The most important clinical actions are not procedures or prescriptions but the judgments from which all other aspects of clinical medicine flow. In the modern era of large randomized trials, it is easy to overlook the importance of this elusive mental activity and focus instead on the algorithmic practice guidelines constructed to improve care. One reason for this apparent neglect is that much more research has been done on how doctors should make decisions than on how they actually do. Thus, much of what we know about clinical reasoning comes from empirical studies of non-medical problem-solving behavior. The following three examples introduce the subject of clinical reasoning:

1. A 46-year-old man presents to his internist with a chief complaint of hemoptysis. The physician knows that the differential diagnosis of hemoptysis includes over 100 different conditions, including cancer and tuberculosis. The examination begins with some general background questions, and the patient is asked to describe his symptoms and their chronology. By the time the examination is completed, and even before any tests are run, the physician has formulated a working diagnostic hypothesis and planned a series of steps to test it. In an otherwise healthy and non-smoking patient recovering from a viral bronchitis, the doctor’s hypothesis would be that the acute bronchitis is responsible for the small amount of blood-streaked sputum the patient observed. In this case, a chest x-ray and Purified Protein Derivative (PPD) skin test may be sufficient.
2. A second 46-year-old patient with the same chief complaint who has a 100-pack-year smoking history, a productive morning cough, and episodes of blood-streaked sputum may generate the principal diagnostic hypothesis of carcinoma of the lung. Consequently, along with the chest x-ray and PPD skin test, the physician refers this patient for bronchoscopy.
3. A third 46-year-old patient with hemoptysis who is from a developing country is evaluated with an echocardiogram as well, because the physician thinks she hears a soft diastolic rumble at the apex on cardiac auscultation, suggesting rheumatic mitral stenosis.

These three cases illustrate two aspects of clinical diagnostic reasoning:

1. Problem Solving: Problem solving research was initially aimed at describing reasoning by expert physicians, to improve instruction of medical students and house officers. The problem solving approach has focused on diagnosis as hypothesis testing, pattern matching, or categorization. Categorization is usually based on matching the case to a specific instance (“instance based” or “exemplar based” recognition) or to a more abstract prototype. In the former, a new case is categorized by its resemblance to memories of instances previously seen. This model is supported by the fact that clinical diagnosis is strongly affected by context even when the context ought to be irrelevant. Errors that can occur in difficult cases in internal medicine include failure to generate the correct hypothesis, misperception or misreading the evidence —especially visual cues— and misinterpretations of the evidence. Many diagnostic problems are so complex that the correct solution is not contained in the initial set of hypotheses; so restructuring and reformulating should occur as data are obtained. However, the generation and evaluation of diagnostic hypotheses is a skill that not all clinicians possess to an equal degree and as a result, a clinician may quickly become psychologically committed to a particular hypothesis, making it more difficult to restructure the problem. Because of this reason, errors in the problem solving approach can occur, and in the patient with serious acute illness these may lead to tragic consequences.
2. Decision Making: Psychological decision research has been influenced from the start by statistical models of reasoning under uncertainty, and has concentrated on identifying departures from these standards. Decision making approach views diagnosis as opinion revision with imperfect information (the clinical evidence). The standard rule for this task is Bayes’ theorem. Bayes’ theorem tells us how we should reason, but probably only a minority of clinicians use it in daily practice and informal methods of opinion revision still predominate. Bayes’ theorem directs attention to two major classes of errors in clinical reasoning: in the assessment of either pretest probability or the strength of the evidence. The psychological study of diagnostic reasoning has focused on these errors, and on the simplifying rules or heuristics that replace more complex procedures. Consequently, this approach has become widely known as “heuristics and biases”. Some of errors in estimation of probability include:

   Availability: People are apt to overestimate the frequency of vivid or easily recalled events and to underestimate the frequency of events that are either very ordinary or difficult to recall. So errors with the availability can come from several sources of recall bias, e.g., ignoring of rare conditions.

   Representativeness: When assessing a patient, clinicians often weigh the probability that this patient’s clinical features match those of the class of patients with the leading diagnostic hypotheses being considered. In other words, the clinician is searching for the diagnosis for which the patient appears to be a representative example. So it can lead to overestimation of probability either by causing confusion of post­test probability with test sensitivity or by leading to neglect of base rates and implicitly considering all hypotheses equally likely.

   Probability transformations: Decision theory assumes that in psychological processing of probabilities, they are not transformed from the ordinary probability scale. But in practice, small probabilities are overweighted and large probabilities underweighted, contrary to the assumption of standard decision theory. This “compression” of the probability scale explains why the difference between 99% and 100% is psychologically much greater than the difference between, say, 60% and 61%.

   Effect of description detail: In clinical practice, a longer, more detailed case description will be assigned a higher subjective probability of the index disease than a brief abstract of the same case, even if they contain the same information about that disease. Thus, subjective assessments of events, while often necessary in clinical practice, can be affected by factors unrelated to true prevalence.

   Conservatism, anchoring and adjustment: In clinical case discussions, data are presented sequentially, and diagnostic probabilities are not revised as much as is implied by Bayes’ theorem; this phenomenon is called conservatism. One explanation is that diagnostic opinions are revised up or down from an initial anchor, which is either given in the problem or subjectively formed. Final opinions are sensitive to the starting point (the “anchor”), and the shift (“adjustment”) from it is typically insufficient. Both biases will lead to collecting more information than is necessary to reach a desired level of diagnostic certainty.

   Order effects: It is difficult for everyday judgment to keep separate accounts of the probability of a disease and the benefits that accrue from detecting it, as required by standard decision theory. Furthermore, there is a tendency to overestimate the probability of more serious but treatable diseases, because a clinician would hate to miss one. In addition, Bayes’ theorem implies that clinicians given identical information should reach the same diagnostic opinion, regardless of the order in which information is presented. However, final opinions are also affected by the order of presentation of information. Information presented later in a case is given more weight than information presented earlier. Other errors identified in data interpretation include simplifying a diagnostic problem by interpreting findings as consistent with a single hypothesis, forgetting facts inconsistent with a favored hypothesis, overemphasizing positive findings, and discounting negative findings.

In conclusion, we emphasize, firstly, that the prevalence of the above mentioned errors has not been established; secondly, we believe that expert clinical reasoning is very likely to be right in the majority of cases; and, thirdly, despite the expansion of statistically grounded decision supports, expert judgment will still be needed to apply general principles to specific cases.