Visual diagnostic reasoning: from theory to practice

Dr. Amy Warren BSc., BVSc. (hons), PhD, DACVP Associate Professor Pathology

Faculty of Veterinary Medicine, University of Calgary, Canada

OVERVIEW 1. What is clinical reasoning? 2. Theories of clinical reasoning in

experts 3. Visual diagnostic reasoning

“Clinical reasoning is as difficult to define as it is to teach”

(Linn et al, 2008, Teaching in General Practice 41:1)

• 15 year old quarter horse mare with a history of recurrent laminitis

• Polyuria and polydipsia • Polyphagic • Hyperhidrosis

Laboratory data: • Fasting hyperglycemia and

glycosuria • Mild anemia • Lymphopenia • Elevated liver enzymes • Elevated serum ACTH

wikimedia

What is clinical reasoning?

The sum of thinking and decision-making processes associated with clinical practice.

(Higgs, 2008)

Clinical reasoning

Cognition (problem solving)

Context

Knowledge

Metacognition

Tolerance to ambiguity

High stakes

Reflexive understanding

ebollerSticky NoteUNIQUE TO VET MED, ESP WITH FINANCIAL CONSTRAINTS THAT LEAVE HIOLES IN OUR INFORMATION SET

The nature of clinical reasoning

• Problems are ill-structured and evidence is ambiguous – Decision making needs to occur based on

dynamic and ill-defined information – Shifting and competing goals and

outcomes • Occurs on action-feedback loops • Decisions have elements of time-

pressure, personal stress and highly significant outcomes for patients

Yet, it is one of the most defining features of expertise and practice excellence.

Research and understanding clinical reasoning

• Complex mostly invisible process • Automatic and highly attenuated • Highly individualized • Requires a range of capabilities:

– Cognition, – Metacognition, – Emotion, – Reflexive and – Social

ebollerSticky Notehighly exp clinician that recognizes dz X coming from 'gut' but how tdo you explain what went into that thinking/reasoning to a student?

ebollerSticky Noteeg forward, backward, hypothetical deductive, etc

Signalment and history

Clinical exam findings

Bowen, J. L. (2006) Educational strategies to promote clinical diagnostic reasoning. NEJM. 355:3317-25.

CBC/ Chemistry

Specialized testing (ACTH/ cortisol levels)

Diagnostic Reasoning Theories (as a cognitive process)

• Analytical reasoning – Hypothetico-deductive – Forward and backward reasoning – Bayes Theorem

• Non-analytical reasoning – Pattern recognition – Knowledge structures – Script theory

• Dual processing theory

ebollerSticky Noteimp in dx testing - using probabilities to test your dx

Analytical reasoning • Hypothetico-deductive reasoning • Forward and backward reasoning

strategies • Bayes’ theorem

– Pre- and post-test probabilities and likelihood ratios

– Errors in frequency estimations: • Overestimation of vivid or easily

recalled events/ Underestimation of ordinary events

• Overestimation of frequency of events that fit a representative case

• Probabilities at the extreme (close to 0 and 1) tend to be over and underestimated

Bowen, J. L. (2006) Educational strategies to promote clinical diagnostic reasoning. NEJM. 355:3317-25.

Diagnostic Reasoning Theories (as a cognitive process)

• Analytical reasoning – Hypothetico-deductive – Forward and backward reasoning – Bayes Theorem

• Non-analytical reasoning – Pattern recognition – Knowledge structures – Script theory

• Dual processing theory

Non-analytical reasoning

• Pattern recognition (Patel, Norman, Eva, Groen, Schmidt)

– Rapid, automatic and non-verbal – Experience – Automated retrieval knowledge

• Knowledge structures • Script theory

ebollerSticky Notein the novice can be dangerous , eg students desperately wnat to pattern revognize, eg when you teach a dx, they think everything has that

ebollerSticky Noteusin gcase exemplars assoicated with a knowledge structure

Knowledge structures

Circulating immune complexes deposit in basement membrane

Complement, humoral and cell-mediated

immune cell damage to glomerulus

Increased leakiness of glomerulus

approximately same MW as albumin

PROTEINURIA

Type III hypersensitivity

Glomerulonephritis

Antibody-antigen

structure and production

Glomerular microanatomy and renal physiology

http://upload.wikimedia.org/wikipedia/commons/d/d8/Dalmatian_black_front.jpgebollerSticky Noteas time goes by, the knowledge drops off and the dz then gets represented by the visual/clinical entity and the things on teh following slide

Knowledge structures

• Causes (Lyme, Babesia, RMSF, Ehrlichia, hepatozoon, Leishmania, CAV-1, Leptospirosis, heart worm, trypanosomiasis, schistosomiasis, GI disease, neoplasia, SLE, hypertension, hyperadrenocortism, juvenile renal dysplasia, congenital glomerulopathy)

• Relative prevalence and likelihood of diagnoses (dependent on animal’s age, breed, location, history, etc.)

• Diagnostic tests that differentiate DDx (expected results, normal reference intervals, acceptable ranges for DDx)

• Possible treatments • Prognosis

Dotty the Dalmatian with proteinuria

Glomerulonephritis

http://upload.wikimedia.org/wikipedia/commons/d/d8/Dalmatian_black_front.jpgebollerSticky Noteyes exp'd clinicians may foget the basics but they become very impo when we get tough cases

Encapsulation theory Co

ncep

tual

Adapted from Schmidt (1990) Acad Med 65(10):611-621

Stage 1: Development of elaborated causal networks Relatedness of concepts Cause and consequence of pathophysiologic processes

Stage 2: Compilation of elaborated networks into abridged ones

Extensive and repeated application High level simplified causal models using diagnosis to explain signs and symptoms

Stage 3: Emergence of Illness scripts Stage 4: Storing patient encounters as Instance scripts Ex

perie

ntia

l

ebollerSticky Noteas they go into clinical as students

ebollerSticky Noteas they go into careers, building experience, collecting cases

Script theory (Schmidt 1990, Charlin 2000, Charlin 2007)

• Illness scripts that are activated by pattern recognition

• Cognitive structures present in experienced clinicians – Constructed during repeated experience – More efficient task performance

• Hierarchical with “slots” pertaining to: – Predisposing conditions – Clinical signs – Pathophysiology – Diagnostic testing – Acceptable and not acceptable values – Treatment options – Prognosis

ebollerSticky Noteyou see this then think about what should be there if it really is PPID

Analytical reasoning • Hypothetico-deductive • Bayes’ theorem • Forward reasoning • Backward reasoning

Non-analytical reasoning • Pattern recognition • Automated retrieval of

knowledge structures • Script theory

Diagnostic Reasoning Theories (as a cognitive process)

• Analytical reasoning – Hypothetico-deductive – Forward and backward reasoning – Bayes Theorem

• Non-analytical reasoning – Pattern recognition – Knowledge structures – Script theory

• Dual processing theory

ebollerSticky Notefor years CR was pigeon holed ionto one or the other, eg novice does AR and expert does NAR

Dual process theory of human judgment (Kahneman 2003, Evans 2008)

Adapted from Evans et al. (2008) Annu Rev Psychol; 59:255-278.

System 1 and System 2 reasoning processes in human decision-making

Non-analytical Analytical

Dual process theory in clinical reasoning

• System 1 default overridden by system 2: – Strong deductive reasoning instructions, dissonant patterns,

high cognitive ability • System 1 highly related to experience

– Repetitive system 2 processing can lead to a system 1 response • System 2 impaired by stresses on working memory

– Includes lack of sleep, competing thought processes, emotion, time pressure

• Clinicians use a mixture of intuition and analysis in decision making

• No “better” system- both are necessary and both are involved in diagnostic error

Our understanding of clinical reasoning so far…..

• System 1 and system 2 dual reasoning theories likely accounts for much of the cognitive processes of clinical reasoning – Experts with familiar cases tend to use system 1 reasoning – Novices tend to use system 2 reasoning – Experts are triggered to use system 2 reasoning in unfamiliar cases,

cases with dissonant data or as a “checks and balance” to their reasoning

– Not likely a dichotomy but a continuum with clinicians switching back and forth between system 1 and system 2 reasoning

• Experts possess an extensive and multidimensional knowledge base = SCRIPTS – Initial exposure of clinicians to a case typically stimulates initial

System 1 retrieval of multiple illness scripts

ebollerSticky Noteeg if it is a legal case

Veterinary medicine is a visual science Visual reasoning

FNA Subcutaneous mass, 6 year old, FS, Rottweiler

Bowen, J. L. (2006) Educational strategies to promote clinical diagnostic reasoning. NEJM. 355:3317-25.

Verbal data (history) Clinical findings (PE) Additional exams or diagnostic tests

Visual Reasoning

When do we use visual reasoning skills?

Pathology Dermatology Diagnostic imaging

Visual Reasoning

Sensation and perception

Comparison of image to sensory, short and long term memory

Visual Reasoning

ebollerSticky Notethis process is much more Type 1 CR

Sensation and perception Entire scene in

peripheral vision interesting

features pop out

Scene disengaged, foveal attention

on area of interest “Fixation”

Eyes reposition to new area of

interest “Saccade”

Visual Reasoning

ebollerSticky Noteeyes dart around the image in a particular pattern as described here. Things taht we 'need' to the see to make a diagnosis

Eye-tracking technology • Eye position represents

visual attention • Fixations and saccades of

eye-movement

Visual Reasoning

ebollerSticky Notered - fixationgrey - saccades

Visual reasoning hypothesis

• Experts develop sophisticated illness scripts triggered by visual images

• Using eye-tracking we expect that experts have more efficient and targeted viewing of a slide, akin to system 1 thinking

• Experts differ from novices by: – Reduced total time which slides are viewed – Select more diagnostically relevant fixation points – Spend more time on diagnostically relevant fixation points

Novice-Expert study

Novice-expert study

Equipment set up

Round cell images 1-5 with AOIs

Dependent variables: 1. Diagnostic accuracy (dichotomous) 2. Percentage time spent viewing the

area of interest 3. Time to diagnosis 4. Eye-movement pattern 5. Talk aloud protocol

Novice-Expert study

ebollerSticky Noteto switch them into system 2 and explain why they came up with the dx

Warren et al (2016) JVME In Press

What differences do we see between novice and experts? 1. Experts have a much shorter time

to diagnosis

2. Experts spend more time visualizing key features and use more individual key features to make a diagnosis than novices

3. Experts have efficient eye-pattern movements

4. Experts use higher order “knowledge collation” terms

5. Experts had a higher level of diagnostic accuracy

ebollerSticky Noteexperts left out 'the noise' eg blood/debris in background and were able to pick ou thte key features that are needed for the dx

What can we conclude?

Experts employ pattern-recognition (System 1) and – System 1 reasoning fast, intuitive (quicker to diagnosis) – More efficient eye-movements

Experts use script inductive reasoning with analytic (System 2) justification

– More diagnostic features identified suggests script induction – Suggest experts develop a sophisticated set of “visual illness

scripts”

Novice-Expert study

ebollerSticky Noteand non-replicable - highly individualised

ebollerSticky Notechecks and balances in system 2 - eg 30% in system 1 then 70% system 2 to make sure they are correct

Multimodal assessment of visual diagnostic reasoning

Where to next?

• In collaboration with Dr. Kent Hecker • fMRI showed different levels of activation within novices

between hard and easy clinical reasoning scenarios • System 2 reasoning is localized to the prefrontal cortex

neural areas of activation in novice and expert clinicians during clinical reasoning tasks

ebollerSticky Noteexperts use in system 2 in hard cases as do students al lthe time. in easier cases they use thalamic, lower level thinking mechanisms

This is all great….. but what does all this mean in teaching and learning?

Stay tuned for tomorrow!

Acknowledgments Collaborators: • Dr. Kent Hecker • Dr. Tyronne Donnon • Dr. Catherine Wagg • Dr. Nicole Fernandez • Dr. Heather Priest • Research technician- Jason Abboud • DVM class of 2014 and 2015

Funding • Zoetis • UCVM veterinary educational

research fund • UCVM clinical research fund

Visual diagnostic reasoning: from theory to practiceSlide Number 2Slide Number 3What is clinical reasoning?The nature of clinical reasoningResearch and understanding clinical reasoningSlide Number 7Diagnostic Reasoning Theories�(as a cognitive process)Analytical reasoningDiagnostic Reasoning Theories�(as a cognitive process)Non-analytical reasoningKnowledge structuresKnowledge structuresEncapsulation theoryScript theory �(Schmidt 1990, Charlin 2000, Charlin 2007)Slide Number 16Diagnostic Reasoning Theories�(as a cognitive process)Dual process theory of human judgment�(Kahneman 2003, Evans 2008)Dual process theory in clinical reasoningOur understanding of clinical reasoning so far…..Veterinary medicine is a visual scienceSlide Number 22Slide Number 23When do we use visual reasoning skills?Sensation and perceptionSensation and perceptionEye-tracking technologyVisual reasoning hypothesisNovice-expert studySlide Number 30What can we conclude?Multimodal assessment of visual diagnostic reasoningThis is all great….. but what does all this mean in teaching and learning?Acknowledgments