commentaries

Heuristics and biases ) a biased perspective on clinical

In a recent issue of the British Medical cerns regarding the nature of the There is substantial evidence in
Journal, Klein1 has reviewed several advice provided to clinicians. health sciences education that
biases in reasoning to which clini- highly performing individuals can
cians may fall victim. These biases • The representativeness heuristic: be underconfident just as readily
are not new; all were identified first people make categorical judge- as poorly performing individuals
by Tversky and Kahneman in a clever ments on the basis of how much can be overconfident.7 In addi-
series of experiments conducted an individual example resem- tion, we think that alternative
mainly in the 1970s on ubiquitous bles the stereotype of the cate- explanations for the mechanism
samples of psychology undergradu- gory, largely ignoring the whereby illusory correlations are
ates.2,3 Since then, physicians have relative likelihood of falling into perceived have better empirical
been shown in numerous studies to each category. support.8 That said, we have no
be susceptible to the same biases, the • The availability heuristic: categor- doubt that the psychological bia-
results and warnings being summa- ical judgements are influenced ses described by Klein are present,
rised in papers such as Klein’s at by how easily similar examples to a greater or lesser degree, in
regular intervals.4,5 can be retrieved from memory. most areas of human judgement,
For example, physicians tend to including medical decision-mak-
overestimate the risk of opioid ing, as described in many of the
Physicians can expect to be suscept- analgesic addiction as a result of references Klein provides. How-
ible to decision-making biases the amount of publicity the ever, as Gigerenzer has pointed
problem has received. out, some studies suggest that
• Overconfidence: people tend to be departures from normative ra-
We do not disagree with Klein overconfident in the accuracy of tional models are not just adaptive
regarding the presence of these their judgements. For example, in terms of providing efficient
biases, but we do think that there is a people tend not to recognise modes of human processing but
danger in suggesting that they are shortcomings in their own also provide reasoning mecha-
universally bad and easily overcome. knowledge of pain control. nisms that can be as (or more)
In fact, we too will travel a familiar • Confirmatory bias: people tend to effective relative to more formal-
road by supporting Gigerenzer, a seek out data to confirm, not ised rational strategies.
psychologist who has, since the disprove, their hypotheses. For
1990s, argued for the utility of the example, the interpretation of
strategies that enable these biases.6 information obtained toward the Heuristics, while logically
In the present paper, we review end of a medical work-up tends irrational, are maintained because
briefly the nature of the biases raised to be biased by judgements made they are adaptive
by Klein, using some of her exam- earlier in the case presentation.
ples, simply so that the reader can be • Illusory correlation: people tend
oriented to the substance of our to emphasise one cell in a Klein’s advice is to avoid the
critique. We then present our con- 2 · 2 table of association. For biases outlined by raising aware-
example, Klein argues that ness: awareness of the likelihood
homeopaths often pay partic- of particular events, of one’s
ular attention to patients who shortcomings, of the diverse fac-
McMaster University, Hamilton, Canada.
improve, thus leading them to tors that influence decision-mak-
Correspondence: Kevin W. Eva, MDCL 3522,
believe erroneously in the ing, of decreased vigilance and of
McMaster University, Hamilton, L8N 3Z5, instances that do not fit with
Canada. E-mail: evakw@mcmaster.ca effectiveness of their treat-
ment. assumed relationships between
doi: 10.1111/j.1365-2929.2005.02258.x

870
Blackwell Publishing Ltd 2005. MEDICAL EDUCATION 2005; 39: 870–872
 871

variables. This advice is worrisome, we have shown that expert physi- Nor is it reasonable. When errors
in that it does not take into cians often use a strategy of Ônon- arise they may be the result of
account the complexity of the analyticalÕ reasoning, matching to reliance on imperfect heuristics.
cognitive mechanisms whereby previous cases, which is both effi- However, were it possible to func-
these biases arise. cient and usually effective.11 An tion in the world of medicine
emergency department physician without such heuristics, errors
These suggestions treat human who encounters a new patient would still arise. Were we to inter-
memory like a light circuit. Inten- under severe distress could Ôration- pret every feature in an acontextual
tionally flip the switch and a light allyÕ waste precious time by collect- manner (i.e. without being influ-
turns on, making the contents of ing dozens of signs and symptoms, enced by confirmation bias) we
the room visible; if you think you being careful to consider all diag- would find ourselves awash in a
are being biased by potentially sali- nostic possibilities, and relating the torrent of apparently independent
ent past experiences, flip the switch findings back to a systematic col- features with no way to organise
and remember the hard-to-remem- lection of base rates through them into a meaningful concept.
ber experiences. Bayes’ theorem; or she could When we have asked research par-
ÔirrationallyÕ act quickly by knowing ticipants to list features objectively
In reality we do not maintain such that this patient is presenting like without attending to diagnostic
control over the processes that another patient seen suffering possibilities, their accuracy has suf-
allow access to the contents of from a heart attack. The latter is fered as a result.12
memory. The tip-of-the-tongue much more efficient and probably
phenomenon provides an excel- as effective. This is not to say that the biases
lent example. It involves knowing identified by Tversky, Kahneman
that the information you seek is in and others should be ignored by
memory, but not being able to It is unrealistic and unreasonable to clinicians. Awareness of these biases
activate that information in a way expect clinicians to think in a non- may be helpful when difficulties
that allows you to use it. Trying human manner arise, even though we do not think
hard is often fruitless and, in that an attempt to be constantly
fact, directing more mental energy vigilant and eliminate any bias
to the task can actually be detri- Clearly, this is an extreme example, when it arises is either possible or
mental (a process known as but the same principle holds for desirable. Psychologists studying
blocking9). many types of health-care decisions. these processes have yielded great
Heuristics (i.e. decision-making insights into the capacities and
short-cuts) such as the ones out- tendencies of human thought; they
Successful heuristics should be lined by Klein evolve because they are not simply in the business of
embraced rather than overcome tend to yield better outcomes than showing how fallible and hopelessly
more careful ÔrationalÕ processes. biased people are. In fact, psychol-
Heuristics that yield an unaccepta- ogists are the only group we know
Perhaps more fundamentally, it has ble quantity or quality of errors of for whom ÔbiasÕ is not a negative
to be noted that much of the become extinct quite naturally. word. It is simply a fact, for better
mental activity in which we engage Intentionally avoiding them by rather than worse.
occurs outside conscious aware- expending enough time and
ness, not due to inattention but energy to solve one problem care-
because we simply do not have fully and systematically may prevent REFERENCES
access to it. Social psychologists one from spending enough time
have collected an impressive body and energy on a dozen others. 1 Klein JG. Five pitfalls in decisions
of evidence to suggest that much of Physicians face a substantial about diagnosis and prescribing.
our behaviour, motivations and amount of routine in their day-to- BMJ, 2005;330:781–4.
feelings are guided by automatic day practice and it is probably 2 Kahneman D, Slovic P, Tversky A.
processes over which we have no impossible to avoid biased short Judgment Under Uncertainty.
control.10 Others have gone so far cuts vigilantly in the face of such Heuristics and Biases. Cambridge:
Cambridge University Press 1982.
as to argue that it has to be this way ongoing routine. Nor is it possible
3 Kahneman D, Tversky A. On the
for humans, with our limited com- to simply force oneself to be aware
psychology of prediction. Psychol
putational power, to manage in an of things of which one is unaware. Rev 1973;80:237–51.
environment that is teeming with Asking physicians to do so is unre-
information.6 In medical diagnosis, alistic.

Blackwell Publishing Ltd 2005. MEDICAL EDUCATION 2005; 39: 870–872

4 Friedman CP, Gatti GG, Franz TM, 7 Eva KW, Regehr G. Self-assessment Press, Harvard University Press
Murphy GC, Wolf FM, Heckerlings in the health professions: a refor- 2002.
PS, Fine PL, Miller TM, Elstein AS. mulation and research agenda, in 11 Hatala R, Norman GR, Brooks LR.
Do physicians know when their press. Influence of a single example
diagnoses are correct? Implications 8 Gilovich T. How We Know What Isn’t upon subsequent electrocardio-
for decision support and error So: the Fallibility of Human Reason in gram interpretation. Teach Learn
reduction. J Gen Intern Med Everyday Life. New York: The Free Med 1999;11:110–17.
2005;20:334–9. Press 1991. 12 Norman GR, Brooks LR, Colle CL,
5 Elstein AS. Heuristics and biases: 9 Schacter DL. The Seven Sins of Hatala RM. The benefit of diag-
selected errors in clinical reason- Memory. How the Mind Forgets and nostic hypotheses in clinical rea-
ing. Acad Med 1999;74(7):791–4. Remembers. New York: Houghton- soning: experimental study of an
6 Gigerenzer G, Todd P, ABC Mifflin Company 2001. instructional intervention for for-
Research Group. Simple Heuristics 10 Wilson TD. Strangers to Ourselves: ward and backward reasoning.
That Make Us Smart. New York: Discovering the Adaptive Unconscious. Cogn Instruct 2000;17:433–48.
Oxford University Press 1999. Cambridge, MA: The Belknap

Outreach is better than selection for increasing diversity

Students admitted to medicine in tical financial reasons for medical to higher drop-out rates, partic-
the United Kingdom do not schools to widen participation, in ularly with regard to long medical
match the national profile of addition to considerations of social courses.
social and ethnic diversity.1,2 This equity and health care.
is not just harmful from the point Because medicine has many more
of view of equity: there is evidence Factors limiting access to medicine applicants than places, selection is
that candidates from particular have been studied insufficiently. the best studied of these compo-
cultural backgrounds work more For access to higher education in nents. Affirmative action strategies
effectively within that background, general, social class is the main to address historical imbalances
and thereby produce better health predictor of academic achieve- are, however, politically and ethic-
care indicators for society as a ment.5 Because medical courses ally controversial. In the United
whole3. are significantly longer, there is States there has been a tradition of
likely to be an additional disin- affirmative action for certain categ-
Addressing these imbalances has a centive for disadvantaged groups.6 ories of candidates. The rationale
high political priority, evidenced by Finally, societal and family expec- for this has been that the societal
the introduction of the Office of tations are significantly different imbalances for some social group-
Fair Access (OFFA), which can within different backgrounds.5 As ings have been so great that selec-
prevent institutions without an a consequence, applicants to tion based on academic
acceptable widening participation medicine and dentistry are signifi- performance is in itself discrimin-
strategy from charging full fees4. cantly skewed in their class com- atory. While there have been chal-
Government statements suggest position (Table 1). lenges to this approach, the most
that OFFA will pay particular recent ruling from the Supreme
attention to socio-economic disad- There are three components to Court (July 2003) is that affirmative
vantage. There are therefore prac- widening participation. The first is action is legal and not an infringe-
outreach, to encourage a greater ment of the rights of others as long
diversity of candidates to apply. as it is individual and not the sole
The second is selection, to ensure basis of selection.3 Guidance within
Peninsula Medical School, Plymouth, UK. that candidates are not disadvan- the United Kingdom also suggests
Correspondence: J. C. McLachlan, Professor of taged by their background. The that while applicants cannot be
Medical Education, Peninsula Medical School, third is retention, to ensure that treated differentially on the basis of
C306 Portland Square, Plymouth PL4 8AA, UK. candidates from disadvantaged their background, individual cir-
Tel. +44(0)1752238005; Fax: +44(0)1752 238000;
E-mail: john.mclachlan@pms.ac.uk backgrounds do not suffer cumstances can be taken into
disproportionate hardships leading account.7
doi: 10.1111/j.1365-2929.2005.02257.x

Blackwell Publishing Ltd 2005. MEDICAL EDUCATION 2005; 39: 872–875