The NEW ENGLA ND JOURNAL of MEDICINE

Perspective june 27, 2013

From Mindless to Mindful Practice — Cognitive Bias

and Clinical Decision Making
Pat Croskerry, M.D., Ph.D.

T he two major products of clinical decision

making are diagnoses and treatment plans. If
the first is correct, the second has a greater chance
and clinicians are not immune
to the problem (see box). More
than 100 biases affecting clinical
decision making have been de-
of being correct too. Surprisingly, we don’t make scribed, and many medical disci-
plines now acknowledge their per-
correct diagnoses as often as we but common illnesses are com- vasive influence on our thinking.
think: the diagnostic failure rate monly misdiagnosed. For example, Cognitive failures are best un-
is estimated to be 10 to 15%. physicians know the pathophysi- derstood in the context of how
The rate is highest among spe- ology of pulmonary embolus in our brains manage and process
cialties in which patients are di- excruciating detail, yet because its information. The two principal
agnostically undifferentiated, such signs and symptoms are notori- modes, automatic and controlled,
as emergency medicine, family ously variable and overlap with are colloquially referred to as
medicine, and internal medicine. those of numerous other diseas- “intuitive” and “analytic”; psy-
Error in the visual specialties, es, this important diagnosis was chologists know them as Type 1
such as radiology and pathology, missed a staggering 55% of the and Type 2 processes. Various con-
is considerably lower, probably time in a series of fatal cases.2 ceptualizations of the reasoning
around 2%.1 Over the past 40 years, work process have been proposed, but
Diagnostic error has multiple by cognitive psychologists and most can be incorporated into
causes, but principal among them others has pointed to the human this dual-process system. This
are cognitive errors. Usually, it’s mind’s vulnerability to cognitive system is more than a model: it
not a lack of knowledge that biases, logical fallacies, false as- is accepted that the two processes
leads to failure, but problems with sumptions, and other reasoning involve different cortical mecha-
the clinician’s thinking. Esoteric failures. It seems that much of nisms with associated neuro-
diagnoses are occasionally missed, our everyday thinking is flawed, physiologic and neuroanatomical

n engl j med 368;26 nejm.org june 27, 2013 2445

The New England Journal of Medicine
Downloaded from nejm.org at UNIV OF SO DAKOTA on July 1, 2013. For personal use only. No other uses without permission.
Copyright © 2013 Massachusetts Medical Society. All rights reserved.
PERSPE C T I V E From Mindless to Mindful Practice

substrates. Functional magnetic

Clinical Examples of Cognitive Failure
resonance imaging scans vividly
Case 1 reveal the changes in neuronal ac-
A 21-year-old man is brought to a trauma center by ambulance. He has been stabbed tivity patterns as processes move
multiple times in the arms, chest, and head. He is in no significant distress. He is from one system to the other dur-
inebriated but cooperative. He has no dyspnea or shortness of breath; air entry is ing learning. Although the two
equal in both lungs; oxygen saturation, blood pressure, and pulse are all within
processes are often construed as
normal limits.
two different ways of reasoning,
The chest laceration over his left scapula is deep but on exploration does not
in fact very little (if any) reason-
appear to penetrate the chest cavity. Nevertheless, there is concern that the chest
cavity and major vessels may have been penetrated. Ultrasonography shows no free ing occurs in Type 1 processing
fluid in the chest; a chest film appears normal, with no pneumothorax; and an ab- — it is largely reflexive and au-
dominal series is normal, with no free air. There is considerable discussion between tonomous. The Augenblick diagno-
the resident and the attending physician regarding the management of posterior sis, made in the blink of an eye,
chest stab wounds, but eventually agreement is reached that computed tomogra-
is an impressive piece of medi-
phy (CT) of the chest is not indicated. The remaining lacerations are cleaned and
sutured, and the patient is discharged home in the company of his friend. cal showmanship and the stuff
Five days later, he presents to a different hospital reporting vomiting, blurred
of television entertainment (and
vision, and difficulty concentrating. A CT of his head reveals the track of a knife corridor consultations), but in real
wound penetrating the skull and several inches into the brain. clinical life it is fraught with
Comment: The cognitive failures identified here are “anchoring” and “search satis- danger.
ficing.” The resident and attending staff both anchored onto the chest wound as Descriptions of the operating
the most significant injury. When they satisfied themselves that the chest wound characteristics of the dual pro-
was stable, the resident failed to conduct a sufficient search to rule out other sig- cessing system in clinical reason-
nificant injuries.
ing provide a useful starting point
Case 2 for learning about medical deci-
An 18-year-old woman is referred by her family doctor to a psychiatric service for sion making.3 Intuitive processes
symptoms of severe anxiety and depression. She has been having frequent episod- are generally either hard-wired or
ic dyspnea, associated with hyperventilation, carpopedal spasm, and loss of con- acquired through repeated experi-
sciousness. The admitting psychiatrist wants to exclude the possibility of a respira-
tory problem and sends the patient to the emergency department (ED) with a re-
ence. They are subconscious and
quest for a chest film to rule out pneumonia. fast and mostly serve us well,
She is seen and assessed by an ED resident. The patient was not noted to be in
enabling us to conduct much of
any significant distress other than feeling breathless. She is obese, has a history of our daily business in all fields
asthma, and smokes cigarettes. She is currently being treated with a benzodiaze- of human activity. We mostly get
pine and anxiolytics and is taking a birth-control pill. Her chest and cardiovascular through life by moving from one
examination are normal. The resident orders routine blood work and a chest film. of the intuitive mode’s associa-
He reviews the film, reads it as normal, and believes the patient can be safely re-
turned to the psychiatric facility. He attributes her respiratory problems to anxiety.
tions to the next in a succession
of largely mindless, fixed-action
While she awaits transfer, she becomes very agitated and short of breath.
Several nurses attempt to settle her, encouraging her to breathe into a paper bag.
patterns. These patterns are in-
Shortly afterward, she loses consciousness. Her monitor shows pulseless electrical dispensable; however, they are
activity and then asystole. She cannot be resuscitated. At autopsy, she is found to also the primary source of cog-
have pelvic vein thrombosis extending from the femoral vein and saddle emboli in nitive failure. Most biases, falla-
both lungs, as well as multiple clots of varying age. cies, and thinking failures arise
Comment: Several cognitive failures probably influenced the outcome in this case. from the intuitive mode (see box).
The patient’s diagnosis of anxiety established “momentum” from her family doctor
When primary care physicians
through to the ED, and although she might well have had hyperventilation due to
anxiety, other possibilities were not ruled out earlier on in her care. Furthermore, trust their intuition that a pa-
bias regarding her psychiatric diagnosis probably influenced her care providers; psy- tient’s chest pain does not have
chiatric patients are more vulnerable to adverse events. “Framing” may also have a cardiac origin, they will usu-
been a problem, since the psychiatrist had specifically asked the ED to rule out an ally be correct — but not always.
­infective process and had not raised the possibility of pulmonary embolus, despite
The clinical gamble of trusting
the patient’s multiple risk factors. “Search satisficing” is again a problem, in that
the resident called off the search for a cause for the patient’s dyspnea after ruling one’s intuitions generally carries
out pneumonia. good odds, but inevitably those
intuitions will fail some patients.

2446 n engl j med 368;26 nejm.org june 27, 2013

The New England Journal of Medicine

Downloaded from nejm.org at UNIV OF SO DAKOTA on July 1, 2013. For personal use only. No other uses without permission.
Copyright © 2013 Massachusetts Medical Society. All rights reserved.
PERSPECTIVE From Mindless to Mindful Practice

The issue is whether we can tol- brace any work that helps us fined as the “rules, knowledge,
erate the current levels of failure think about our thinking (meta- procedures, and strategies that a
— or is there room for improve- cognition) and that it would be person can retrieve from memo-
ment? beneficial both to include basic ry in order to aid decision mak-
Analytic processes, by con- psychology courses in the medi- ing and problem solving.”4 It in-
trast, are conscious, deliberate, cal school curriculum and to ex- cludes knowledge about the
slower, and generally reliable. They pand medicine’s lexicon to in- properties of the particular bias
follow the laws of science and log- corporate terms from cognitive and what strategies might elimi-
ic and therefore are more likely to psychology. nate or reduce it. This process
be rational. Despite the ubiquity If cognitive biases are so appears to be uncommonly dif-
and usefulness of intuitions, they abundant and troublesome in ficult, although there have been
are not reliable enough for us to clinical decision making, why not some successes. A variety of
use them to send a spaceship to simply identify them and use a debiasing strategies have been
Mars. By contrast, when a patient
undergoes analytic assessment
for chest pain in a cardiac clinic Becoming alert to the influence
that culminates in angiography,
the conclusion is invariably cor- of bias requires maintaining keen vigilance
rect. Analytic failures can occur, and mindfulness of one’s own thinking.
but usually when the wrong
rules are followed or other fac- When a bias is identified by a decision maker,
tors come into play, such as
cognitive overload, fatigue, sleep a deliberate decoupling from the intuitive mode
deprivation, or emotional per- is required so that corrective “mindware”
turbations. The biggest down-
side of analytic reasoning is that can be engaged from the analytic mode.
it’s resource-intensive. Although
analytic reasoning can often be
done quickly and effectively, in “debiasing” strategy to avoid proposed, and they lead to a few
most fields of medicine, it would them? Unfortunately, that’s not important conclusions: debias-
be impractical to deal with each as easy as it sounds. First, many ing is not easy, no one strategy
clinical decision analytically. decision makers are unaware of will work for all biases, some
Given the substantial impact their biases, in part because our customization of strategies will
of our evolving understanding psychological defense mechanisms be necessary, and debiasing will
of cognition over the past few prevent us from examining our probably require multiple inter-
decades, it is somewhat surpris- thinking, motivation, and desires ventions and lifelong mainte-
ing that these major social sci- too closely. Second, many clini- nance.
ence findings have not readily cians are unaware of, or simply Cognitive failures like those
made their way into medicine. don’t appreciate the effect of, described in the box can be ad-
Although our awareness of re- such influences on their decision dressed by educational strategies
search biases led to the develop- making. that embrace critical thinking —
ment of the randomized, pro- Becoming alert to the influ- the “ability to engage in purpose-
spective, double-blind clinical ence of bias requires maintaining ful, self-regulatory judgement.”5
trial, we remain unrealistic keen vigilance and mindfulness Regulating judgment requires
about the scale of everyday cog- of one’s own thinking. When a training that can permit judi-
nitive and affective biases and bias is identified by a decision cious interventions by the ana-
their effect on clinical reason- maker, a deliberate decoupling lytic mode when needed — spe-
ing. Cognitive psychology has from the intuitive mode is re- cifically, in its capacity to override
not historically been considered quired so that corrective “mind- the intuitive mode. This critical
within the remit of medicine, ware” can be engaged from the step has been referred to as de-
but I believe that we should em- analytic mode. “Mindware” is de- coupling, metacognition, mind-

n engl j med 368;26 nejm.org june 27, 2013 2447

The New England Journal of Medicine
Downloaded from nejm.org at UNIV OF SO DAKOTA on July 1, 2013. For personal use only. No other uses without permission.
Copyright © 2013 Massachusetts Medical Society. All rights reserved.
PERSPE C T I V E From Mindless to Mindful Practice

fulness, and self-reflection. Most element of training in critical From the Division of Medical Education,
Dalhousie University, Halifax, NS, Canada.
of us never reach our ceilings thinking should be a review of
for critical thinking, and many the major cognitive and affec- 1. Berner ES, Graber ML. Overconfidence as
people go through life unaware tive biases and the ways they af- a cause of diagnostic error in medicine. Am J
of their thinking limitations. We fect thinking. Greater effort is Med 2008;121:Suppl:S2-S23.
2. Pineda LA, Hathwar VS, Grand BJ. Clinical
are not born critical thinkers. needed to develop effective cog- suspicion of fatal pulmonary embolism.
Like any other skill, however, nitive debiasing strategies in Chest 2001;120:791-5.
critical thinking can be taught medicine. All clinicians should 3. Croskerry P. A universal model for di­
agnostic reasoning. Acad Med 2009;84:
and cultivated, but even accom- develop the habit of conducting 1022-8.
plished critical thinkers remain regular and frequent surveil- 4. Stanovich KE. What intelligence tests
vulnerable to occasional undis- lance of their intuitive behavior. miss: the psychology of rational thought.
New Haven, CT: Yale University Press, 2010.
ciplined and irrational thought. To paraphrase Socrates, the un- 5. Abrami PC, Bernard RM, Borokhovski E,
I believe that medical educa- examined thought is not worth et al. Instructional interventions affecting criti-
tors should promote critical thinking. cal thinking skills and dispositions: a stage 1
meta-analysis. Rev Educ Res 2008;78:1102-
thinking throughout undergrad- 34.
Disclosure forms provided by the author
uate, postgraduate, and continu- are available with the full text of this arti- DOI: 10.1056/NEJMp1303712
ing medical education. One key cle at NEJM.org. Copyright © 2013 Massachusetts Medical Society.

Uncertainty — The Other Side of Prognosis

Alexander K. Smith, M.D., M.P.H., Douglas B. White, M.D., and Robert M. Arnold, M.D.

R ecently, there has been a re-

surgence of interest in prog-
nosis. This interest has been
tions in sample sizes. Second,
most prognostic indexes have not
been tested in heterogeneous
colleague Dr. Faith Fitzgerald as
the “punctilious quantification
of the amorphous.” In other
driven by a recognition that clinical settings.3 Third, in clin- words, no matter what we do,
prognosis plays a central role in ical practice, clinicians must ex- there will always be some un-
medical decision making, from trapolate from population-level certainty in prognosis.
counseling outpatients about stop- estimates to make judgments This uncertainty is difficult
ping cancer screening to making with or for individual patients. for patients and their families to
decisions with patients’ surrogates Even if a risk estimate is very deal with. For patients, not
about withdrawal of life support precise — say, a 25% risk of knowing what the future will
in intensive care units.1,2 Patients death within 6 months — it is bring is psychologically diffi-
say that understanding prognosis not clear whether the patient is cult. Worrying about the future
is important for making life 1 of the 25 out of 100 who will may impede their ability to enjoy
choices, such as engaging in fi- die or 1 of the 75 who will live. the present. They may be con-
nancial planning, arranging cus- Some people believe that the sumed by trying to figure out
todial care, and deciding when best approach to this problem is whether things are getting bet-
it’s important for long-distance to generate and analyze more ter and therefore become hyper-
family members to visit.2 data so that we can know what aware of any physical changes
Despite a proliferation of data the future will bring. Improving that occur. Families may spend
about prognosis and life expec- the accuracy of our prognostic a great deal of time acquiring
tancy, our best estimates still estimates is indeed critically im- information in an effort to learn
carry a high degree of uncer- portant — reducing uncertainty more about what the future will
tainty.3 First, 95% confidence is helpful for clinicians and pa- bring and may focus excessively
intervals indicate variation in tients alike.2 On the other hand, on the medical details. For both
the survival of people with simi- the quest for prognostic certain- patients and family members,
lar health conditions and limita- ty has been described by our anxiety may increase.

2448 n engl j med 368;26 nejm.org june 27, 2013

The New England Journal of Medicine

Downloaded from nejm.org at UNIV OF SO DAKOTA on July 1, 2013. For personal use only. No other uses without permission.
Copyright © 2013 Massachusetts Medical Society. All rights reserved.