Errors in Oncology Pathologies

Swerlick, Robert A rswerli at EMORY.EDU
Thu Aug 15 17:55:00 UTC 2013

I have been mulling over this exchange. I actually find the whole idea of the history or clinical information "creating" bias rather odd. Of course it creates bias but it that such a bad thing? Is the influence of history before looking at test data any worse than the influence of test data before you look at other clinical data?

In my training we were taught to do skin exams before we took any history and I was never convinced that it yielded superior results than knowing the history first.

As I noted above information can and should bias our judgments. That is the whole purpose of garnering such information, assuming we can avoid premature closure. In my opinion the issues I see in this realm are more driven by inadequate information being used by pathologists than too much. I do not have much interaction with radiologists.

Bob Swerlick

-----Original Message-----
From: Graber, Mark [mailto:Mark.Graber at VA.GOV]
Sent: Wednesday, August 14, 2013 9:56 AM
Subject: Re: [IMPROVEDX] Errors in Oncology Pathologies

I like Ehud's suggested approach too, a lot.  (I copied it just below my message, as it seems to be on a different string at this point).  Reviewing the clinical history AFTER you've performed your first look seems to be an ideal solution.  You avoid the initial possibility of framing and ascertainment bias, you have the chance to incorporate the clinical information, and you enjoy the added benefit of second thoughts -  along the lines of Robert Trowbridge's diagnostic 'time out' concept, or Mamede's 'reflective practice'.

This would seem to be a great research project (comparing accuracy of diagnosis with first vs last review of clinical data) for someone out there (I'm thinking Jena .......).

I hadn't really thought about it, but I actually enjoy using Ehud's approach when I do a urinalysis (I'm a Nephrologist). I do the UA first and then try to GUESS the clinical scenario after I see the findings.  It changes the clinical puzzle into  a game, making it more interesting for me and the trainees.  There is actually some good literature that this kind of mental activation improves decisions and problem solving skills ( and I can use all the help I can get!)

Listserv discussions are one thing, but the real question is how we could translate this suggestion into practice.  Is it ready for prime time or does it need research validation?  How could we move this idea into trials or practice ??  Someone (Ehud - this has your name on it) needs to publish this !


Mark L Graber, MD FACP
Senior Fellow, RTI International
Professor Emeritus, SUNY Stony Brook School of Medicine Founder and President, Society to Improve Diagnosis in Medicine
Phone:   919 990-8497

The earlier comments from Ehud Zammir:

"I believe both of the above are true. Accurate clinical info should be provided to anyone who is required to diagnose, whether it is a radiologist, a pathologist or a colleague clinician asked to provide a second opinion. However, the radiologist/pathologist/clinician providing a second opinion should, in my opinion, keep that referral information to the end of their diagnostic routine. An unbiased, fresh diagnostic process is priceless. If, at the end of an independent assessment (without reading the referral), one reaches a conclusion which is consistent with previous diagnoses,  the diagnosis is likely to be correct (the odds of two independent observes making the same error are fairly slim). If there is a discrepancy, one should go back and reassess, with the clinical information in mind. The conclusion may remain the same (e.g. if referrer's diagnosis is incorrect) or change (initial impression of diagnostician/second opinion incorrect).

If, however, one reads the referral prior to independently assessing the patient/Xray/path slide, there is a potent bias introduced into the process. I call it "the bias of the question posed" and I have seen examples where even the most obvious findings are repeatedly missed by multiple doctors due to such questions. If not allowed to independently assess the data first, many competent doctors will be distracted by the referrer question in the same way the participants in the "Invisible gorilla" experiment did (see YouTube if you don't know it). As an ophthalmologist I see patients who come from other ophthalmologists or from optometrists for a second opinion . I make it a point to ignore the referral information until after I have taken an independent history and examined the patient myself. Many pathologists and radiologists follow the same routine. I think the issue is not whether or not clinical info is available, but whether that info is allowed to colour the diagnostic process. Keep it till the end and you will have a truly SECOND opinion, free of assumptions and distractions."

From: "'croskerry at eastlink. ca'" <croskerry at>
Reply-To: Society to Improve Diagnosis in Medicine <IMPROVEDX at LIST.IMPROVEDIAGNOSIS.ORG>, "'croskerry at eastlink. ca'" <croskerry at>
Date: Wed, 14 Aug 2013 09:01:32 -0400
Subject: Re: [IMPROVEDX] Errors in Oncology Pathologies

Good discussion. Certainly, the context of specific findings is critical to their correct interpretation, so ideally some minimum set of objective data should be provided with specimens and for imaging studies.
But the framing bias that Mark mentions, and ascertainment bias (slightly different but amounting to 'you see what you expect to see') are extremely powerful and can be shown to influence decisions without the decision maker being aware of them.
I strongly support Ehud's approach. The potential for bias to interfere with judgment seems to depend very much on what stage someone else's thinking is introduced - it seems best to delay any decision until you have independently made your own. One strategy we use in emergency medicine is not to read the triage note or the nurse's note (or even listen to any comments from others) until after we have seen the patient. This is cumbersome and redundant at times, but seems to work. I remember a savvy and well regarded radiologist who used to insist on not being told anything about the patient until after he had made his first interpretation of the film. He also used to talk about the benefit of 'not having seen the patient' - implying that patients too can be misleading in the way they frame information to you.

Pat Croskerry MD, PhD, FRCP(Edin)
Professor,Department of Emergency Medicine, Director, Critical Thinking Program, Division of Medical Education, Faculty of Medicine, Dalhousie University, QE II - Health Sciences Centre, Halifax Infirmary, Suite 355
1796 Summer Street, Halifax, Nova Scotia, B3H 2Y9 CANADA

Phone:  902 821 2014 (home)
               902 225 0360 (cell)

From: Michael.H.Kanter at KP.ORG [mailto:Michael.H.Kanter at KP.ORG]
Sent: Wednesday, August 14, 2013 2:03 AM
Subject: Re: [IMPROVEDX] Errors in Oncology Pathologies

I am a pathologist who has practiced for over 20 years.  I think that there is a big difference between improving diagnosis and decreasing diagnostic errors that is not often talked about.  A pathologist will almost never make an error if he/she hedges their diagnosis with words like "suspicious, atypical, cannot rule out, suggest clinical correlation, ect.  In my view, the main reason to look at the clinical histtory and information is to understand what question is being asked by the person doing the biopsy and what information will best help the patient.  So, if one gets a lung biopsy in the clinical setting of a lung mass that is suspicious for cancer and sees interstitial fibrosis, the interpretation and report will be very different than if the biopsy is from someone with diffuse interstitial lung disease seen on x ray where the pathologist should be noting if the findings are consistent with UIP rather than is cancer present or are there atypical cells.  So, in the setting of diffuse interstitial lung disease, without the history, the pathologist will likely just give a description of what is on the slide rather than an informative interpretation.  The idea is that if the pathologist has some sense of the clinical setting, in cases where the slides are not clearly diagnostic of something, a more reasonable differential diagnosis can be made.
         Moreover, a better diagnostic work up can be done by the pathologist.    In the lymph node biopsy example below, the interpretation if benign looking may consider infectious diseases are cause the pathologist to order fungal stains or other stains/cultures than if in an afebrile pt with a history of melanoma.  Sometimes outright errors can be prevented as well.   A bone biopsy in a benign looking x ray can look very worrisome for malignancy if interpreted without thinking of the x ray.
Perhaps the analogy would be if an internist tried to do a physical exam without any history.  What we he/she examine and how would it all be interpreted?

In terms of getting second opinions in all cases with small biopsies, I usually reccommend that people get an odd number of opinions.

Michael Kanter, M.D.
Regional Medical Director of Quality & Clinical Analysis
(626) 405-5722 (tie line 8+335)
THRIVE By Getting Regular Exercise

NOTICE TO RECIPIENT:  If you are not the intended recipient of this e-mail, you are prohibited from sharing, copying, or otherwise using or disclosing its contents.  If you have received this e-mail in error, please notify the sender immediately by reply e-mail and permanently delete this e-mail and any attachments without reading, forwarding or saving them.  Thank you.

From:        Alan Morris <Alan.Morris at IMAIL.ORG>
Date:        08/13/2013 05:00 PM
Subject:        Re: [IMPROVEDX] Errors in Oncology Pathologies


This discussion concerns a version of confounding by intention.  However, since all diagnostic conclusions are influenced by prior probabilities, withholding information is often a bad idea.  In the 1970s, when we developed our first rules for interpreting lung function tests, we required a brief history and physical, and (most important) an articulation of the question the clinician wanted addressed, before we would perform lung function testing.  The current state of clinical practice no long allows that kind of rigor.  Diagnostic rules that incorporate the pertinent clinical information, with their associated probability inferences, can deal with this issue of confounding by intention.  More information, as long as the interpretive rules are comprehensive and adequate explicit, seems clearly the better approach.
Have  a nice day.

Alan H. Morris, M.D.
Professor of Medicine
Adjunct Prof. of Medical Informatics
University of Utah

Director of Research
Director Urban Central Region Blood Gas and Pulmonary Laboratories Pulmonary/Critical Care Division Sorenson Heart & Lung Center - 6th Floor Intermountain Medical Center
5121 South Cottonwood Street
Murray, Utah  84157-7000, USA

Office Phone: 801-507-4603
Mobile Phone: 801-718-1283
Fax: 801-507-4699
e-mail: alan.morris at
e-mail: alanhmorris at

On 8/13/13 4:33 PM, "Robert Bell" <rmsbell at ESEDONA.NET> wrote:

>Good point Mark.
>Would symptoms only and other "tests" work better?
>Rob Bell
>Sent from my iPad
>On Aug 13, 2013, at 1:50 PM, "Graber, Mark" <Mark.Graber at VA.GOV> wrote:
>> Thanks Jena for corroborating the value of Peggy's advice. I've heard
>>very similar comments from my Radiology colleagues, that supplying
>>clinical information improves their ability to make the right call.
>> I'm worried though that this may sometimes bias the review and 'frame'
>>the case prematurely.  These kind of framing effects are SO common
>>elsewhere, as illustrated by the Rory Staunton case (His pediatrician
>>communicated to the ER that she was sending over a boy with
>>gastroenteritis, and no surprise that was the ER diagnosis too,
>>missing his sepsis).  So if I send you a lymph node saying 'Pt with
>>weight loss, night sweats, lymphadenopathy, suspected lymphoma',
>>aren't you going to be biased towards a diagnosis of lymphoma by that?
>> Maybe the better way to frame this is to ask your advice:  How can
>>clinicians best provide clinical information WITHOUT inducing undue
>>bias in the subsequent pathologist's (or radiologist's) diagnosis?
>> Mark
>> Mark L Graber, MD FACP
>> Senior Fellow, RTI International
>> Professor Emeritus, SUNY Stony Brook School of Medicine Founder and
>> President, Society to Improve Diagnosis in Medicine
>> Phone:   919 990-8497
>> ________________________________
>> From: Jena Giltnane <jennifer.giltnane at>
>> Reply-To: Society to Improve Diagnosis in Medicine
>><jennifer.giltnane at>
>> Date: Tue, 13 Aug 2013 15:43:56 -0400
>> Subject: Re: [IMPROVEDX] Errors in Oncology Pathologies
>> Peggy Zuckerman writes, "Clinicians can assist pathologists with more
>>accurate analysis of underlying disease by providing pertinent
>>clinical information and radiologic testing to the pathologists. 4" As
>>a junior pathologist, I am ever so grateful that she included this
>>simple statement. All of the missed and delayed diagnoses I have
>>encountered so far (thankfully, few, but more than I imagined) had a
>>significant component of minimal to no supporting clinical
>>information, or even misleading information. I insist that all of my
>>family and friends get a second oncology and/or pathology opinion when
>>a major treatment decision is based on a limited biopsy. Thankfully,
>>referral pathology is often a "package deal" when a patient seeks a
>>second oncology opinion, and so this may be the best route for
>>patients to pursue. It works best, however, when the pathology
>>material can be received and reviewed with ample time before the patient's visit.
>> Best regards, Jena
>> Jennifer M. Giltnane, MD, PhD
>> Dept. of Pathology, Microbiology, and Immunology Division of
>> Investigative Pathology Arteaga Lab @ Vanderbilt University School of
>> Medicine jennifer.giltnane at

Moderator: Lorri Zipperer Lorri at, Communication co-chair, Society for Improving Diagnosis in Medicine

To unsubscribe from the IMPROVEDX list, click the following link:<br> <a href="" target="_blank"></a>


This e-mail message (including any attachments) is for the sole use of
the intended recipient(s) and may contain confidential and privileged
information. If the reader of this message is not the intended
recipient, you are hereby notified that any dissemination, distribution
or copying of this message (including any attachments) is strictly

If you have received this message in error, please contact
the sender by reply e-mail message and destroy all copies of the
original message (including attachments).

More information about the Test mailing list