clinical reasoning

Clinical reasoning processes: unravelling complexity

through graphical representation
Bernard Charlin,1 Stuart Lubarsky,2,3 Bernard Millette,1,4 Françoise Crevier,5 Marie-Claude Audétat,1,6
Anne Charbonneau,7 Nathalie Caire Fon,1 Lea Hoff8 & Christian Bourdy6

CONTEXT Clinical reasoning is a core skill in then refined through an iterative validation process
medical practice, but remains notoriously difficult involving the same group of doctors, after which
for students to grasp and teachers to nurture. To other groups of clinicians were asked to solve a
date, an accepted model that adequately captures clinical problem involving simulated patients.
the complexity of clinical reasoning processes does
not exist. Knowledge-modelling software such as RESULTS A hierarchical model depicting the
MOT Plus (Modelling using Typified Objects multifaceted processes of clinical reasoning was
[MOT]) may be exploited to generate models produced. Validation rounds suggested generalis-
capable of unravelling some of this complexity. ability across disciplines and situations.

OBJECTIVES This study was designed to create a CONCLUSIONS The MOT model of clinical
comprehensive generic model of clinical reasoning reasoning processes has potentially important
processes that is intended for use by teachers and applications for use within undergraduate and
learners, and to provide data on the validity of the graduate medical curricula to inform teaching,
model. learning and assessment. Specifically, it could be
used to support curricular development because it
METHODS Using a participatory action research can help to identify opportune moments for learn-
method and the established modelling software ing specific elements of clinical reasoning. It could
(MOT Plus), knowledge was extracted and entered also be used to precisely identify and remediate
into the model by a cognitician in a series of reasoning errors in students, residents and practis-
encounters with a group of experienced clinicians ing doctors with persistent difficulties in clinical
over more than 250 contact hours. The model was reasoning.

Medical Education 2012: 46: 454–463

doi:10.1111/j.1365-2923.2012.04242.x

Discuss ideas arising from this article at

www.mededuc.com ‘discuss’

1 7
Centre for Applied Pedagogy in Health Sciences (Centre de Department of Oral Health, Faculty of Dental Medicine, University
Pédagogie Appliquée aux Sciences de la Santé [CPASS]), University of Montreal, Montreal, Quebec, Canada
8
of Montreal, Montreal, Quebec, Canada Department of Medicine, CPASS, University of Montreal,
2
Centre for Medical Education, McGill University, Montreal, Montreal, Quebec, Canada
Quebec, Canada
3
Department of Neurology, Montreal General Hospital, Montreal,
Correspondence: Professor Bernard Charlin, Médecine Direction,
Quebec, Canada
4 Centre de Pédagogie Appliquée aux Sciences de la Santé (CPASS),
Department of Family and Emergency Medicine, Hôpital Cité de la
Université de Montréal, CP 6128, succ. Centre-Ville, Montreal,
Santé de Laval, Quebec, Canada
5 Quebec H3C 3J7, Canada. Tel: 00 1 514 343 7864; Fax: 00 1 514
Consultant, Educational Engineering, Faculty of Pharmacy,
343 7650; E-mail: [email protected]
University of Montreal, Montreal, Quebec, Canada
6
Department of Family and Emergency Medicine, Faculty of
Medicine, University of Montreal, Montreal, Quebec, Canada

454 ª Blackwell Publishing Ltd 2012. MEDICAL EDUCATION 2012; 46: 454–463
 Clinical reasoning: graphical representation

rules of grammar for guiding modelling.11 It may be

INTRODUCTION
used to map clinical reasoning processes and render
How doctors develop clinical reasoning skills has knowledge networks and strategies explicit.11,12 The
fascinated researchers in medical education for goals of this study were: (i) to develop a new,
decades.1–3 Yet unravelling the complexity of clinical comprehensive representational model of clinical
reasoning continues to present a formidable chal- reasoning processes using MOT Plus software intended
lenge to teachers and learners in medicine. Part of for use by teachers and learners, and (ii) to provide
the difficulty stems from the observation that data on the validity of this model.
seasoned doctors often reason through cases using
rapid, tacit cognitive processes, at least when dealing
METHODS
with common or routine clinical situations.1,4,5
When asked to render these processes explicit, they
This research was situated in a socio-constructivist
typically find it difficult to slow them down and to
paradigm whereby the processes and guide were
retrace their cognitive steps6. It is therefore often
co-constructed with clinical educators who were
hard for medical trainees to grasp the various
involved at each step. We used a qualitative method-
dimensions and nuances of their tutors’ clinical
ology known as participatory action research, a
reasoning processes, and for medical teachers to
process of development carried out collaboratively by
foster clinical reasoning skills in medical learners.
a group of people interested in changing practice in
Another challenge routinely facing
their setting.15
medical educators is the identification and remedi-
ation of specific errors in clinical reasoning.7–10
MOT software
A model capturing the richness and complexity of
The MOT software can be described as a semi-
clinical reasoning processes would therefore be very
structured cognitive mapping tool. It includes a
useful to inform teaching, learning and assessment.
typology of knowledge objects and a typology of links.
Such a model is difficult to create because clinical
This software was developed in 1992 by the LICEF
reasoning depends on mobilising and processing vast
(Laboratoire en Informatique Cognitive et Envi-
networks of knowledge that may not be easily accessible
ronnements de Formation) Research Center, Mon-
to conscious scrutiny. Early attempts to develop
treal, Quebec, Canada.12 Its current version, MOT Plus
generic models of clinical reasoning, which depicted
1.6.7, is available in English and French, free of
its processes in a primarily linear fashion, largely fell by
charge, at www.licef.ca (under ‘Realisations/
the wayside years ago. However, modern methods of
Produits’). It has been used to help design courses, to
concept mapping11,12 have the potential to reinvigo-
promote knowledge transfer in corporations, to
rate the field and prompt significant advances in the
enhance e-learning courses and to facilitate infor-
establishment of a comprehensive model of clinical
mation management.11,12
reasoning.
A MOT model is the graphical expression of a field’s
Concept maps are graphical tools for organising and
knowledge objects and the links that unite them. The
representing knowledge.13,14 Recently, computer
grammar of MOT uses geometric symbols to repre-
software has greatly facilitated the dissemination of
sent types of knowledge objects, such as concepts
concept mapping as a research tool in many domains,
(rectangles), procedures (ovals) and principles
including medical education. However, the concept-
(hexagons). It also defines the types of link that are
mapping software in current use provides neither
permissible between the various types of knowledge.
guidance nor constraints, such that users are free to
There are six types of link: composition (C: ‘is
create an infinite array of possible models. Knowl-
composed of’); specialisation (S: ‘is a sort of’);
edge modelling software that includes grammatical
instantiation (I: ‘is an example or instance of’);
constraints during model construction may allow
precedence (P: ‘precedes’); input ⁄ product (IP: ‘is an
users to generate maps that are better organised,
input to’ or ‘produces’), and regulation (R: ‘is a
more complete, more useful in practical settings and
regulating principle for’). Using these types of
more efficient in communicating information than
knowledge objects and types of link, the MOT
their traditional non-constrained counterparts.11,12
software enables the modelling of any field of
knowledge or competence. When a competence or a
The Modelling using Typified Objects (MOT) soft-
ware and technique is an innovative tool with built-in

ª Blackwell Publishing Ltd 2012. MEDICAL EDUCATION 2012; 46: 454–463 455
B Charlin et al

knowledge field is complex, a model can be Round 2: groups of clinician-validators

expanded into several sub-models.11,12
The model produced at the end of the first round was
Participants then submitted to a sample of 12 doctors. The sample
was purposeful. To ensure the assembly of a diverse
Six clinician-educators (hereafter referred to as panel, participants were selected based on their
clinician participants [CPs]) were interviewed as a interest and experience in medical education and
group in a series of sessions occurring over more than their clinical specialties. Potential participants were
250 contact hours. The participants were selected given information on the design of the study. The
from different disciplines: four were medical doctors study was approved by the university’s institutional
(two in family medicine, one in otolaryngology, one review board and all those contacted gave informed
in geriatrics); one was a dentist with expertise in oral consent prior to participation.
and facial pain, and one was a psychologist affiliated
with a department of family medicine, who works During this second round, a printed copy of the MOT
directly with residents experiencing clinical reasoning model prepared during Round 1 was presented to
difficulties. Three participants were male and three clinician-validators (CVs). Clarification questions
were female. All CPs had > 15 years of clinical were answered and CVs were then randomly divided
experience and were educators with a particular into four groups of three participants that were sent
interest in clinical reasoning. into separate rooms. Each group encountered a
simulated patient (SP) trained to depict the same
Construction and validation of the model clinical scenario involving a chief complaint of
chronic fatigue. The simulated situation was chosen
The process of participatory action research is often because clinicians from all disciplines were felt to be
described as a spiral consisting of several reflective likely to have some familiarity with this common
rounds.15 Following this methodology, we submitted clinical presentation. In each room, one CV volun-
the MOT model to a three-round validation process. teered to interview the SP, while the other two CVs
During each round, we solicited critical reflections observed the interaction, looked at their printed
and comments from the participants, and adjusted copies of the model and took notes. If the CV
the model accordingly. This method therefore conducting the interview requested a specific exam-
enabled us to concurrently construct, amend and ination or laboratory result, the SP provided the
validate the model. relevant data. In all rooms, a CP recorded all
observable indicators of reasoning (questions,
Round 1: construction and initial validation hypotheses generated, signs, investigations, etc.). All
encounters were videotaped. When the SP left the
A cognitician, who was familiar with the MOT room, the video of the encounter was played and
software and technique, interviewed CPs to ‘extract’ the three CVs were given the opportunity to comment
and analyse their knowledge, reflections on action16 on reasoning processes they observed or might have
and cognitive pathways, and subsequently converted used instead. The goal of this activity was to place CVs
them into a graphic representation adhering to the in a clinical reasoning situation with the model in
grammatical rules of MOT. Knowledge extraction hand with which they could compare their own
and graphical construction of the model were inte- reasoning pathways.
grated and based on multiple successive iterations in
2-hour sessions held over 2 years. A concurrent Finally, all 12 CVs returned to a common confer-
process of validation occurred during this first stage ence room, where a plenary session was held in
as the cognitician, who led the discussion, helped CPs which all the models annotated by the CVs were
to unravel their reasoning processes and identify shared. The CVs were instructed to: (i) specify which
inconsistencies and gaps in the emerging model. The elements within the model were relevant, erroneous
process was repeated until all participants felt that the or missing; (ii) determine if the model was congru-
model was ‘saturated’ (i.e. that it provided an ent with their perceptions of their own clinical
adequate depiction of their conscious cognitive reasoning processes; (iii) discuss the strengths,
pathways during clinical encounters). During ses- weaknesses and potential missing elements of the
sions, CPs were also encouraged to identify important model, and (iv) discuss the utility of the model for
definitions and concepts from the clinical reasoning learning and teaching clinical reasoning. The entire
literature that they felt were useful for constructing second round of validation took approximately
and validating the model. 4 hours to complete.

456 ª Blackwell Publishing Ltd 2012. MEDICAL EDUCATION 2012; 46: 454–463
 Clinical reasoning: graphical representation

Round 3: CPs and the cognitician than ‘diagnosis’ better described the product of
initial phases in clinical reasoning.
In this final round of model validation, data obtained
in the second round were used by CPs and the Types of knowledge used
cognitician to make final adjustments to the MOT
model with the added aim of producing a model that There is growing agreement among researchers and
would be detailed enough to reflect the multi- medical educators that several types of knowledge
dimensional structure of clinical reasoning, but interact in a medical encounter.1,4,19,20 The CPs felt it
would not be prohibitively unwieldy. In other words, was necessary to specify these in order to produce a
CPs were requested to fine-tune the model such that model that would be useful for teaching and learn-
important processes and concepts that emerged in ing. The illness script theory18–23 assumes that
Rounds 1 and 2 were included, but excessive details knowledge networks adapted to clinical tasks develop
that might render the model cumbersome or through experience and operate autonomously
impractical to use as a teaching and learning tool beneath the level of conscious awareness. These
were avoided. networks consist of associative links among illnesses
and their attributes, consequences, investigation or
treatment, and links with memories of previously
RESULTS encountered exemplars of the illness (instances). In
ambiguous situations, clinicians search for a fit
The Modelling using Typified Objects modelling between the available information and appropriate
sessions confirmed that clinical reasoning is a highly scripts.23,24 The CPs reported that, in clinical situa-
complex and multifaceted process. The CPs were all tions, they effectively mobilise specific knowledge
seasoned educators, who were well versed in classical (scripts) relevant to the particular situation (e.g.
clinical reasoning concepts, such as hypothesis gen- knowledge related to vertigo and then knowledge
eration or hypothetico-deductive reasoning, but in linked to cervical masses when they see two successive
the early stages of modelling they realised that certain patients whose main complaints are, respectively,
concepts required operational definition to ensure a dizziness and progressive swelling of the upper neck).
common base of language and understanding among The group reported that clinical reasoning occa-
themselves during the construction and validation of sionally taps into sources of knowledge of a different
the model. Space limitations preclude the inclusion nature and that the concept of knowledge encapsu-
in this paper of all discussed concepts. The following lation20,25 describes that type of knowledge well. In
section defines concepts taken from the medical the model, these latter sources of knowledge are
education literature that prompted discussion among referred to as ‘biological, psychological and socio-
CPs and were incorporated within the model. logical knowledge’.

Concepts Semantic transformation, problem solving, problem

representation and metacognition
Salient concepts that emerged in the research process
concerned diagnosis versus categorisation, types of For cognitive psychologists,26 problem solving in-
knowledge used in clinical encounters, semantic volves processes that transform the initial state of the
transformation of information, problem solving, problem into a state in which the goal is achieved. The
problem representation and metacognition. key to solving a problem is to represent it in such a way
that the required processes to understand and solve it
Diagnosis versus categorisation can apply.26 Problem representation has a crucial role
in problem solving.26,27 It was clear to participants
Cognitive psychologists consider that diagnosis is a that the clinician’s representation of a patient’s
categorisation task that consists of placing patients’ problem changes over the course of the encounter
illnesses in different classes according to their attri- with the discovery of additional data. Semantic trans-
butes.2,17,18 Because for a doctor ‘the diagnosis’ is a formation emerged as another key concept.28 Clini-
precise entity and because the primary function in cians ascribe meaning to the presenting symptoms
many medical encounters (such as in emergency and clinical findings by transforming the data using
medicine) is to understand the situation enough to semantic qualifiers that represent conceptualisation
begin action, such as investigation or treatment, or abstraction of the clinical findings. For instance, ‘a
rather than to obtain a precise diagnosis, CPs agreed patient’s painful, swollen, right knee that began
that ‘categorisation for the purpose of action’ rather 2 nights ago with attacks 2 and 9 years ago’ is

ª Blackwell Publishing Ltd 2012. MEDICAL EDUCATION 2012; 46: 454–463 457
B Charlin et al

transformed in ‘an acute, recurrent attack of abrupt, orient the process ‘Categorise for the purpose of
nocturnal and extremely severe pain in a single, large action’. This process has two possible outcomes:
joint’. Metacognition refers to an individual’s knowl- ‘Categorisation is suitable for purposeful action’ and
edge concerning his or her own cognitive pro- ‘Categorisation is NOT suitable for purposeful
cesses.2,29 In medical practice, the application of action’. The former outcome feeds another process,
metacognition implies that the clinician is aware of ‘Implement purposeful action’, which has ‘Investiga-
encounter goals and verifies that his or her cognitive tions’ and ‘Therapeutic interventions’ as outputs.
processes and actions are contributing to the Results of these actions are then evaluated. If
achievement of the goals of the clinical encounter. satisfactory, the episode of care is terminated. If not, the
categorisation process or the therapeutic interven-
The MOT model tions put in place are reconsidered.

The MOT model is hierarchical. It is depicted on six Within Fig. 1, the core clinical reasoning processes
screens that show, respectively, a main model and five are depicted mainly along the screen’s horizontal
sub-models. Four screens are presented here. axis. However, the vertical axis of the figure is also
The complete, updated model (with appendices) important. Doctors’ repertoires of knowledge are
can be accessed at www.medent.umontreal.ca/ represented beneath the series of core processes by
clinical-reasoning/. Figure 1 features the main mod- rectangles entitled ‘Clinical knowledge organised for
el, a graphical representation of the core clinical action (= illness scripts)’ and ‘Biological, psycholog-
reasoning processes that emerged during develop- ical, sociological knowledge’. Links show that the
ment of the model. Beginning from the left side of latter underpins the former and that each clinical
the screen, the concepts (rectangles) ‘Context’ and encounter modifies and enriches illness scripts.
‘Patient’ are rich input sources for the process (oval) Above the series of core processes, problem repre-
‘Identify early cues’. The product of this process, sentation is horizontally depicted by three rectangles
‘Initial data’, then feeds into the process ‘Determine designating, respectively, ‘Initial representation’,
the objectives of the encounter’. These objectives ‘Dynamic representation’ and ‘Final representation’.

Throughout the process, the doctor considers

Semantic transformation of data Regulate one's own the problem from the patient's perspective,
is a key process throughout the cognitive process R taking into account the patient's beliefs, fears,
clinical encounter (Metacognition) expectations, choices, and feelings, as well as
R the impact of the problem on the patient's life

R IP
IP

Initial representation Final representation

Dynamic representation of the problem of the problem
of the problem

IP IP
IP IP
IP
Context IP Objective(s) of the
IP Categorisation is suitable IP

encounter for IP
IP Investigations
IP IP
purposeful action IP
IP

IP
Determine the Implement Evaluate the
Identify early cues Categorise for the purposeful action
objectives of the IP
results
purpose of action
encounter
IP
Categorisation is
NOT suitable
IP
IP IP for purposeful action Therapeutic IP

Patient Initial data interventions

IP IP

Find alternative
hypotheses
Each clinical encounter Clinical knowledge organised for action
modifies and enriches R
( = illness scripts)
IP

illness scripts.

IP

Organise knowledge for clinical

Clinical experience IP
action IP

IP

Biological, psychological, sociological knowledge

Figure 1 Graphical representation of core clinical reasoning processes; main model

458 ª Blackwell Publishing Ltd 2012. MEDICAL EDUCATION 2012; 46: 454–463
 Clinical reasoning: graphical representation

Above all is the metacognition process. A regulating produces as outcome (another rectangle) the
principle, contained in a hexagon, describes the ‘Patient’s needs according to the patient’. ‘Request
importance during clinical reasoning of taking into from the referring doctor’ (an input) feeds the
account the patient’s perspective and the impact of process ‘Acknowledge the request’, which produces
the illness on the patient’s life. as outcome ‘Patient’s needs according to the refer-
ring doctor’. ‘Identify the patient’s needs’ is a
On the site (www.medent.umontreal.ca/clinical- process that results in ‘Patient’s needs according to
reasoning), clicking on red arrows allows the user to the doctor’. The four outcomes feed the fifth
navigate between the main model and the five sub- process, ‘Establish priorities’, which ultimately pro-
models that depict specific processes from the main duces the ‘Objective(s) of the encounter’. This
model in greater detail. Examples of these sub- process is regulated by a principle (contained in a
screens are shown in Figs 2–4. hexagon): ‘When establishing priorities, factors such
as urgency, appropriateness and efficiency must be
Figure 2 depicts the five processes (ovals) that considered.’
underlie the process ‘Determine the objective(s) of
the encounter’: (i) ‘Clarify the patient’s request(s)’; Figure 3 depicts the sub-model ‘Categorise for the
(ii) ‘Acknowledge the request’; (iii) ‘Acknowledge purpose of action’. This process is informed by
the information’; (iv) ‘Identify the patient’s needs’, patient data and the objectives judged to be relevant
and (v) ‘Establish priorities’. As shown, ‘Patient’s in the encounter. It bears upon repertoires of
request(s)’ (input, represented by a rectangle) feeds knowledge for action (illness scripts). It is composed
the process ‘Clarify the patient’s request(s)’, which of a succession of six sub-processes: (i) ‘Search for

When establishing priorities,

factors such as urgency,
Clinical information provided by a appropriateness, and effciency
Patient's Request from the
third party (e.g. another professional Initial data must be considered
request(s) referring doctor
or a patient's relative)

IP R
IP IP IP

Clarify the
Acknowledge the Acknowledge the Identify the Establish Objective(s) of the
patient's IP
request information patient's needs priorities encounter
request(s)

IP IP
IP IP

IP
Patient's needs Patient's needs Clinical information Patient's needs
according according to the provided by the third according
to the patient referring doctor party to the doctor

IP

IP

IP

Figure 2 Process: ‘Determine the objective(s) of the encounter’

Objective(s) of Initial hypothesis and New set of activated scripts and Categorisation is suitable
Patient associated activated script their associated attributes for purposeful action
the encounter
IP
IP IP
IP IP IP
IP IP

Evaluate the fit between

Search for a Deliberately search Deliberately generate Orient search for further the clinical data and the
Search for relevant
pattern or an for an explanatory other explanatory data to confirm or refute attributes of the activated
data
analogous case hypothesis hypotheses hypotheses script(s)

IP IP IP IP
IP IP
IP
IP

Initial clinical data No initial hypothesis New clinical data

Categorisation is NOT suitable for
purposeful action
IP
Clinical knowledge
organised for action
( = illness scripts)

Figure 3 Process: ‘Categorise for the purpose of action’

ª Blackwell Publishing Ltd 2012. MEDICAL EDUCATION 2012; 46: 454–463 459
B Charlin et al

Investigations
Evidence-based Particularity
Categorization is suitable data of the situation
for purposeful action Therapeutic
interventions
C
Practical Particularity Clinician IP

considerations IP of the patient experience

IP IP
Activated script(s)
IP IP
IP

IP
Address the Implement
Consider factors that might Evaluate the fit between Discuss the
INVESTIGATION and investigation and/or
influence the choice of the clinical data and the option with the
INTERVENTION therapeutic
particular investigative or attributes of the relevant patient
components of the intervention plan
management options options
activated script(s)

IP IP
IP IP IP
IP IP IP
IP
Investigation plan IP
Investigative and Relevant options Selected options
therapeutic options
Therapeutic
intervention plan

Figure 4 Process: ‘Implement purposeful action’

relevant data’; (ii) ‘Search for a pattern or an

DISCUSSION
analogous case’; (iii) ‘Deliberately search for an
explanatory hypothesis’ (if there is no initial hypoth-
Toward a new ‘enriched’ generic process model of
esis); (iv) ‘Deliberately generate other explanatory
clinical reasoning
hypotheses’; (v) ‘Orient search for further data to
confirm or refute hypotheses’, and (vi) ‘Evaluate the
The present study represents a modern revival of
fit between the clinical data and attributes of the
attempts to model generic clinical reasoning path-
activated scripts’. The activation of hypotheses and
ways.3,30–32 Our own model, derived through a com-
their associated scripts gives access to script attributes
puter-generated cognitive mapping methodology,
that inform the oriented search for data to confirm or
depicts clinical reasoning as a series of highly complex,
refute hypotheses. The process ‘Categorise for the
multidimensional, non-linear processes that depend
purpose of action’, composed of these six sub-
on the mobilisation of specific knowledge held in
processes, has two possible outcomes: ‘Categorisation
long-term memory. Our view is that the MOT model is
is suitable for purposeful action’ and ‘Categorisation
not at odds with current concepts regarding the
is NOT suitable for purposeful action’. Subsequent
quality and content of the knowledge base, or postu-
processes are launched by each of these outputs.
lated mechanisms such as hypothesis generation or
problem representation; in fact, it encompasses them
Figure 4 depicts the series of processes that are
and are necessary elements of the model.
launched by the output ‘Categorisation is suitable for
purposeful action’. Investigation and treatment compo-
Research has shown that clinicians follow divergent
nents of selected script(s) are then activated. Factors that
lines of reasoning when solving a problem33 and that
might influence the selection of particular investiga-
there rarely exists a ‘single correct pathway’ to a
tive or management options, such as practical con-
clinical solution. Nevertheless, our study suggests that
siderations (e.g. cost, availability), evidence from the
doctors across different disciplines traverse common
literature, patient profile and preferences, the par-
cognitive signposts when reasoning through cases in
ticularity of the situation, and the clinician’s level of
their respective domains. These clinical reasoning
experience, are then considered. The myriad possible
signposts, initially identified by our CPs, were subse-
options are thus narrowed down to several that are
quently acknowledged and validated within a large
relevant to the current case and fit well with the
array of medical specialties such as neurology,
clinical data. These are discussed with the patient,
psychiatry and internal medicine, and even dentistry.
leading to the selection of appropriate actions and
This indicates that there are common reasoning
producing ‘Investigations’ and ‘Therapeutic inter-
processes that clinicians across different disciplines
ventions’ that are then implemented.
routinely employ, and common steps that all seem to
be essential for success in problem solving during
The two other sub-models, which describe the
clinical encounters. We suspect that, depending on
processes ‘Identify early cues’ and ‘Regulate cognitive
the discipline, some processes may be more salient
processes’, are available online (Figs S1 and S2).

460 ª Blackwell Publishing Ltd 2012. MEDICAL EDUCATION 2012; 46: 454–463
 Clinical reasoning: graphical representation

but that the essence of the processes is similar. For reasoning processes even while they concentrate on
instance, categorisation ⁄ diagnosis is often central in specific aspects in isolation in dedicated teaching
medicine, whereas in dentistry a greater part of sessions.
reasoning may revolve around treatment planning.
Under the complex and ambiguous conditions that
Lessons drawn from the development of the MOT characterise daily practice, seasoned clinicians are
model attentive to a wide variety of dimensions of a clinical
encounter and reasoning cannot be successful if
The extraction and graphical representation of the some of these dimensions are neglected. A narrow
clinical reasoning processes of our participants raised view of the multidimensional nature of clinical
several important points. Firstly, it confirmed the reasoning may explain the paucity of methods avail-
dynamic nature of problem representation2,27 able to medical educators for fostering the acquisi-
throughout a clinical encounter. Secondly, CPs found tion of clinical reasoning skills, detecting learners
the processes they used when considering manage- with deficiencies in clinical reasoning, and offering
ment (i.e. investigation and treatment) options very appropriate remediation to those who have been
similar to those they employed during categorisation identified as learners in trouble.7–10 Although some
processes: both entailed a search for a fit between clinical reasoning difficulties may be linked to clearly
activated scripts and information derived during the delineated steps in the reasoning process (e.g. ‘data
clinical encounter. For instance, performance of a gathering’ or ‘hypothesis generation’), others may be
lumbar puncture (an investigative option) ‘fits’ with ascribed to less commonly identified factors, such as
the investigation script ‘measure the patient’s intra- ability to appreciate the clinical context, ability to
cranial pressure’ only if there is no clinical evidence deal with uncertainty, or ability to communicate
of elevated intracranial pressure (such as papilloe- effectively. The relationship between these factors
dema, which might be considered an unacceptable and clinical reasoning is rarely discussed in the
value for the ‘measure intracranial pressure’ script literature. Our model may be a useful tool for
prior to the acquisition of a computed tomography bringing these crucial yet often overlooked dimen-
scan of the head). Thirdly, the core reasoning sions of clinical reasoning to light for the benefit of
processes depicted on the horizontal axis of the main discussion and instruction among teachers and
screen of the model suggest that, from the moment a learners.
clinical encounter begins, a whole series of concom-
itant cognitive actions occur: the clinical reservoir of Written assessments of clinical reasoning typically
knowledge is tapped; scripts are mobilised and provide a brief clinical context and then ask ques-
enriched, and from time to time encapsulated bio- tions such as ‘What is the correct or most probable
psycho-social knowledge is accessed. All these pro- diagnosis?’ or ‘What is the most relevant investiga-
cesses work in parallel and are under the control of tion or treatment option?’ Such formats assess the
metacognition, indicating that clinical reasoning is outcome of clinical reasoning, rather than its pro-
not a linear process consisting of a succession of cesses. As long as the right answer is provided by the
steps, but, rather, a complex deployment of numer- learner, reasoning is considered acceptable and no
ous cognitive processes. regard is paid to how the answer is obtained. In such
instances, educators who assess clinical reasoning do
Potential educational applications of the MOT model not actually have access to learners’ reasoning
of clinical reasoning processes and are therefore unable to identify
whether important elements are lacking or are being
The identification and graphical representation of misused. Educators are therefore generally good at
the complexity of clinical reasoning has potentially detecting ‘those who do not have it’, but are rarely
important applications for use within undergraduate able to point out precisely where in a learner’s
and graduate medical curricula to inform teaching, reasoning processes about a particular case the
learning and assessment. The MOT model could be problem lies, and thus have difficulty designing
presented early in these curricula as an advance appropriate remediation activities.10 The MOT
organiser, thus providing students with an overview of model can be used as a guide to test specific parts of
experts’ clinical reasoning processes. It could be used learners’ clinical reasoning processes, and therefore
to support curricular development because it can to help pinpoint flaws in reasoning as learners
help to identify opportune moments for students to reason through clinical encounters. Remediation
learn specific elements of clinical reasoning, and help activities can then be designed to target these
students gain a global appreciation of clinical specific areas of difficulty.

ª Blackwell Publishing Ltd 2012. MEDICAL EDUCATION 2012; 46: 454–463 461
B Charlin et al

Study limitations always shows that knowledge is incorrectly interpreted

or that the model is incomplete. Within the sessions,
This study has several limitations. The 12 CVs were all resolving these issues often revealed tacit knowledge
doctors; health professionals from other health- that had not surfaced before.
related disciplines were therefore under-represented
in this particular cohort. Furthermore, the six
clinicians who participated in the iterative knowledge CONCLUSIONS
extraction and modelling sessions were all strongly
implicated in the teaching of clinical reasoning, were The MOT model constitutes an explicit graphical
all informed of the literature on clinical reasoning, representation of the multifaceted processes of clin-
and had more than basic knowledge of cognitive ical reasoning. Its validity across disciplines and
sciences. They were inevitably influenced by these situations needs to be confirmed, but it has the
concepts from the literature and perceived them as potential to inform the acquisition, teaching and
critical both for unravelling their own clinical assessment of clinical reasoning skills, to help clinical
reasoning processes and for facilitating clinical edu- educators identify flawed reasoning processes in their
cation. However, that the CPs held these characteris- students and residents and design appropriate reme-
tics can be viewed as a strength as well as a limitation of diation activities, and to orient future research studies
the study because the research method – participatory that aim to unravel more of the complexities of
action research – borrows techniques from qualitative clinical reasoning.
methodology, in which the co-construction process is
carried out collaboratively by a group of people who
are interested in changing practice in their setting. Contributors: BC co-led the conception and design of the
study and is a principal co-author of the manuscript. SL
It is interesting to consider whether reasoning participated in the validation session and is a principal
processes and their representations would have been co-author of the paper. BM conceived and organised the
validation session. FC co-led the conception and design of
different had the study involved other CPs and
the study. BC, BM, FC, M-CA, AC, NCF and CB served as
another cognitician. It is likely that the model would clinician-participants and were responsible for developing
differ in detail, as symbols (ovals, rectangles, arrows) the ideas and analysing the findings of the study. LH
can be arrayed in slightly different manners, with no described the technical aspects of MOT methodology and
ultimate consequence for conceptualisation. How- wrote a preliminary version of the paper. All authors
ever, the nature of the processes and the nature of contributed to the critical revision of the paper for
links are another matter. A first concern refers to intellectual content and approved the final manuscript for
whether all key processes have been unveiled. As submission.
knowledge extraction was undertaken in multiple Acknowledgements: Valérie Dory, Institute of Health and
sessions in an iterative way, we think that participants’ Society (IRSS), Université catholique de Louvain, Brussels,
conscious reasoning processes continued to reveal Belgium (VD) and Mark GoldszmidtDepartment of
themselves until saturation was reached; only findings Medicine, Schulich School of Medicine & Dentistry,
London, ON, Canada (MG), provided useful comments on
from research using similar or other techniques will
drafts of the paper.
confirm or refute this. A second concern refers to
Funding: none.
how realistically these cognitive operations are
Conflicts of interest: none.
depicted. In this respect the MOT software and
Ethical approval: the validation session described in this
technique offer significant advances over previous study was approved by the University of Montreal/Faculty of
tools used for the purpose of deriving cognitive maps. Medicine Institutional Review Board.

The rules inherent in the modelling technique in

MOT Plus often led to the detection of logical errors REFERENCES
during the construction of models. Links that did not
accurately express the experts’ knowledge indicated 1 Norman G. Research in clinical reasoning: past history
that the model probably contained a logical error. By and current trends. Med Educ 2005;39:418–27.
probing participants further it was possible to correct 2 Gruppen L, Frohna A. Clinical reasoning. In: Norman
the mistake and enhance the precision of the map. GR, van der Vleuten CP, Newble DI, eds. International
Additionally, the graphical layout of the model often Handbook of Research on Medical Education. Dordrecht:
revealed possible faulty interpretations. For example, Kluwer Academic 2002;205–30.
the grammar in MOT Plus very rarely allows links to 3 Schwartz A, Elstein AS. Clinical reasoning in medicine.
In: Higgs J, Jones M, Loftus S, Christensen NC, eds.
cross. When this situation arises, analysis almost

462 ª Blackwell Publishing Ltd 2012. MEDICAL EDUCATION 2012; 46: 454–463
 Clinical reasoning: graphical representation

Clinical Reasoning in the Health Professions, 3rd edn. 20 Schmidt H, Rickers R. How expertise develops in
Edinburgh: Elsevier 2008;223–34. medicine: knowledge encapsulation and illness script
4 Eva KW. What every teacher needs to know about formation. Med Educ 2007;41:1133–9.
clinical reasoning. Med Educ 2005;39:98–106. 21 Feltovich PJ, Barrows HS. Issues of generality in
5 Bowen J. Educational strategies to promote clinical medical problem solving. In: Schmidt HG, De Volder
diagnostic reasoning. N Engl J Med 2006;355:2217–25. ML, eds. Tutorials in Problem-Based Learning. Assen ⁄
6 Bargh JA, Chartrand TL. The unbearable automaticity Maastricht: Van Gorcum 1984;128–42.
of being. Am Psychol 1999;54:462–79. 22 Schmidt HG, Norman GR, Boshuizen HPA. A cognitive
7 Bordage G. Why did I miss the diagnosis? Some perspective on medical expertise: theory and implica-
cognitive explanations and educational implication. tions. Acad Med 1990;65:611–21.
Acad Med 1999;74 (Suppl):138–43. 23 Charlin B, Boshuizen HP, Custers EJ, Feltovich PJ. Scripts
8 Mamede S, Schmidt HG, Rikers R. Diagnostic errors and clinical reasoning. Med Educ 2007;41:1178–84.
and reflective practice in medicine. J Eval Clin Pract 24 Lurie S. Towards greater clarity in the role of ambiguity
2007;13:138–45. in clinical reasoning. Med Educ 2011;45:326–8.
9 Hauer K, Ciccone A, Henzel T, Katsufrakis P, Miller S, 25 Boshuizen HPA, Schmidt HG. On the role of bio-
Norcross W, Papadakis M, Irby D. Remediation of the medical knowledge in clinical reasoning by experts,
deficiencies of physicians across the continuum from intermediates and novices. Cogn Sci 1992;16:
medical school to practice: a thematic review of the 153–84.
literature. Acad Med 2009;84:1822–32. 26 Anderson J. Problem solving. In: Anderson J, ed.
10 Audétat MC, Dory V, Nendaz M, Vanpee D, Pestiaux D, Cognitive Psychology and its Implications, 3rd edn. New
Junod Perron D, Charlin B. What is so difficult about York, NY: W H Freeman 1990;219–55.
managing clinical reasoning difficulties? Med Educ 27 Nendaz M, Bordage G. Promoting diagnostic problem
2012;46:216–27. representation. Med Educ 2002;36:760–6.
11 Basque J, Pudelko B. Intersubjective meaning-making 28 Bordage G. Prototypes and semantic qualifiers: from
in dyads using object-typed concept mapping. In: past to present. Med Educ 2007;41:1117–21.
Torres PL, Marriott RCV, eds. Handbook of Research on 29 Flavell JH. Metacognition and cognitive monitoring, a
Collaborative Learning using Concept Mapping. New York, new area of cognitive–developmental inquiry. Am
NY: Hershey 2009;180–206. Psychol 1979;34:906–11.
12 Paquette G. Mode´lisation des Connaissances et des Compe´- 30 Elstein AS, Shulman LS, Sprafka SA. Medical Problem
tences. Un Langage Graphique pour Concevoir et Apprendre. Solving: An Analysis of Clinical Reasoning. Cambridge,
Sainte-Foy, QC: University of Quebec Press 2002. MA: Harvard University Press 1978;273–302.
13 Novak JD, Cañas AJ. The Theory underlying Concept Maps 31 Balla JI. The Diagnostic Process: A Model for Clinical
and How to Construct and Use Them. Technical Report Teachers. Cambridge: Cambridge University Press
IHMC CmapTools. Florida Institute for Human and 1985;49–106.
Machine Cognition 2008. http://cmap.ihmc.us/ 32 Barrows HS, Feltovich PJ. The clinical reasoning
publications/researchpapers/theorycmaps/theory process. Med Educ 1987;21:86–91.
underlyingconceptmaps.htm. [Accessed 27 February 33 Grant J, Marsden P. Primary knowledge, medical
2012]. education and consultant expertise. Med Educ 1988;22:
14 Daley BJ, Torre DM. Concept maps in medical educa- 173–9.
tion: an analytical literature review. Med Educ
2010;44:440–8.
15 Kemmis S, McTaggart R. Participatory action research: SUPPORTING INFORMATION
communicative action and the public sphere. In:
Denzin N, Lincoln Y, eds. Handbook of Qualitative Additional supporting information may be found in the
Research, 3rd edn. Beverley Hills, CA: Sage Publications online version of this article.
2005;271–330.
16 Schön D. From technical rationality to reflection in Figure S1. Process: ‘Identify early cues’.
action. In: Dowie J, Elstein A, eds. Professional Judgment:
a Reader in Clinical Decision Making (1988). Cambridge:
Figure S2. Process: ‘Regulate cognitive processes’.
Cambridge University Press 1983;60–77.
17 Gilhooly KS. Cognitive psychology and medical
Please note: Wiley-Blackwell is not responsible for the
diagnosis. Appl Cogn Psychol 1990;4:261–72.
content or functionality of any supporting materials sup-
18 Charlin B, Tardif J, Boshuizen HP. Scripts and medical
plied by the authors. Any queries (other than for missing
diagnostic knowledge: theory and applications for
material) should be directed to the corresponding author
clinical reasoning instruction and research. Acad Med
for the article.
2000;75 (2):182–90.
19 Custers EJ, Regehr G, Norman GR. Mental represen-
tations of medical diagnostic knowledge: a review. Acad Received 18 July 2011; editorial comments to authors 12 October
Med 1996;71 (Suppl):55–61. 2011; accepted for publication 1 November 2011

ª Blackwell Publishing Ltd 2012. MEDICAL EDUCATION 2012; 46: 454–463 463