credible. Many schemes have been devised to determine when the probability of a diagnosis is sufficiently high to consider the diagnosis verified, but all are arbitrary. Neither a specific probability value (e.g., 0.95) nor a scoring mechanism based on approaches that give credit for findings that are present and subtract credit for findings that are not present have a solid scientific basis as a verification tool. Beyond verification, these schemes do not account for the benefit and risk of appropriate treatment for that diagnosis or the risk of harms for unnecessary treatment.
( Case 5, Case 8, Case 12 , Case 33, Case 39, Case 53, Case 56)
Accepting a diagnosis before it is fully verified is often known as “premature closure.”[ 52 ],[ 53] Premature closure cannot be avoided simply by avoiding hypothesis generation. Indeed, physicians almost invariably generate hypotheses early in a diagnostic encounter when only a few cues are available. Nonetheless, inappropriate and premature acceptance of a diagnostic hypothesis may be avoided if one insists that all data be considered before accepting a diagnostic hypothesis as verified.
The “tests” considered in the previous chapter on causal reasoning, namely adequacy and coherency, are mechanisms that help to avoid making a diagnosis that is incomplete or wrong. Sometimes, an acceptable diagnosis may not emerge even after all available data have been collected and considered. In such circumstances further testing, further patient observation, or both may be necessary. Alternatively, sometimes a detailed re-examination of the patient and a reanalysis of the significance of individual clinical features or findings reveal hypotheses not yet considered. Some observers have hypothesized that seasoned clinicians may be at greater risk of premature closure.[ 54]
As data emerge, the process of diagnostic refinement requires probabilistic revision, which can be a source for errors.[ 6] Overconfidence bias refers to physicians having a greater belief in the correctness of their diagnosis than necessarily warranted,[ 55],[ 56] which may lead to premature closure. Conservatism reflects the tendency to insufficiently raise or lower the likelihood of a diagnosis as would occur with formal Bayesian probabilistic revision.
Other errors include acquiring redundant evidence or confirmatory bias (confirmation testing at the expense of elimination strategies), incorrect interpretation (placing greater weight than warranted on confirmatory data and discounting or ignoring negative findings), base-rate neglect (ignoring or not considering appropriately pretest likelihoods, especially those that are very high or very low), and ordering effect (the influence of the sequence of the presentation of information).[ 6],[ 56]
The Penultimate Result: A Working Diagnosis
( Case 9, Case 39, Case 41)
Working diagnoses should be highly likely and parsimonious. They should explain all principal clinical findings and should be coherent, in the sense of causal and physiologic relations. They survive the test that no competing diagnostic hypotheses are plausible. Such hypotheses usually produce valid predictions, both of test results and of the patient’s future clinical course. After all efforts have been made to identify the most plausible diagnostic hypothesis, considerable uncertainty may still remain. At some point, despite remaining uncertainty, prognostic and/or therapeutic decisions must be made. The basis for making such decisions is discussed in the next chapter.