Chapter 18 A history of clinical reasoning research
Clinical reasoning has been a topic of research for several decades. The history of this research is important as it provides insights into the various ways in which both cognition and clinical reasoning have been conceptualized over the years and provides a context for current understanding of clinical reasoning and the ways in which it is taught to novice health professionals. In this chapter we draw on two recent research studies which have investigated clinical reasoning as used by different health professionals. These studies (Loftus 2006, Smith 2006) were situated within an understanding of clinical reasoning derived from the variety of research approaches that have been used to study clinical reasoning.
Early research into clinical reasoning was based predominantly within the empirico-analytical paradigm. The first studies came from behavioural psychology, and were followed by studies based on cognitive psychology. A separate but related body of research, generally referred to as medical decision theory, adopted a more probabilistic and statistical approach to conceptualizing clinical reasoning. Research into novice/expert differences has also constituted a distinct topic throughout the history of research into clinical reasoning. In more recent years, research situated in the interpretive and critical paradigms has appeared and grown in volume, especially in healthcare professions other than medicine.
The oldest research tradition in clinical reasoning is behaviourism. Behaviourism is the view that mental phenomena like clinical reasoning can be understood only by analysing behaviour. Behaviour such as clinical reasoning is taken to be a dependent variable, and the independent variables that produce it are the stimuli that might lawfully cause that behaviour. The behavioural laws that link stimuli to behaviour are assumed to be similar in kind to the laws of physics and chemistry. Internal states of consciousness are excluded from this view of psychology as being beyond scientific study. Some research into clinical reasoning has been conducted within the behaviourist paradigm. For example, Rimoldi (1988) tested diagnostic skills of medical practitioners and students in the 1950s and 1960s, showing that as expertise increased so the numbers of questions asked and the time taken to solve diagnostic problems decreased.
Behaviourism has affected the teaching of clinical reasoning and other skills. For example, the notion that students should receive immediate corrective feedback on their performance comes from behaviourism, as does the precept of providing explicit aims and objectives that are related to measurable outcomes (Custers & Boshuizen 2002, Greeno et al 1996, Smith & Irby 1997).
Many features of modern medical curricula that have a direct bearing on the way that clinical reasoning is taught and practised can be traced to influences from behaviourist principles. These features include frequent and progressive testing, and close monitoring of students (Custers & Boshiuzen 2002). Behaviourism may have many weaknesses but it has been of some benefit when intelligently applied. However, as an explanation of all learning, behaviourism is conceptually weak and does not go far enough. It ignores context, sociocultural interaction and intersubjectivity. In the endeavour to address some of these weaknesses cognitivism emerged as a more powerful conceptual model for thinking about mental phenomena such as clinical reasoning (Patel & Arocha 2000).
Cognitive science seeks to account for intelligent activity as exhibited by living organisms or machines. Cognitivism replaced the behaviourist metaphor of cognition, as a black box having environmental inputs and behavioural outputs, with the metaphor of cognition as a form of computation and information processing, similar in kind to that carried out by computers.
Cognitivism allows for ‘mental’ structures and processes, whereas behaviourism does not. Information processing, memory representation and problem solving are three core concepts (Case & Bereiter 1984). There have been a number of attempts to characterize knowledge structures according to a cognitive view, and these feature prominently in much clinical reasoning research within the cognitive paradigm. The mental structures which purportedly play such a prominent role have included successively: categories, prototypes, instances, schemas, scripts and networks (Gruppen & Frohna 2002). Each concept was introduced in turn as a response to the perceived weaknesses of its predecessors. For example, according to the theory of instances, knowledge organization occurs around an individual instance rather than as an abstract based on several cases. This idea was proposed in response to the weaknesses perceived in the construct of prototypes (Brooks et al 1991, Homa et al 1981).
This preoccupation with mental structures and access to them is typical of cognitivism and is symptomatic of the underlying conceptual model of cognition as a form of computation. Along with the concern for cognitive structures is an interest in the cognitive processes by which individuals make use of such structures. The most popular process for utilizing these cognitive structures in clinical reasoning is held to be hypothetico-deductive reasoning.
Research investigating the hypothetico-deductive method as a foundation in clinical reasoning was divided by Bradley (1993) into two groups. Researchers in the first group used think-aloud protocols with patients or simulated patients (e.g. Elstein et al 1978). Those in the second group used case vignettes (e.g. Eddy & Clanton 1982). There were weaknesses with both kinds of study, such as the artificial nature of the think-aloud protocols that tended to be used. However, the concept of hypothetico-deductive reasoning is generally considered to be a robust element of the cognitive paradigm, and one that could be adopted in different models that may reject many of the assumptions of cognitivism. The cognitivist body of research also highlighted the differences between experts and novices in clinical reasoning.
Much of the effort in cognitivist research into clinical reasoning has consisted of attempts to delineate differences between novices and experts, which is therefore sometimes called the contrastive method. Most of this research has been experimental. A problem-solving approach is generally used, in which cognitive processes are studied in tasks that attempt to represent medical thinking. Typically, protocol analysis has been used, as in the work of Ericsson & Simon (1993), who claimed that experts’ use of forward-directed reasoning was ‘one of the most robust findings’ (p. 132) of research in this field. Forward reasoning is supposed to occur when someone gathers data and, with the aid of a great deal of pattern recognition invoking ‘if-then’ production rules, eventually reaches a conclusion (Patel & Groen 1986). Backward reasoning is supposed to occur when someone selects a hypothesis early and then proceeds to test it by gathering data that will confirm or refute it. This is believed to work well if the hypothesis is correct, but means that the problem-solver may need to start again if it becomes clear that the data being gathered tend to refute the hypothesis. This view of expert–novice difference is widespread in the clinical reasoning literature. It began about 1980 when researchers claimed that these differences existed between experts and novices in physics (Larkin et al 1980). These studies influenced the research of Elstein et al (1978) into clinical reasoning, seeking the same phenomenon of forward and backward reasoning.
The finding that forward and backward reasoning distinguish experts and novices has now been extensively investigated and ‘confirmed’ within medicine (Patel et al 1990), and is now widely accepted. However, the relevant studies were experimental and can be criticized as being highly artificial. In general they used written protocols, with all the relevant information presented simultaneously on a single page. The researchers asked individuals to read the case and verbalize or write down their thoughts. Analysis of these verbalizations produced the apparent distinction between forward and backward reasoning.
Variations on the research into novice–expert differences in reasoning have continued to recent times. For example, Norman & Schmidt (2000) also devised experiments to test forward and backward reasoning strategies among novices and experts. Their findings showed clearly that novices did better when using backward reasoning. This kind of finding has been used to provide a theory of what happens during problem-based learning, and this is why the hypothetico-deductive model is still an important theory in the teaching of clinical reasoning (Barrows & Feltovich 1987).
However, as Norman et al (1999) have observed, the concept of forward and backward reasoning is problematic owing to the artificial nature of the decontextualized settings in which it was established. In other words, these findings may be a laboratory artefact. Lemieux & Bordage (1992) discussed the issue of research into forward versus backward clinical reasoning at length. They concluded that laboratory-based studies were far too limiting, and that the results were often more a reflection on the method of investigation than the actual reasoning of the clinician. This criticism is supported by the work of Laufer & Glick (1996), who investigated novice–expert differences in real-world work settings, using an ethnographic approach informed by ideas from the cultural psychology of Vygotsky (1978, 1986).
Cognitivism entails an essentially individualistic view of expert–novice differences. Even as early as 1980, some researchers were dissatisfied with cognitivism as an explanatory model. For example, Norman (1980) complained that cognitivism was inadequate for conceptualizing the influence of interaction with others and the ways in which an individual’s personal life history and cultural background could affect reasoning skills. If cognition is in fact not a computational process then the search for the purported cognitive structures and processes may be misguided and doomed to failure. It can be argued that the similarities between cognition and computation are trivial, such as the ability to do simple mental arithmetic in one’s head. Much of the research referred to above, which sets out to establish the nature of the cognitive structures in clinical reasoning and other forms of cognition, assumes what it sets out to prove. The underlying metaphor of cognition being a form of computation is open to challenge. Humans undertake procedures such as mathematical calculations differently from computers, and the way they do them varies depending on the circumstances (Dreyfus 1992). Cognitivism has an essentially individualistic view that expertise in skills such as clinical reasoning is a collection of behaviours and thoughts which are unique personal constructions. This directly contrasts with the sociocultural view that expertise is fundamentally best viewed as a social phenomenon. From this perspective expertise would, in large part, be selective assimilations of prevalent social practices and values. There is limited research into clinical reasoning from a sociocultural perspective. Engeström (1995) used a sociocultural approach to study medical expertise in clinical consultations with real patients, and was able to richly describe and articulate his findings in a manner that would have been precluded by a purely cognitivist framework.