1 INTRODUCTION

All thinking occurs in some sort of context, rendering the relation between context and clinical reasoning a matter of significant interest.1-3 Context, however, has a notoriously vague and contested meaning in medical education. We are so accustomed to our contexts that they literally become invisible to us.4 Various interpretations of context have been put forward. Context has been defined as a fabric ‘weaving together’ the different elements of the clinical situation,2 an interplay of dynamically interacting patterns,4 an emergent product of activity or an element surrounding the individual.5 Others have emphasised particular aspects of context such as constituting the clinical plot of a medical case or the physical settings and its importance in mobilising (or not) commitment and engagement.6, 7

This profound disaccord in the meaning ascribed to context may be frustrating, but should come as no surprise. Indeed, clinical reasoning constitutes an object of study claimed by diverse academic disciplines, such as cognitive psychology, medical anthropology or medical sociology. Each discipline accommodates various competing ‘research traditions’ each embracing a certain worldview, a legitimate way of understanding clinical reasoning, bounded by distinct habits of thinking, values, theories and concepts as well as valid research questions and methodologies.8 These research traditions are grounded, tacitly or explicitly, on fundamental epistemological assumptions concerning the nature of knowledge and knowing. Such ‘guiding assumptions’8 not only authorise what counts as knowledge but also endorse legitimate methodologies for knowledge creation.9

Epistemological claims determine importantly how different research traditions conceptualise clinical reasoning and consequently context. Advancing our understanding regarding context requires then taking the epistemological underpinnings of clinical reasoning into consideration. Context, in this respect, is far from being a fixed entity. Instead, its meaning is actively produced, negotiated and legitimised within the discourses of different research traditions.10 Understanding the meaning of context necessitates understanding of how clinical reasoning is conceptualised.11

1.1 Why does context really matter?

In recent years, many researchers have called for addressing context.3, 12, 13 We argue, however, that empirical evidence examining the impact (or not) of context on clinical reasoning cannot be interpreted without reference to the meaning ascribed to context. Furthermore, making sense of context has important implications for the widely observed phenomenon of context specificity where successful performance in a clinical problem in one situation is poorly predictive of successful performance in a different situation.14, 15 As Eva remarks, context specificity raises a crucial matter with resonating implications for the assessment of competences of whether cognitive skills (including clinical reasoning) are indeed stable dispositions or rather context-bound states.14 This issue has obviously attracted much attention.13, 14, 16 We argue that the path forward involves putting the meaning of context into perspective. Lastly, in an era of growing diversification of learning environments in medical education, examining context could reveal the potential implications these diverse learning contexts have on the learning of clinical reasoning.4

1.2 Aim of this article

In previous work, we identified three main conceptualisations of clinical reasoning each grounded on particular epistemological assumptions.17 The current work builds on this. Our intention is to provide a sound epistemological framework of clinical reasoning that puts context into perspective and demonstrates how context is understood and researched in relation to clinical reasoning. Possible ramifications concerning the issue of context specificity and learning of clinical reasoning are explored. The proposed framework offers a roadmap to navigate the complex terrain of clinical reasoning literature. It may be worth emphasising that this article is a critical appraisal of the literature and not a review of the available empirical evidence.

2 EPISTEMOLOGICAL DIMENSIONS OF CONTEXT

We suggest three main epistemological dimensions of context in relation to clinical reasoning; ‘Representation’, ‘Interaction’ and ‘Interpretation’, each of them corresponding to ‘fundamental patterns of knowing’.18 (Figure 1) Under each dimension, we identify the main theories of clinical reasoning (corresponding to particular research traditions) and position them inside the illustrated triangle, according to their epistemological claims (Figure 2). Tables 1 and 2 summarise important points. Theories are not static constructs but are constantly advanced, modified and interpreted in different ways. In this respect, the positions of different theories inside the triangle in Figure 2 should not be understood as exact coordinates but as rough approximations.

​

The epistemological dimensions of context in clinical reasoning

The main theories of clinical reasoning in relation to the epistemological framework

TABLE 1. Summary of the three epistemological dimesions of context Epistemological dimension Main theoretical affiliations Knowledge Elements of interest Main methodologies Representation Problem-solving and decision-making cognitive psychology Mental representations Long-term and working memory Mostly Quantitative (ie controlled interventions) Interaction Situated & Distributed cognition, Ecological psychology Situated in activity Patterns in Interactions Both quantitative qualitative (ie statistical correlations, Critical Decision Analysis, Ethnography) Interpretation Narrative theory Rhetoric Hermeneutics, Phenomenology Socially negotiated Language and Culture Mostly Qualitative (ie Narrative analysis, Discourse analysis Ethnography) TABLE 2. Context metaphors Epistemological dimension Clinical reasoning Context metaphor Representation Act of categorisation Context as a container encasing the mind4 Interaction Cognitive state emergent from interaction Context as a complex natural ecosystem Interpretation Act of intersubjectivity and professionalisation Context as the light immersing the various objects in a room 3 REPRESENTATIONAL DIMENSION

A dominant line of thought originates from the problem-solving tradition in cognitive science and considers clinical reasoning an act of categorisation. The physician recognises salient features of a disease, which, in turn, activates the recall from long-term memory of the corresponding category that best fits these features.19, 20 This pattern recognition constitutes the basis of non-analytical reasoning, affording the physician to make inductive inferences from knowledge acquired ‘from previous experience with respect to a given class of stimuli and events’ which ‘gives meaning and guides anticipation with respect to similar stimuli and events in the future’.21 Non-analytical reasoning enables making a prediction (diagnosis) for a new patient based on features observed in previously encountered patients.

Much work in cognitive science has concentrated on identifying the cognitive structures responsible for the organisation of past experience in long-term memory. Such cognitive structures, or mental schemas, constitute mental representations of previous experiences organised in particular ways. Prototypes, exemplars, semantic qualifiers and illness scripts represent different approaches to how mental representations are organised in memory.19, 20, 22, 23

Not all reasoning is non-analytical. Apart from mental schemas, researchers have postulated the existence of cognitive processes that retrieve, transform and store these schemas.24 In Hypothetico-deductive thinking, for example, a diagnostic problem is solved by the generation initially of a limited number of tentative hypotheses that determine subsequent clinical data collection to deduce or refute these initial hypotheses. In recent years, dual-process theories embracing both analytical and non-analytical pathways have gained increasing popularity. Dual-process theories originate from a separate research tradition, namely behavioural decision-making research that is not concerned with diagnosis as categorisation, but as opinion revision under conditions of uncertainty.25, 26 These theories postulate the coexistence of two minds; one tacit, rapid and intuitive, and the other analytical, effortful and conscious.27 While the overlap between dual-process and knowledge organisation theories is debated,28 there is growing consensus in both research traditions that physicians employ analytic and non-analytic strategies interchangeably when facing a diagnostic problem.15

There are significant differences between these two theories that are beyond the scope of this article. Nevertheless, they share common epistemological ground as both implicitly acknowledge the existence of an external, pre-given world, reconstructed in mental representaions.29 The pragmatist philosopher Richard Rorty wonderfully articulated this mind-world dichotomy between ‘stuff happening out there’ and mental operations ‘inside the head’:30

To know is to represent accurately what is outside the mind; so to understand the possibility and nature of knowledge is to understand the way in which the mind is able to construct such [internal] representations.31

From this perspective, research endeavours to unravel the essential ‘principles’ of how the physician's mind represents the pre-given world and acts upon such representations. These principles are carved out by systematically manipulating individual variables under tightly controlled experimental conditions. The rationale of the experimental method is to reduce and control context so that it acts solely as a stimulus for the particular mental process under investigation. It would be unfair to claim that cognitive studies reject the impact of context entirely. Context, however, assumes significance inasmuch as it is relevant as a stimulus (input) for the mental task under investigation. As some claim ‘while some context will often be required to assess a component [of cognition], there are always bounds on how complex such a context need be’.32 Context should be kept simple and relevant to the cognitive task under study. From this point-of-view, studying reasoning in naturalistic settings is problematic at best. Unreduced context from ‘real-life’ complexity would serve as a too broad, non-specific stimulus eliciting a multitude of mental operations in parallel underlining the view that ‘the more complex a phenomenon, the greater the need to study it under controlled conditions, and the less it ought to be studied in its natural complexity’.33 Furthermore, such complexity risks overwhelming working memory and impair learning. Context in this non-reduced form is inevitably disregarded as noise.34

Clinical reasoning research in line with the tradition of cognitive science is predominantly conducted under strictly manipulated experimental conditions, where context is reduced and objectified into measurable factors. Context assumes the quality of a ‘factorable’ and ‘quantifiable’ external world sharply demarcated from human activity and cognition. The measurable factors are regarded as having unambiguous meaning across different situations and cultures; they are universally valid. There is an abundance of studies investigating the impact of contextual factors on physicians’ diagnostic accuracy (proxy for problem-solving) or on prescriptive models of rationality (proxy for decision-making). Examples of quantifiable factors include time pressure,35, 36 repeated work interruptions,37 patients with disruptive behaviours,38, 39 patients’ demographics, physician's years of experience,40 or recent exposure to media information about a particular disease.41

4 INTERACTIONAL DIMENSION

Recent years have seen a surge of studies that take a different perspective.2, 42-44 These studies share a strong scepticism towards the mind-world dichotomy of cognitivism. They view instead cognition as manifesting itself in activity and as contingent on mind-world interaction. They draw their theoretical framework from Situativity Theories and on the work of the pragmatist philosophers Dewey, James and Mead and the psychologist Bartlett.24, 30, 45, 46 Knowledge is not something solely residing in organised cognitive structures in a mind sharply demarcated from an external world. In this sense, it cannot be divorced from the world as it is situated in activity and can only be actualised through mind-world interaction. Overcoming the dichotomy of knowing and doing entails that knowledge and cognition emerge from the particularities of the given situation.47

This epistemological shift from moving the unit of analysis from the individual mind to the relation between the mind and its environment has clear repercussions. Clinical reasoning embodies an activity of creating meaning ‘on the fly, rather than reading it back from something (representation or schematic) stored in the head’.48 This does not mean that situativity theorists deny the existence of cognitive structures. Rather they place these structures in the interaction between the individual and the context. The physician's knowledge is entangled in the activity of providing care for the patient.49

We identify Situated Cognition,24 Ecological Psychology50 and Distributed Cognition,51 as well as the separate research tradition of Situated Learning and Activity Theory52 as theories having relevance to clincial reasoning in medical education. It should be emphasised here, however, that they make up an agglomeration of theories with varying claims regarding the structure of mind-world interaction, which is beyond the scope of this article. To better illustrate this heterogeneity, the different Situativity Theories have been dispersed accordingly along the ‘Interaction’ side of the triangle accordingly in Figure 2. Those theories placed in proximity to the ‘Interpretation’ side, such as Activity Theory and Situated Learning give primacy to the social world (society in the mind), and will be discussed in the following sections. Distributed and Situated Cognition as well as Ecological Psychology, discussed in this section, take a balanced position between society and mind and are therefore placed in the middle of the ‘Interaction’ side of the triangle.51

Situated Cognition argues that clinical reasoning (cognition) emerges from moment-by-moment interaction with the environment.46 In Ecological Psychology, the environment provides certain enabling conditions (affordances) that make possible but do not determine certain types of individual activity (effectivities).53 Distributed Cognition assumes a systemic perspective where clinical reasoning is distributed in the interrelations of interacting people and artefacts, becoming a property of the system itself. In this respect, clinical reasoning at the individual level constitutes only a small piece of clinical reasoning operated at the system level.54

While these three theories differ in their focus of analysis, what they share in common is the view that clinical reasoning is determined by the interdependency of the various factors of the system (physician, patient, artefacts, physical settings) specific to the particular situation. This affords for non-linearity and places clear limitations on determining the outcome of clinical reasoning in every situation (predictability).34

Invariably, such epistemological claims have significant consequences for the meaning of context. The environment is not regarded as a fixed source of objectified and measurable inputs whose reception triggers cognitive operations. Context emerges from activity in the sense that the environment ‘is instead interactive and responsive to the agent's actions’.46 This reciprocity between agent and environment places context on equal footing with the agent as a co-constructor of the activity whereby context enables or restraints this acticity.2 Context in other words ‘isn't just there, but is actively produced, maintained, and enacted in the course of the activity at hand’.55 From this angle, researchers have investigated how the diagnostic outcome of a clinical case is related to the idiosyncratic attunement of the particular affordances with the effectivities of the physician solving it (Ecological Psychology) or to the contextual factors that alter the diagnostic outcome of patient-doctor interaction (Situated Cognition).12, 56, 57

4.1 Crossroads of representation and interaction

Naturalistic decision-making (NDM), is a distinct research tradition examining decision-making everyday practice settings.58, 59 One of the better known theoretical frameworks in NDM is Klein's recognition-primed decision model (RPD),60 based on empirical work on experts such as firefighters, military personnel and intensive care nurses managing complex real-life situations under conditions of uncertainty. According to the RPD model, practitioners draw on a wide repertoire of mental schemas formed from previous experience. Schemas fitting the situation at hand are intuitively activated. Such schemas enable practitioners to make sense of the situation, by suggesting patterns of causal relations, appropriate goals and action plans encoded in the schema itself. The suitability of the action plan is assessed before the schema is adopted. Reasoning has a biphasic character: an initial intuitive ‘skilled recognition’ of the appropriate schema followed by a deliberate act of assessment.

NDM lies at the crossroads between two epistemological dimensions. It acknowledges the mind-world divide and embraces a representational view of knowing by advocating that the experience of an external world becomes consolidated in knowledge structures (mental schemas). Likewise, it strives to carve out generalisable cognitive rules and principles. On the other hand, it distinguishes itself from experimental psychology in the sense that it recognises the pervasive impact of context on cognition by emphasising the role of action. Cognition cannot be separated from the environment inside which it occurs.60 Experts face ill-defined complex problems, under time constraints, with only incomplete information, thus burdening them with often competing or shifting goals and no clear-cut solutions.61 They operate, under conditions of uncertainty that are inherently complex, volatile and dynamic.62, 63 Clinical reasoning in this respect involves taking the best justified action in a specific context.64 Studying such cognitive processes in experimentally controlled settings deprives the opportunity to register how this natural complexity impinges on cognition.

This difference between cognitive science and NDM lies beyond a mere methodological dispute on the ecological validity of laboratory experiments.63 It highlights a deeper rift between two distinct research traditions that concerns the importance of context. In NDM research, it is the authentic world in its full complexity that puts higher cognitive functions into action. Context, embodied by the ill-defined, complex real-life situations, enables the emergence of ‘macro-cognitive processes’, such as situation awareness, planning, problem detection, and uncertainty management.65, 66 Context drives cognition implying that cognition cannot be studied in isolation in the experimental laboratory. Reductionistic experimental techniques break down context in manageable input units, and eliminate the very phenomenon that actually brings these ‘cognitive systems in context’ to life.67 For this reason, NDM is positioned at the lower left angle of the triangle (Figure 2). NDM has been applied in the study of experts’ decision-making in various healthcare-related contexts, such as intra-operative decision-making,68 laparoscopy,69 emergency medicine,70, 71 and critical care.72, 73

5 INTERPRETATIONAL DIMENSION

The scholarship of clinical reasoning through the lens of sociology of medicine, anthropology of medicine and the humanities follows a different path. Clinical reasoning is regarded as an interpretative social practice, embedded in culture and coming into existence through language.74 Language is not merely considered a neutral linguistic medium conveying information (content) but assumes a central role in the study of how physicians reason. Meaning imparted by language is socially and culturally situated and dependent on the particular social and cultural contexts.75 Language, however, is also a social act. ‘Learning sanctioned ways of talking’ legitimise entry into a professional community.76 Learning to think like a physician entails learning to talk like a physician. Language and thought are inextricable.77 Language in this sense is both situated and situating. It both reflects and constructs the cultural community it serves.78 This fundamental entanglement of thought, language and culture has tangible implications for clinical reasoning and context. It is not simply what is said (the content) but how it is said that becomes important.79, 80

Two forms of language of particular attention are rhetoric and narrative.81 Rhetoric concentrates on the persuasive nature of verbal and written communication. Contextual elements of communication such as the occasion, audience and purpose of the communication assume primary importance.82 Researchers for example have demonstrated how medical students tacitly learn to master specific rhetorical strategies in case presentations as a means of achieving credibility.78 Mastering such rhetoric strategies becomes an integral aspect of developing a professional identity allowing legitimate access to the medical community. Importantly, rhetoric acts reflect tacit assumptions concerning which reasoning strategies are more credible. Students internalise these strategies during their indoctrination into the medical profession.83 Indeed, as practitioners ‘construct’ their talk ‘to match the audience, context and purpose’84 their narratives come to implicitly favour particular ways of reasoning, which embody assumptions of rationality, that have legitimacy in the scientific community.84-87 Reasoning strategies such as hypothetico-deductive reasoning signify, to paraphrase Lingard, sanctioned ways of thinking and talking.76

Narrative reasoning which Bruner carefully distinguishes from hypothetico-deductive reasoning denotes a fundamental way of thinking involving the construction and interpretation of narratives.75 Medical practice is awash in narratives of talking about patients (narrative construction) or talking with patients (narrative interpretation).87 Narrative knowing embodies a retrospective construction of a narrative integrating the patient's illness experience into a meaningful plot where events, beliefs, motives and actions are placed in a temporal and meaningful (causal) relation that render the experience plausible and comprehensible to the practitioner.85 It is the plot that makes the individual events comprehensible88 and provides the background against which human agency can be understood. A number of authors have investigated narrative reasoning in clinical practice and how, in particular, the practitioners’ narratives emphasise particular reasoning strategies.86, 87, 89

Context grounded in this epistemological dimension, expresses the world of experience or, as Sandberg and Tsoukas remark, ‘the meaningful totality into which practitioners are immersed […] in which things, people, actions and options already matter in specific ways’.90 Clinical reasoning in this sense cannot be accounted for solely by focusing on intrapsychic (cognitive) dispositions without taking account of the world in which such depositions exist.75

Two clarifications should be made at this point. First, it would be unfair to claim that other research traditions dismiss the value of interpretation in clinical reasoning. The majority of theories elaborated in previous sections do indeed acknowledge some significance to how individuals give meaning to contextual stimuli of the situation.21, 32, 91 This concern in cognitive science discourse is reflected in the concept of construal, which signifies the interpretation of information in light of one's own subjective understandings.92 It has its roots in Piaget's famous words that ‘intelligence organizes the world by organizing itself’ signifying that the mind provides the categories of knowing while experience provides the content.93 Cognitive structures give meaning to experience and guide anticipation of similar stimuli in the future.21 The meaning of interpretation implied in this section, however, departs from this representational ground of clinical reasoning as an act of categorisation and adopts a phenomenological-hermeneutical epistemology where clinical reasoning embodies the intersubjective, social practice of making sense of the patient's illness experience.

The other point concerns the decision to discriminate the interpretative from the interactive epistemological dimension. Does not interpretation, seen as participation in a social world, presuppose an interactive element with this world as well? Indeed, it does. However, the meaning of interaction discussed in the previous section emphasises a connectionist aspect, whereby clinical reasoning is an emergent state determined by the interaction of cognition with elements in the environment (Situated Cognition and Ecological Psychology) or by the distribution of cognitive operations over a network of agents and artefacts (Distributed Cognition). On the contrary, clinical reasoning in this view emphasises engagement and participation in a social world. This participatory connotation distinguishes it from the connectionist (interactive) dimension described previously.

5.1 Crossroads of interaction and interpretation

Two relevant theories lie at the intersection between interpretation and interaction: Situated Learning and Activity Theory. These theories have their routes in historical-cultural psychology originating from the works of Vygotsky, Luria and Leontief and further developed by Lave,94 Cole95 and Engeström.52 Both theories place particular emphasis on both meaning and interaction. Situated Learning, being primary a learning theory, will be discussed later in relation to learning of clinical reasoning.

Activity Theory regards cognition as goal-oriented action that cannot be understood without considering the broader activity system in which it is embedded. An activity system includes individuals (subjects) and tools (artefacts) to achieve the individual's goals (objects) while simultaneously being informed by cultural, social and organisational norms (rules, community, division of labour). These interconnected systemic elements have the potential to conflict with each other underlining the instability and volatility of activity systems.96 Goals, plans and actions do not exist in isolation in the subject's mind; they are embedded in a system of relations within a sociocultural context from which they derive their meaning.94 As Engeström remarks, ‘the notion of situation is alone insufficient as a unit of contextual analysis of clinical cognition’.52 He sees clinical cognition instead as being embedded in ‘broader institutional cultures and long-term historical trajectories of development and change’.

Cognition manifested through action derives its meaning from the system in which it occurs (interpretative dimension) and is distributed across the elements of the activity system (interactional dimension). Clinical reasoning becomes a distributed activity in wider historical-cultural systems.97 What takes place in the activity system itself is the context, irreducible to enumerable external factors but yet constituted through the enactment of the activity in the system.42 It is worth noting that despite the robust theoretical framework provided by Activity Theory there is a surprising scarcity of empirical studies in medical education.97

6 REFLECTIONS ON CONTEXT SPECIFICITY

The thorny issue of context specificity remains even today without a satisfying and broadly accepted explanation.13, 98 But should context specificity be regarded as a problem in the first place? At a time when researchers were postulating on the existence of universal problem-solving strategies, it was Elstein's pioneering work that brought the problem into awareness.99 Elstein's work demonstrated that physicians employed the hypothetico-deductive method to solve clinical problems. The method, however, was all too general and non-specific. Everybody was doing the same thing irrespective of level of expertise.20 More importantly, it did not correlate with better diagnostic accuracy.100 Success at one case was a poor predictor for success in the next.

Attempting to interpret this finding, he labelled the phenomenon content-specificity suggesting that effective problem-solving was ‘to be found in the repertory of their [physicians’] experiences, organized in long-term memory’ rather ‘than in differences in the planning and problem-solving heuristics employed’.99 This idea proved influential enough to stimulate research on the organisation of knowledge in the mind.101-104 Finding the correct diagnosis became an issue of organisation of content knowledge, in essence knowledge from previous experience organised as mental representations in memory.19, 26, 100, 105, 106 However, it soon became evident that while mastery of the content was necessary for diagnostic accuracy, it was not sufficient to explain this case-to-case variation in performance.16, 98, 106 For example, evidence suggested that slightly modifying the presentation of a clinical case or even presenting exactly the same case on a different occasion affects diagnostic accuracy.16, 107 The way content was ‘framed’ in a particular case affected how physicians responded to the case.