How do doctors think?
- eg how do they come up with a diagnosis
Training: start with symptoms and history, results of initial assessment, list of possibilities, appropriate treatments
However, heuristics/pattern recognition is part of it
=> disparity between bet. training and their practice
=> 20 seconds to form an opinion
What can we learn from the case study of Anne Dodge?
Misdiagnosis: bulimia
Actual diagnosis: celiac disease
- availability heuristic (expectations): the doctors thought that a young woman’s medical history like hers would fit the diagnosis of bulimia
- the initial diagnosis and psychiatric patient referral are anchors
=> must look beyond the previous diagnoses and look at it from a fresh perspective
Errors in misdiagnosis
(cognitive biases account for 80% in medical errors that are documented)
Availability heuristic: how you get started, what is the available info
Anchoring: starting point
Confirmation bias: what info/evidence to look for
What were the errors in the diagnosis of Blanche Begaye?
- the doctor missed the point/possibility that it could have been aspirin toxicity
- initial bias was availability: the doc. thought of the idea of pneumonia because of the many cases (recent exposure effect)
- Confirmation bias: persistence in the misdiagnosis as some data consistent with it, while some were not (discounted this)
What were the errors in the diagnosis of Maxine Carlson?
- only diagnosed one of out the two
- Availability: IBS served as the available explanation after 1 year when new symptoms occurred
- Anchoring: did not consider other options as the evaluation and interpretation of results were focused on the IBS diagnosis
- Attribution error: problems attributed to patient (hypochondriac - anxious abt health) rather than physical problems
Heuristics and biases in medicine
- doctors must constantly evaluate evidence and make decisions under uncertainty
- generally, they rely on heuristics and can be effective but susceptible to biases as well => this is a natural tendency
(med pros cons)
Is more data a solution to this issue?
Radiology
radiologist: specializes in interpreting x-rays or MRIs, 12000-15000 images / year
- expected to quickly evaluate from the first impression (get the gist/”gestalt”)
- data is not clear cut, opportunity for cognitive biases
- can’t fix the problem
Study with missing clavicle that 60% of radiologists miss
=> have to know what you are looking for to some extent
=> attention (inattentional blindness) and expectations matter to detect (especially the absence of something)
It’s easy to miss things when u don’t hv an idea of what to attend to
Radiology: is more evaluation worse for the diagnosis?
- yes, they start to seethings that are not there!
- increase chance of hurting the patient
=> perceptual “filling in”
Issue of false positives
(med pros cons)
- sets u on path for increasingly invasive procedures
- eg for mammograms, the current recommendation is for those over 50 years old => the group most at risk
=> avoid false positives!
Improving medical decisions
- strategies and approaches by doctors
Strategy: blank state
=> Anne Dodge case study: not necessarily always easy to do, extra effort, mbe work against your instincts
Strategy: highlight structured checklist for evaluation, be thorough and don’t skip out on things - avoid “search satisfaction” and more than one problem can occur
Strategy: ABC checklist in the ER
- counters against preconceptions that cause the doctor to miss something important
Strategy: short list of alternatives - avoid making mistakes and anchor to the first possibility of what you “think you know”
Improving medical decisions
-questions that patients can ask
- kinda like a partnership
Ask questions like:
- what else could it be? especially if the condition persists
- any test or x-ray that contradicts the current prediction? (note and think of any discrepancies)
- what is the worst that it could be? other body parts near the symptoms? (avoid anchoring)
Improving medical decisions
- clinical decision support systems (CDS)
Software database and analysis
- based on statistical analysis of large samples
- database includes full range of possible diagnoses consistent with the data
+ allows quick access to summaries of relevant clinical studies
+ less likely to base judgments on selective subset of evidence
+ counteracts confirmation bais
Improving medical decisions
- training better intuitive judgment
ICU admission
- found that doctors did not give enough weighing to the danger admitting to the ICU (not a good judgment)
- did not reliably differentiate bet. patients that did and did not need the ICU
Physocal measures eg EKG are most informative for evaluation, but doctors overweigh the longterm risk factors
=> heart disease predictive instrument (statistical instrument)
=> judgments improve
=> learned to pay attention to the right factos, the instrument helped them develop pattern recognition
