Chris Feddock, MD, MS Andrew Hoellein, MD, MS Clinical Reasoning Chris Feddock, MD, MS Andrew Hoellein, MD, MS
Learning Objectives Describe the cognitive processes of expert diagnosticians Recognize the shortcomings of various diagnostic strategies Foster the development of clinical reasoning in bedside teaching
Systems System 1 Intuitive & Heuristic System 2 Analytic & Systematic Fast Slow Little conscious control High cognitive awareness Low reliability High reliability Errors are relatively common Rare errors (but significant ones) High emotional attachment Low emotional attachment Low scientific rigor High scientific rigor Am J Med. 2008;121: S24–S29.
Intuitive System
Pattern Recognition Representativeness Heuristic Recognition of the prototypical pattern of disease “If it looks like a duck, walks like a duck, quacks like a duck, then it is a duck.” Med Educ. 2004;39:98-106.
Shortfalls of Inductive Reasoning
Poggendorff Illusion, 1860.
Previously Encountered 16 clinical cases reviewed and discussed 4 cases similar to those encountered previously Provide a diagnosis 4 cases distinct from those encountered previously First and Second Year Internal Medicine Residents 0 = incorrect diagnosis 0.5 = partially correct 1 = fully correct Which residents had greater diagnostic accuracy? Does it matter if they have seen the case previously? Previously Encountered Novel First Year Residents 2.03 (1.55-2.51) 1.42 (0.92-1.92) Second Year Residents 1.55 (1.15-1.96) 2.19 (1.73-2.66)
Analytical System
Hypothetico-Deductive Process Accept Hypothesis Data Collection New Hypothesis Prediction Adequacy Coherency Parsimony Reject Hypothesis Med Educ. 2004;39:98-106.
Shortfalls of Analytical Reasoning
The “Door” Study http://www.youtube.com/watch?v=FWSxSQsspiQ Psychonomic Bulletin & Review. 1998;5:644-9.
What diagnostic strategy is best?
First and Second Year Internal Medicine Residents 16 clinical cases reviewed and discussed 4 cases similar to those encountered previously Quick diagnosis Analytic Diagnosis Read the case Write down the diagnosis previously given for the case List the findings in the case description that support this diagnosis List the findings that speak against this diagnosis List the findings that would be expected to be present if the diagnosis were true but were not described in the case Nonanalytical Analytical First Year Residents 2.03 (1.55-2.51) 2.31 (1.89-2.73) Second Year Residents 1.55 (1.15-1.96) 2.03 (1.73-2.66)
No single diagnostic strategy Diagnostic error occurs with both strategies Heuristic error Cognitive overload – can only process small amounts Experts move between analytical and intuitive Efficient tailoring of key information Synthesize data to reach a diagnosis (forward reasoning) Knowledge organized into schemes
Teach Clinical Reasoning Make learner’s commit to a diagnosis or a narrow differential Must be actively engaged in decision-making Emphasize pattern recognition Teach context specificity Continual development of more sophisticated illness scripts Think about your thinking (out loud!) Self-reflect on diagnostic decision-making Admit to clinical reasoning failures
Learner #1 “I think it is most likely GERD vs. anxiety.” 65 year-old woman Lost mother recently “chest burning” began today Radiates to right arm Nausea/vomiting Did not respond to anti-acids History of diabetes “I think it is most likely GERD vs. anxiety.”
Learner #1 – insufficient knowledge Metacognition – dual process thinking Problem representation Illness scripts Likelihood ratios
Learner #1 65 year-old woman with diabetes (45% chance of MI) and chest discomfort 2 hours of “chest burning” (LR = 1.3) Radiates to right arm (LR = 7.3) Nausea/vomiting (LR = 3.5)
Learner #2 54 year-old woman “Boring” abdominal pain for 1 day Nausea Drinks 3-4 beers/night Non-smoker Abdomen with voluntary guarding “I think it is a AAA because my uncle had one and he presented like this.”
Learner #2 - biased Look for bias/heuristics - shortcuts What else could this be? Why would a non-smoking 54 y/o woman have a AAA?
Bias-prone Heuristics Definition Anchoring Focus on features too early without adjusting with more data Emotional Feelings influence diagnosis (non-adherent patient) Availability Recently encountered Confirmation Look for confirming evidence only Blind obedience Undue deference to authority/technology Framing Source of info and how it is framed Unpacking Failure to elicit all pertinent info Diagnostic momentum Failure to consider diagnostic eval due to prior diagnostic label Academic Internal Medicine Insight. 2015
Learner #2 Pancreatitis Cholecystitis Appendicitis AAA Older women < 2% Smoking Family history Absence of mass/bruit (LR- .43)
Learner #3 “This must be COPD because he is a smoker.” 40 y/o man Progressive shortness of air for 3 days 20 pack year smoking history Non-productive cough Modest dyspnea on exertion Temp 100.5 “This must be COPD because he is a smoker.”
Learner #3 - overconfident State confidence about diagnosis Describe biases What else could this be? Do all smokers develop COPD?
Learner #3 Anchoring bias Availability bias Confirmation bias Smoking < 30 years + no wheezing symptoms + no wheezes on exam -posterior odds of obstructive airways = 0.02%
Learner 4 18 year-old woman hurt left knee playing lacrosse “popping” sensation when planted left foot Difficulty bearing weight Small effusion present Lachman positive McMurray indeterminate “This could be a ACL tear but still could be a meniscal injury or maybe even torn patellar tendon… I don’t know!”
Learner #4 - underconfident Diagnostic “time out” Review history and physical
Learner #4 Classic injury pattern for ACL Women RR > 2 Lachman LR+ 42
Thank you Chris Feddock Andrew Hoellein cafedd00@uky.edu arhoel0@uky.edu