Abductive Strategies:
How does the physician use abduction?
1.Gestalt: analogous or empirical matching; the most frequent approach in dermatology, radiology, microscopic anatomy; may be disastrous for there are no immediate facts to account for abduction; anecdotal or a pearl at best. Heavily weighted by prevalence hence population statistics. Difficult to teach and learn: distinguishing features may be subtle at first encounter.
2. Proof by exclusion; an exhaustive list as might be provided by an artificial intelligence program as noted above. Demonstrate decision tree: ‘foliate and prune’.
3.Muliple branching: employs the method of elenchus: patient history is the first question: symptom, where, how long, occurred before, family history of symptom. If no differential diagnosis then the patient must be watched. For each question the answer may be considered ampliative.
4.Steepest ascent: shooting from the hip; e.g. one supposes hypothesis is one that gives maximum probability but employing prior probability, derived from experience. One scans the details of the interview, testing, and seeks for a local maxima by a grid of trial points: review history, review previous medical history, and consider that there may be several maximal points and no guaranteed that the global maximum is attained; similar to mountaineers selecting the ascent of a peak: either straight upwards with severe gradients, or switching trails for moderation. Eample:A traveler from Ethiopia arrives by airplane in New York and begins to complain of severe pain in his knees. The physician guesses sickle cell anemia, hypoxemia and bone infarcts.
5. Bayesian: consider the joint probability of the evidence and the conclusion or incidence and evidence but also consider the assumptions
a. New conclusions arise from populations for which full incidence and prevalence data are available for the diagnosis in consideration and it evidence i.e. symptoms and tests.
b. Population is static
c. A definite diagnosis is made in every case in order to support the
chain of reasoning.
6. Hypothetico-deductive: the abduction is based on all of the usual investigatory maneuvers, plus curiosity or ambition to verify one’s initial hypothesis. Perhaps nothing comes out of this at first, or even from sustained exploration of the hypothesis, but further pursuit turns up an anomaly that one had previously overlooked; point is that all efforts gathering data, facts, experiments is a complicated and partly subconscious process. Including chance and does not, like chess or Go, follow the rigorous syllogistic logic of the Greeks. It is a strategy to be emulated.
First, she considers the usual cause of chest pain in this age group and population by way of prevalence—that is, the number of patients per 100,000 persons per year in the population who have the disease. Her initial abductive inference may be reconstructed as follows:
Premise 1: 54 yr. old man has severe chest pain and related symptoms.
Premise 2: If he were suffering from acute coronary syndrome (ACS) the severe chest pain and most of the related symptoms would be explained.
Conclusion: It is plausible to believe, fallibly and pending further testing, that he is suffering acute coronary syndrome.
This abductive inference draws on her knowledge and experience of the most prevalent cause of severe chest pain in the population including this patient. But the diagnosis of cardiac ischemia is only a tentative first hypothesis.
Second, she starts to search for alternative explanatory diagnoses. We can think of this abductive process as an internal dialogue of questions and possible answers where the physician rapidly constructs a diagnostic plan for herself to follow, e.g. “is it possible that the patient suffers from diagnosis X? If so, would X best explain the observed phenomena? And do I have an efficient way of testing for X or treating it?” Through this process she prepares a short list of actionable differential diagnoses. The clinical scene, as it manifests itself in subjective (i.e. patient reported and medical record) and objective criteria, pain distribution and scale, clinical and family history to further refine the search for the commonest plausible and even implausible causes for chest pain, the less common, and even the rare causes. And the rule-out scenario must be efficient and rapid.
Third, while she keeps in mind the adage “horse’s hoofs first, zebra’s hoofs later”, the initial priority is to determine which possible diagnoses are potentially life threatening and to rule out those conditions in the most rapid (or efficient) fashion. Since acute coronary syndrome, if left untreated, would cause severe damage to the patient, this is another reason to prioritize the diagnosis. Some artificial intelligence programs may be helpful to remind us of rare but dangerous conditions as above see below.
One naive approach would be for the physician to consider all the plausible disease candidates to rule out. Obviously this list would encompass too many diagnostic possibilities, especially when the patient is unstable. For instance, computer programs to assist medical diagnosis (e.g. Isabel, DXplain) are based on the prevalence and cardinal weighting of signs and symptoms based on sensitivity and specificity linked to diseases. Red-flagged diagnoses, as above, should be considered higher in the list of possibilities, as they would threaten life. The cardinal arrangement of the diagnoses is based on the experience of the writers of the program and the epidemiology of diseases commonly encountered. They are based on usual signs and symptoms in the western world. The program links the signs and symptoms to diseases that manifest in this way. These are aids to abduction in the sense that the clinical picture would include most or all of the findings.
Most of the likely diagnoses (1 &2) are already in the physician’s diagnostic space as they are commonly related to the clinical presentation. Most experienced physicians would not find such a long list of possible diagnoses helpful in an emergent situation. In the software-generated list, the physician is faced with local prevalence data and clinical experience, and that is specific to the geographic area and population at risk. She may not know how the computer programmers assign weights to each of the signs and symptoms, test characteristics and to the prevalence of disease. Furthermore, the list is too extensive and cannot specifically consider clinical context (emergency) and the physician’s observations, experience and insight. Experienced physicians would be more attuned to the clinical setting (e.g. medical history, country of origin, foreign travel, use of drugs, smoking) the patient’s response to therapy (e.g. pain relief and normalization of heart rate, breathing), and would be more likely to use blood levels of enzymes, radiography, and ECG in order to rapidly shorten the list of differential diagnoses.
Computer programs are based on common epidemiologic findings in specific diseases, not on interpreting whether or not the findings are properly perceived and integrated into the diagnostic picture. Similar reasoning is applied in view of randomized controlled trials i.e. results apply only to the population studied, but
Travel to sub-Saharan Africa might associated anemia with malaria (prevalence); sedentary life style would place coronary ischemic heart disease higher in the list of diagnoses; family history of diabetes in mother and grandmother would integrate weight gain, hypertension, and dysuria into the spectrum of metabolic syndrome. Thus the acute signs and symptoms (the cues or clues) must be cardinally ranked by clinical context, and experience of the clinician. Perspicacious observation at this point is an intricate part of the reasoning process itself. As Peirce might put it, “abductive inference shades into perceptual judgment without any sharp line of demarcation between them” (1998, 227). Knowledgeable, judicious observation is crucial to further hypothesis generation or revision and selection. It is the wide arm of common sense.