Example 3. Learning diagnostic reasoning
The final example is taken from my own work. When I began my research on knowledge organisation in memory and clinical reasoning, I turned to psychology and chose prototype theory, an accepted theory of knowledge organisation at the time, as a conceptual framework. Prototype theory portrays knowledge organisation as categories that have representative (prototypical) exemplars at their centre, that serve as anchor points for the remaining members of the category (e.g. pneumonia, common cold, otitis media and HIV are prototypical exemplars of infections).14 Based on prototype theory, I hypothesised that representative exemplars of medical categories would be ‘recalled faster and more accurately than peripheral members’.14 The results from four experimental studies with medical students and experienced doctors confirmed these hypotheses.15 Other researchers, by contrast, used an alternative conceptual framework to explore disease categorisation, namely, the notion of particular instances as anchor points for knowledge in memory, such as patient characteristics and context of care.16,17 The two frameworks illustrate how different conceptual frameworks illuminate and magnify certain aspects of the situation and disregard others. For example, different conceptual frameworks highlight different variables, such as family resemblance in prototype theory compared with context characteristics in instance theory, and different outcomes, such as response time versus diagnostic accuracy.
Prototype theory also predicts that prototype formation will be facilitated by early exposure to representative and intermediate-level exemplars (e.g. pneumonia), as distinguished from the broad range of members in a category, and to superordinate classes (e.g. viral or bacterial infections). A follow-up observational study confirmed that prototype formation during organ-system courses was greatest in courses concentrating on fewer and more intermediate-level disorders, leading to the conclusion that ‘less is more’ when it comes to prototype formation.14,18
After this initial foray into knowledge organisation, I became interested in the nature of the relationships that link knowledge in memory.14 Structural semantics, borrowed from anthropology, provided a better theoretical framework than prototype theory to study the inter-relatedness of knowledge in memory. According to structural semantics, the networks of relationships are represented by dichotomous abstract qualifiers called semantic axes; for example, a severe, acute, lower right quadrant abdominal pain related to appendicitis and ectopic pregnancy, whereas a mild, chronic, upper left quadrant pain related to stomach, colon, spleen or pancreatic disorders. The relationships, as expressed by abstract qualifiers (severe–mild, acute–chronic…), can be readily observed and measured in the clinical discourses of doctors as they work through cases. In a series of think-aloud, qualitative studies for categories of neurological, gastrointestinal, intensive care and rheumatological disorders, the clinicians who exhibited greater diagnostic accuracy and more comprehensive understanding of the clinical issues, be they students or experienced doctors, were those who expressed the greatest and most diversified sets of semantic qualifiers.19,20 These results eventually led to the expansion of the theory itself by proposing four types of discourse organisation:
reduced discourses (few semantic qualifiers and short discourses);
dispersed discourses (few semantic qualifiers but extensive discourses);
elaborated discourses (many semantic qualifiers and extensive discourses), and
compiled discourses (many semantic qualifiers and concise discourses).14,21
Gruppen et al.22 showed that diagnostic accuracy is greatly enhanced by considering the correct diagnosis early during the clinical encounter, as early as the chief complaint. They used a staged information-processing model of diagnostic problem solving and errors (i.e. information gathering versus information integration) as their conceptual framework, to which they added Bayes’ theorem.22 In relation to their findings, we hypothesised that diagnostic accuracy would be associated with greater use of semantic qualifiers to characterise the chief complaint. The results of a case-control study23 showed that indeed diagnostic accuracy was associated with greater use of semantic qualifiers to mentally represent the chief complaint, a result that mirrored findings from mathematics education research24 showing the key role of abstract problem representation in solving algebra problems.
Although conceptual frameworks can be used to frame a problem or guide solutions, as in the previous examples, this example illustrates how conceptual frameworks can be applied to design studies and to interpret study results, as well as how theories themselves can be challenged. Theories are dynamic entities,25 as was illustrated in the further classification of clinical discourses according to their semantic content and the extent of the discourses, a new addition to structural semantics at the time. Competing theories, such as abstractions versus instances in prototype theory, emerge and challenge one’s own work, leading to new hypotheses and theories. This dynamic interplay constantly helps move the field forward, rather than having it descend into stagnation and dogmatism (see also Bordage14). At a broader conceptual level, Eva26 portrays knowledge organisation and clinical reasoning according to analytical and non-analytical processes, such as deliberate semantic problem representation compared with rapid pattern recognition.
Conceptual frameworks help identify important variables and their potential relationships; this also means that some variables are disregarded.
Conceptual frameworks are dynamic entities and benefit from being challenged and altered as needed.
The methodology of qualitative studies illustrates the dual role of conceptual frameworks in framing questions and interpreting results. In a deductive qualitative inquiry, a conceptual framework can be used to formulate the questions and identify important variables to be analysed. In an inductive, grounded theory approach, theories are postulated de novo as the researcher analyses the data.27–30 The newly formulated theory emerges as a hypothesis to be tested through further qualitative or quantitative studies. In this case, conceptual frameworks and critical appraisal of the literature are used post facto to interpret or contextualise the newly proposed theory. Eventually, the evidence can build up to the level of proposing a theory that can be used for prediction: a theory in scientific terms (e.g. behaviourism, gravity, evolution) embodies an idea or set of ideas that have been confirmed by observations or experiments; in lay terms, however, theories are often considered as hypotheses yet to be proven. Whether they are using conceptual frameworks to frame a problem or to interpret results, scholars are building upon one another’s work in ways that allow individual researchers to develop their own programmes of research that lead to explanatory (clarification31) studies and deeper understanding that help to move the field forward. Sporadic and opportunistic research, by contrast, runs the risk of being superficial and non-cumulative.