CHAPTER 7 Clinical reasoning
the nuts and bolts of clinical education
Introduction
Clinical reasoning has been the subject of research for several decades. How it is both conceptualised and researched has reflected the dominant paradigms of the times in which the research was done. The earliest research was conducted within the paradigm of behaviourism, followed soon after by its successor, cognitive science (Elstein et al 1990). In behaviourism and cognitivism there is an analytical focus on the changes occurring within the health professionals who are learning and doing the clinical reasoning. The focus is on the clinician as individual decision maker. The teaching of clinical reasoning within this paradigm naturally follows the imperative to bring about the required cognitive changes in newcomers to the health professions, who will then be able to behave appropriately (e.g. Custers & Boshuizen 2002).
In more recent years, different forms of research have emerged which stem from different sets of assumptions. These newer forms of research are based on more humanistic thinking from the humanities and social sciences. They have included research based on narrative thinking (Charon 2006, Mattingly 1994), professional artistry (Fish 1998), critical theory (Trede & Higgs 2003) and the use of language (Ajjawi 2006, Loftus 2006). These humanistic approaches can be included within the paradigm of interpretivism. Here, the analytical focus is much more on the social world within which clinical reasoning occurs, and the role of the clinician within this social world. In this chapter we synthesise current thinking about clinical reasoning within these different paradigms and present practical implications for developing clinical reasoning during clinical education. In order to do this, in the first half of the chapter the authors draw on models of reasoning founded in behaviourism and cognitivism and highlight the key features of ‘the stage theory of expertise development’ and acquisition in the health professions. The second half of the chapter focuses upon the interpretivist paradigm in clinical reasoning, and the strength of using narrative and sociological theory as a lens to view and better understand clinical reasoning in the healthcare setting.
Clinical reasoning as a cognitive phenomenon
There are many anecdotes about remarkable diagnosticians; physicians who do not heavily rely on inquiry because they already seem to ‘know’. Can you imagine a scenario where the doctor is an elderly man who has been in family practice for about thirty years. He usually does not enquire very deeply into the nature of our symptoms; he appears absentminded when we present our complaints. He does not make an extensive use of diagnostic tools like laboratory tests or X-rays. And yet, he hardly ever misses a diagnosis. When we visit him with a penetrating pain in the chest, he does not refer us to the cardiologist but, after some questions, sends us home with the advice to take some rest because ‘stress can do these things you know, but no doubt it will disappear in a few days’. And his advice turns out to be correct. When he sees our young daughter late in the evening having convulsions and a high fever, he decides to send her to the hospital but reassures us by saying that it does not seem to be something related to brain dysfunction, but rather the result of some infectious process, probably of urogenital origin. And his diagnosis proves to be accurate.
In this section of the chapter, we will discuss some of the reasons why experienced physicians display such remarkable diagnostic performance and how these skills have come to develop over the years in practice. The initial attempts to address these questions in the early 1970s, looked for expert doctors’ superior reasoning processes. The well-known studies conducted by Elstein et al (1978) exemplify these ‘processing theories’. They proposed the notion of the ‘hypothetico-deductive method’: early in the clinical encounter expert doctors generate diagnostic hypotheses and subsequently gather information to confirm or refute these hypotheses. Although this may describe the essential elements of the clinical reasoning process, it does not account for expert performance for a simple reason: subjects at all levels of expertise were shown to reason through similar processes.
As the idea of a general problem-solving process failed, research shifted towards ‘structure theories’, which focus on underlying knowledge structures that generate diagnostic hypotheses (Norman 2005, Ericsson 2007). Empirical research within this paradigm has concentrated on how expertise develops in medicine, by exploring how knowledge is acquired, organised in memory and used by experienced doctors for diagnosing clinical problems (Schmidt & Rikers 2007). These researchers generated a theory that considers the development of expertise as progressing through a number of transitory stages, each characterised by qualitatively different knowledge structures underlying diagnostic performance (Schmidt & Boshuizen 1993a, Schmidt et al 1990). In the next section, we sketch this theory and subsequently summarise findings of more recent studies that have clarified the process of clinicians’ diagnostic reasoning. In the final section, we briefly discuss their implications for clinical education.
A stage theory of expertise development
STAGE 1
The development of these causal networks can be illustrated by findings from a study by Schmidt et al (1988) in which students at different levels of training—first year health sciences students and second and fourth year medical students—were shown the case description in Box 7.1.
The students were asked to explain the signs and symptoms presented in the case (which is one of acute bacterial endocarditis due to intravenous drug use) in terms of the underlying pathophysiological processes. Analysis of the protocols generated by the three groups of students showed an almost linear increase in the number of propositions used to interpret the case.1 This demonstrates the rapid development of elaborated causal networks explaining the causes and consequences of disease in terms of underlying pathophysiological processes.
Students at this stage of their education try to make sense of clinical cases presented to them by analysing isolated signs and symptoms and relating each of them with the pathophysiological mechanisms they have learned. As students do not yet recognise patterns of symptoms that fit together, processing of case information is effortful and detailed. This explains the ‘intermediate effect’ consistently found in the studies of clinical case recall, where intermediate-level students remember more details of cases than medical experts (Rikers et al 2000, Schmidt & Boshuizen 1993b). When asked to think aloud while solving cases, intermediate-level students were also shown to use detailed knowledge of the basic sciences in explaining for themselves the signs and symptoms of the patient. In contrast, references to basic science concepts were almost absent in think-aloud protocols of experienced doctors (Boshuizen & Schmidt 1992). This was thought to occur because the experienced doctors operate upon different knowledge structures. This difference leads to the second stage in our theory of expertise development.
STAGE 2
Through extensive, repeated application of acquired knowledge and, particularly, exposure to patient problems, these elaborate networks of concepts and their interrelations become compiled into high-level, simplified causal models explaining signs and symptoms, and are subsumed under diagnostic labels (Boshuizen & Schmidt 1992).
The transition from the first to the second stage in students’ knowledge structures (referred to as knowledge encapsulation) can be explained using data from the same study by Schmidt et al (1988). Box 7.2 displays the protocols produced by two medical students when they explained the case of endocarditis.
BOX 7.2 Pathophysiology protocols of medical students at different levels of training
The effects of compilation become evident when the two protocols are compared. The fourth year student used many words to explain the mechanisms involved in shock due to sepsis (in some respects, inadequately). The sixth year student did not refer to the word shock at all. The whole case was explained in terms of sepsis and its secondary effects. An internist would probably be even more concise and say: ‘This drug user has developed a sepsis due to the use of contaminated needles’. Used in this way, the concept of sepsis encapsulates the fourth year student’s detailed pathophysiological explanation. It is sufficient to fully explain the condition of the patient. Having a concept such as sepsis available to the reasoner enables them to see patterns of symptoms as wholes, considerably speeding up processing of a case, and adding to accurate diagnosis.
Several studies have confirmed the predictions derived from the notion that biomedical knowledge becomes encapsulated into clinical concepts. Pathophysiological explanations of experts were shown to contain less biomedical and more encapsulated concepts than those of students (Van de Wiel et al 2000), and recall protocols of experts contained more encapsulations than protocols of sub-experts (Rikers et al 2002). Experts have many encapsulating concepts available, describing syndromes or simplified causal mechanisms. This knowledge, often called clinical knowledge (as opposed to biomedical knowledge), tends to be used preferentially by experts (Boshuizen & Schmidt 1992, McLaughlin 2007). Indeed, biomedical knowledge apparently only indirectly relates to clinical competence (De Bruin et al 2005). Recent studies, however, have suggested that biomedical knowledge may play a more important role than is presently assumed (Woods et al 2005, 2006), but this requires further investigation.
STAGE 3
Illness scripts are cognitive structures containing relatively little knowledge about pathophysiological mechanisms, but a wealth of clinically relevant information about the disease (Feltovich & Barrows 1984). A general structure of an illness script consists of enabling conditions, faults and consequences. Enabling conditions are factors that generally make the occurrence of a certain disease or family of diseases more likely. The fault is a description of the malfunction, which may consist of a diagnostic label or a simplified description of a pathophysiological mechanism, for example, invasion of pathogenic organism into body tissue. The consequences generally are the signs and symptoms that arise from the fault (Feltovich & Barrows 1984).
Expertise development is associated with the emergence of illness scripts rich in knowledge of enabling conditions. Studies comparing students and physicians at different levels of expertise have shown that the number and richness of enabling conditions associated with particular diseases increase with expertise (Custers et al 1998), and experienced doctors tend to make extensive use of enabling conditions (Hobus 1994, Van Schaik et al 2005).
When physicians review a patient like the young man described in Box 7.1, they would search for an appropriate illness script in memory and when one (or a few) are selected, verify the script by matching its elements to the information provided by the patient. In this course of script verification, the script is said to become instantiated (Schmidt & Rikers 2007, Schmidt & Boshuizen 1993b). These instantiated scripts do not necessarily become decontextualised but remain available in memory as episodic traces of previous patients. Illness scripts exist, therefore, at various levels of generality, ranging from representations of disease categories to prototypes, to representations of individual patients previously seen. Storing these different representations constitutes another transition, a fourth stage in the course of expertise development.
STAGE 4
Throughout years of clinical practice, doctors store in memory more and more instances of individual patients. Expert clinicians’ reasoning is largely based on recognition of similarities between the case at hand and these examples of previous patients (Schmidt & Boshuizen 1993b, Schmidt et al 1990). This so-called pattern-recognition reasoning occurs in routine situations, rapidly and effortlessly, as a largely unconscious process without requiring physicians to analyse individual signs and symptoms or explain their causal mechanisms (Norman 2005, Norman & Brooks 1997, Ericsson 2004). Nevertheless, the knowledge structures acquired in the earlier stages of expertise development, such as pathophysiological knowledge, do not decay but remain available in memory and may be activated when pattern-recognition reasoning fails (Schmidt & Boshuizen 1993a, Schmidt et al 1990, Patel & Groen 1986).
The role of experience and examples of prior patients in diagnostic reasoning
The influence of prior examples on the generation of diagnostic hypotheses was first demonstrated by empirical studies in the domain of dermatology conducted by Brooks, Norman and colleagues in the 1980s (Brooks et al 1991). Medical students were asked to diagnose dermatological conditions, and similarity with a previously seen example of the particular condition was shown to dramatically influence diagnostic accuracy. Subsequent studies in other domains, such as electrocardiography and psychiatry, reaffirmed non-analytical reasoning based on similarity to prior examples as a crucial component of diagnostic reasoning (Hatala et al 1999, Norman et al 2007). Moreover studies have shown that, far from being objective, interpretation of signs and symptoms in a case tends to be influenced by the diagnosis under consideration. When medical students were presented with photographs of classical signs of diseases, features were shown to be easily misinterpreted due to the influence of an initial hypothesis (LeBlanc et al 2002).
These findings raise the question of how reasoning strategies affect the quality of diagnoses. Studies on diagnostic errors have pointed to the negative consequences of relying excessively on non-analytical reasoning (Croskerry 2003, Graber et al 2002). Diagnostic errors derived from premature closure, which seem to increase with ageing, exemplify these possible deleterious effects (Eva 2002). Apparently non-analytical reasoning is highly effective in routine situations but may provoke failures when doctors encounter complex or unusual problems (Ericsson 2004, Croskerry 2005). Concerns with avoidable medical errors have contributed to attention being directed to the other pole of the spectrum of diagnostic reasoning. It is known that expert doctors may shift from the usual automatic way of reasoning to an analytical, effortful diagnostic approach in some situations (Patel & Groen 1986, Rikers et al 2002). This has been reported when doctors diagnose cases outside of their own domain of expertise; in such cases they adopted an elaborate biomedical processing approach for understanding signs and symptoms (Rikers et al 2002).
Some of our recent empirical studies have confirmed that doctors may engage in effortful reflection for diagnosing cases (Mamede & Schmidt 2004, Mamede et al 2007). Findings of these studies shed light on the analytical mode of diagnostic reasoning and will be briefly discussed here.