I would like to add one comment to this.
In Medicine (as in many other fields of evidence and decision
which affect the public and the providers), there are two points
of view in any interaction involving decision.
The Administrator's View.
Patients are exchangeable; their individuality is irrelevant.
Our objective is to deliver a globally optimised provision.
The Patient's View.
I am not exchangeable with anyone, dammit. I want the best
treatment for me. What is best for anyone else may or may not
be best for me. Evidence gathered from other people may or
may not be relevant to what is best for me.
A couple of examples of the dilemma:
[A]: Breast Cancer Screening (using round figures)
Say true incidence in women who present with suspicious superficial
signs is 1 in 500. Say a mammography procedure has accuracy 95%
(probability that result = true condition is 0.95).
A woman is sent for a screening, and the result is positive.
Then the chance that she has the disease, given the result,
is (Bayes's Theorem) 19/518 = 0.037
From the Administrator's point of view, the procedure catches
19 out of 20 cases; but even so there are more than 27 false
positives for each true positive. Is it worth doing? Count the
From the Patient's point of view, she has just got a positive
result from a procedure with 95% chance of being correct. Shock!
How relevant to her is the low (1/500) incidence of the disease
in the population?
Her personal risk of the disease is a compound of the chances
arising from her inherited propensity and her personal history
of exposure to specific risk factors. If these average out to
the population risk (an "ergodic" hypothesis) then she can (at
least approximately) equate her own risk to the population risk,
given the result. But she can also view the result as indicating
that her personal risk-factors are in the higher-risk region of
the distribution. If she can get that far with the calculation ...
[B]: Evidence of Crime
It is a basic principle of Criminal Law (in the UK and, as far
as I know, in the US too) that one may only be convicted of a
crime if the evidence establishes guilt "beyond reasonable doubt".
Generally, evidence about the incidence of the crime amongst
people with similar characteristics is not admissible as evidence
of guilt of a particular accused. Evidence must be specific to
the circumstances of the particular person accused.
Thus, even if 99% of those driving away from a bar at 11pm on a
Saturday night are over the legal alcohol limit, the Police
cannot arbitrarily arrest a random one of these and then simply
claim in Court that "he must be guilty because nearly everone who
does this is guilty". In other words, you can't appeal to Bayes's
Theorem using the population propensity as a prior.
It can certainly be reasonable grounds for suspicion, justifying
arrest, but the only way to get a legitimate conviction is to
carry out an alcohol test on that individual; only if that result
is so high that the individual cannot reasonably plead that they
were under the legal limit would a conviction in Court ensue.
There has been a period in recent times when the Legal Mind has
been a bit fuddled about the statistical principles involved in
evaluating evidence -- such as DNA evidence in Court, or the
notorious series of trials (Sally Clarke and Trupti Patel being
particularly well known) of parents convicted of killing their
children because they had the misfortune to be associated with
three "Sudden Infent Deaths" (SIDS) amongst their children (this
having been held so improbable to have arisen by chance that
they must be guilty, despite the fact that, very rare though
it be, every few years in a 100-million population such a triple
can be expected to occur).
There was in English Law one exception to this principle, now
repealed but still in operation as late as 1970, namely the crime
of Riotous Assembly. According to this law, if N (N=12 as I recall)
or more people are gathered together for a common purpose, and M
(I think 2) or more act in such a way as to put reasonable persons
in fear, then all present are guilty of Riotous Assembly.
Thus if say 100 turned up for a peaceful demonstration, and a few
of them lost their heads, then (provided it could be proved beyond
reasonable doubt what these few had done) all present were guilty,
regardless of their intent or of their views of the actions of the
few. The last time this was invoked was in relation to the notorious
"Garden House Riot" in Cambridge (England) in February 1970: some
hundreds of people turned up outside the Garden House Hotel one
evening, when a politically-motivated "Greek Evening" was held
(at the time of the Military Junta "regime of the Generals" in
Greece) -- Greek food, speeches from Greek dignitaries, etc.
A few people, most not identified, banged on the windows and broke
them. Some got inside. The diners inside were frightened. The Police
turned up, some people still present were arrested, and brought to
trial on the charge of Riotous Assembly. Whether or not these had
done anything, it was decided that what took place was a Riot, and
that these people were present, and therefore they were guilty (and
got long prison sentences).
So that is a unique instance of evidence about what other people
do being direct (and admissible) evidence of your own guilt. But
that is because it was so defined in the Law (which is no longer
Just a few throughts ...
On 08-Jul-09 18:47:59, Jim Walker wrote:
> Hi, Ben.
> It may be theoretically impossible, but I feel like I do it. And when I
> feel like I am doing it, the process feels different and seems to lead
> to different outcomes than when I don't try do it.
> Can clinical expertise enable patients to assess what they will make of
> and do with evidence (along with being the basis of the clinician's own
> decision making)?
> Is this the sort of conundrum that prompted the adage, "In theory,
> theory and practice are different, but in practice they aren't."?
> James M. Walker, MD, FACP
> Chief Health Information Officer
> Geisinger Health System
> If the human mind was simple enough to understand, we'd be too simple
> to understand it.
> - Emerson Pugh
>>>> "Djulbegovic, Benjamin" <[log in to unmask]> 7/8/2009 2:15 PM
> Neil, Thank you for this. I am really happy to see that this is finally
> happening...for a long time the EBM camp has not talked to
> decision-making researchers, and I hope this will finally open the long
> neglected discussion between evidence and decision-making (necessary
> not sufficient for decision-making)...This also provides the
> to re-define EBM. Most textbook definitions of EBM still talk about
> **integration of best research evidence with clinical
> expertise and patient values.** Based on theoretical ground that
> Neil alluded to this is not possible. Most decision scientists would
> argue that no theory can be both descriptively and normatively (and
> prescriptively) valid. Hence, definition of EBM aiming to achieve the
> stated goals of **integration of best research evidence with
> clinical expertise and patient values" is theoretically impossible:
> integration of normative concepts (evidence and patient values) with
> descriptive theories of decision making (clinical expertise) to become
> prescriptive theory (improve decision making) cannot be done (at least
> at today's level of scientific development and as long as humans think
> the way they do...).
> Which is not to say that there is no much to improve how we practice
> medicine... but we need to know what we can and cannot do...
> Benjamin Djulbegovic, MD, PhD
> Professor of Medicine and Oncology
> University of South Florida & H. Lee Moffitt Cancer Center & Research
> Co-Director of USF Clinical Translation Science Institute
> Director of USF Center for Evidence-based Medicine and Health Outcomes
> Mailing Address:
> USF Health Clinical Research
> 12901 Bruce B. Downs Boulevard, MDC02
> Tampa, FL 33612
> Phone # 813-396-9178
> Fax # 813-974-5411
> e-mail: [log in to unmask]
> Campus Address: MDC02
> Office Address :
> 13101 Bruce B. Downs Boulevard,
> Tampa, FL 33612
> -----Original Message-----
> From: Evidence based health (EBH)
> [mailto:[log in to unmask]] On Behalf Of Maskrey
> Sent: Wednesday, July 08, 2009 12:41 PM
> To: [log in to unmask]
> Subject: Re: List of biases
> Thanks for starting this string Jon. We've discussed the importance of
> how human beings, clinicians and patients and both together, acquire
> use information to reach decisions in health care in the past and
> finally prompted me to tap to this group.
> Apologies for the length of this. I've kept it as short as I can, and
> tried though failed to avoid stating the obvious. No doubt you'll tell
> Ian Scott has an excellent paper in the 4 July BMJ - "Errors in
> clinical reasoning: causes and remedial strategies". For those
> interested, two easy-reads as introductions to the topic are Gerd
> Gigerenzer's "Gut Feelings" and Stuart Sutherland's "Irrationality: the
> enemy within".
> I'll summarise. There are essentially two systems we use in acquiring
> information and reaching a decision. So this is any decision, including
> in health care decisions both diagnosis and management. A system 1
> approach involves acquiring a limited number of information items and
> using them to make a "fast and frugal decision". The human brain has
> limited capacity and we need to truncate the volume of information and
> process it in order to be able to make a decision.
> System 2 is more systematic with lengthy and detailed collection and
> analysis of data. If we were buying a new car we'd almost certainly
> some magazine articles and talk to some people who we trusted for the
> knowledge about cars. I'd guess almost no one would go to the
> reviews re automotive engineering on the Society of Automotive
> website. And if we did, that might not improve the decision.
> System 2 is what we'd expect to underpin health care decisions,
> especially those big decisions that don't thankfully have to be made
> more than a few times in a lifetime. Unfortunately the data shows that
> clinicians make decisions based on brief reading and talking to other
> people (Gabby and le May BMJ 2004). That sounds uncomfortable but
> there's some reassurance in that Gigerenzer shows in some settings
> system 1 based decision making can out perform system 2 based decision
> However, that seems less credible in more complex decisions and the
> list of cognitive biases described in the literature looks like
> testimony to the fallibility of humans.
> Obviously its not a simple as the above and e.g. the recent BMJ paper
> describing GP reasoning from Carl Henegan et al provides more
> and insights.
> So what's to be done. Well, firstly we've recognised the issue.
> Secondly, it seems sensible to explore teaching some of this in our
> work. Ian Scott summarises the evidence for that and as you'd expect it
> looks like its a slim volume as yet. There might be some harms from
> taking this too far too soon, but we are all too aware of the greater
> harms from the under and over implementation of health care
> For me this links into where we're going with EBM. I think we're in the
> 3rd age and tip-toeing into the 4th. EBM v1.0 was conceptual. EBM v2.0
> was and is technical development - searching skills, MA stats etc. EBM
> v3.0 was and is the industrial age - Cochrane collaboration, teaching
> critical appraisal to anyone who shows interest etc. And so I'd profer
> that if we want to make EBM really live for people (cf threads on this
> group on a number of occasions) EBM v4.0 needs to be Customer Focussed.
> The customers are clinicians, patients, commissioners and no doubt
> others. And if we want to reach those groups we need to think about
> as humans use EBM outputs as information and how they use those to make
> decisions. People need to be careful who they talk to and what they
> (briefly) read.
> And in order to broaden our understanding of why high quality evidence
> still doesn't get incorporated into practice quickly, we need to have a
> dialogue with other specialties - sociologists, ethicists and as in
> area the cognitive psychologists.
> In summary, we teach health care professionals so they can make
> decisions. It seems perverse not to describe to them how they as human
> beings make decisions - so they at least stand some chance of being
> aware of some of their own and others potential foibles, doesn't it?
> There's a few of us who've produced (at last) 5 papers which include
> the above and lots more. They've been accepted by the UK RCGP journal
> for GP Registrars InnovAiT and are in press. I'll post links as and
> they appear.
> Love to talk to people about this.
> Neal Maskrey
> Director of Evidence Based Therapeutics
> National Prescribing Centre
> -----Original Message-----
> From: Evidence based health (EBH)
> <[log in to unmask]>
> To: [log in to unmask]
> <[log in to unmask]>
> Sent: Wed Jul 08 07:38:24 2009
> Subject: List of biases
> Hi All,
> I thought I'd share this fascinating list of cognitive biases found on
> wikipedia http://en.wikipedia.org/wiki/List_of_cognitive_biases.
> What struck me was two things:
> 1) The sheer volume of potential cognitive biases
> 2) We often focus on methodological biases, but irrespective of the
> purity of methodology, there is still a mountain to climb when it comes
> to overcoming cognitive bias.
> Best wishes
> PS I feel the need to play a game of cognitive bias bingo!
> IMPORTANT WARNING: The information in this message (and the documents
> attached to it, if any) is confidential and may be legally privileged.
> It is intended solely for the addressee. Access to this message by
> anyone else is unauthorized. If you are not the intended recipient, any
> disclosure, copying, distribution or any action taken, or omitted to be
> taken, in reliance on it is prohibited and may be unlawful. If you have
> received this message in error, please delete all electronic copies of
> this message (and the documents attached to it, if any), destroy any
> hard copies you may have created and notify me immediately by replying
> to this email. Thank you.
E-Mail: (Ted Harding) <[log in to unmask]>
Fax-to-email: +44 (0)870 094 0861
Date: 08-Jul-09 Time: 22:16:43
------------------------------ XFMail ------------------------------