Autopsies (in connection with thread about The Value of a Second Opinion at the Mayo Clinic)

Bruno, Michael mbruno at PENNSTATEHEALTH.PSU.EDU
Mon Apr 10 13:21:13 UTC 2017

That’s terrific—Bruce, do you have any estimate of what the rate of Class 1 and Class 2 errors are for autopsies?  Even a rough estimate?


Michael A. Bruno, M.S., M.D., F.A.C.R.
Professor of Radiology & Medicine
Vice Chair for Quality & Patient Safety
Chief, Division of Emergency Radiology
Penn State Milton S. Hershey Medical Center
• (717) 531-8703  |  6 (717) 531-5737
• mbruno at<mailto:mbruno at>  |
[inspired to keep patient safe]

*****E-Mail Confidentiality Notice*****
This message (including any attachments) contains information intended for a specific individual(s) and purpose that may be privileged, confidential or otherwise protected from disclosure pursuant to applicable law.  Any inappropriate use, distribution or copying of the message is strictly prohibited and may subject you to criminal or civil penalty.  If you have received this transmission in error, please reply to the sender indicating this error and delete the transmission from your system immediately.

From: Goldman, Bruce I [mailto:Bruce_Goldman at URMC.ROCHESTER.EDU]
Sent: Monday, April 10, 2017 8:26 AM
Subject: Re: [IMPROVEDX] Autopsies (in connection with thread about The Value of a Second Opinion at the Mayo Clinic)

At my institution, autopsies that discover class I errors are reviewed in a consensus/ QA conference with the other autopsy pathologists, and a consensus as to the interpretation is usually arrived at.

Bruce Goldman MD
Prof Pathol and Lab Med
University of Rochester Medical Center

From: Bob Swerlick [mailto:rswerli at GMAIL.COM]
Sent: Sunday, April 09, 2017 8:52 PM
Subject: Re: [IMPROVEDX] Autopsies (in connection with thread about The Value of a Second Opinion at the Mayo Clinic)

If there is a disagreement, how does one adjudicate the differences? Does one assume the second one is right or do you need to get a third opinion?


On Sun, Apr 9, 2017 at 7:38 PM, Bruno, Michael <mbruno at<mailto:mbruno at>> wrote:
The only method I can think of is to perform blinded second-opinion/second-look autopsies.  Sort of an “autopsy peer review.”

From: Bob Swerlick [mailto:rswerli at<mailto:rswerli at>]
Sent: Sunday, April 09, 2017 7:20 PM
To: Society to Improve Diagnosis in Medicine <IMPROVEDX at<mailto:IMPROVEDX at>>; Bruno, Michael <mbruno at<mailto:mbruno at>>
Subject: Re: [IMPROVEDX] Autopsies (in connection with thread about The Value of a Second Opinion at the Mayo Clinic)

How might one structure such a study? I am at a loss to conceive this. What would be the controls? Obviously if there some gross anatomic finding (ruptured aorta, large thrombosis in a key structure), this could be viewed as definitive, But what of more subtle findings? One would need to know the frequency of such findings in different clinical contexts.
I am reminded of the perception of anatomic pathology as an invariant gold standard and how this has tended to break down. What do you do when the gold standard is not so golden?
Bob Swerlick

On Sun, Apr 9, 2017 at 6:43 PM, Bruno, Michael <mbruno at<mailto:mbruno at>> wrote:

It would be very desirable to know that!  I did a lit search on the topic not too long ago, and could not find a single published study which reported on the error rate of autopsies.  All of the studies I found used the autopsy as the "gold standard" to find other types of errors.

Surely autopsies are also subject to error, no?


-----Original Message-----
From: Traian Mihaescu [mailto:traian at MIHAESCU.EU<mailto:traian at MIHAESCU.EU>]
Sent: Sunday, April 09, 2017 4:18 PM
Subject: Re: [IMPROVEDX] Autopsies (in connection with thread about The Value of a Second Opinion at the Mayo Clinic)

1.A local study from Romania (attached to this mail) found that "the pathology of the respiratory system is the main source of diagnostic discrepancies" and

2.taking into account that frequent discrepancies of tracheobronchial content findings occurres in postmortem CT versus forensic autopsy
(<>)<> the context of my anectodical experiences in some local autopsy misdiagnosis (eg TB)

my question was:

"what is the error rate in autopsies diagnosis itself"?

Traian Mihaescu MD
Clinic of Pulmonary Diseases
Iasi, Romania<>

As an FYI in connection with the autopsy comments in this thread, there are indeed data on the value of autopsies.
The systematic review below (from JAMA in early 2000s and indexed here
in the AHRQ website PSNet
included some studies in 'modern era' and also took into account theidea that clinicians might only ask for autopsy when they are already worried about a diagnostic error. Interestingly that turns out to be not very true. What we did in the analysis in the systematic review was look at error rates as a function of autopsy rates. If clinicians were great at selecting cases for autopsy ( i.e., they asked for autopsy mostly when
they suspected they had missed the correct diagnosis), then studies with high autopsy rates (most deceased patients sent for autopsy) would have low error rates, because all these extra deaths with no concerns about errors would be added. But, in fact, studies from places with high autopsy
rates showed fairly high error rates. In other words, all the patients with no autopsy include a fair number with errors.
For anyone who is interested, what is meant by error here (ie in the very large literature on autopsy detected errors in clinical diagnosis)
usually amounts to the following (based on a classic study by Lee Goldman, the senior author on the review below from UCSF, but then at the Brigham, now Dean at Columbia, I believe) from the early 80s in NEJM and a second study  in the late 80s from same group but with Seth Landefeld as 1st author):
Class I errors: knowing the right diagnosis (ie the one detected at autopsy) would for sure have changed treatment and probably/possibly
would have altered outcome (ie, patient more likely to have left hospital alive).  Major Errors: includes Class I plus Class II, where the latter amounts to saying the diagnosis was really very different than the clinicians thought but it’s not clear that treatment would have been any different. So, for instance, if a patient with chest pain in the ED is diagnosed as an acute MI but in fact has aortic dissection, that’s a Class I error. Not only is the treatment for acute MI not the same as for aortic dissection, it would actually make death more likely.
A less dramatic Class I error might be (bacterial) pneumonia as the clinical diagnosis, but the correct diagnosis is TB.  For an example of a major error, pneumonia but the real diagnosis is post-obstructive pneumonia from metastatic lung cancer. The idea behind calling this a “major error” (but not a Class I error) is that no one could say that missing metastatic lung cancer is unimportant (the way one might say about many ‘incidentalomas’ found at autopsy). But, the fact that treatment probably would not have been any different if the lung
cancer were already  metastatic.  Yes,  I know that some metastatic cancers are stil treatable, but example just meant to convey the general idea that sometimes, even with a major missed diagnosis, there is no impact on treatment. Plus, the point of this classification system was to
convince clinicians using a conservative framework that would not appear to ‘overcall’ diagnostic errors
Anyway, there continue to be studies of autosy detected errors, espiecally in settings like ICUs, where autopsy rates remain high (Winters B et al.
Diagnostic errors in the intensive care unit: a systematic review of autopsy studies. BMJ Qual Saf. 2012 Nov;21(11):894-902.<>) this systematic review, similar to the older JAMA review below, showed that clinically important missed diagnoses stil occur at a substantial rate.
Keep in mind that these error classifications are particularly conservative in that they count the clinicians as having made the diagnosis if they made it any point while the patient was still alive (ie even if they made it too late to save the patient. For instance, recognizing that  a patient was dying of TB or an acute dissection even if  that recognition came too late to save the patient – as long as someone said that’s the likely diagnosis before the patient actually died, it didn’t get counted as an error.)
In terms of the point recently made in this thread about not all diagnostic errors leading to death. That’s of course true in many ways – instead of death, missed diagnoses may ‘just’ lead to years of untreated symptoms, disability and anxiety over no answer. But, the point of these sorts of autopsy studies, especially when undertaken by clinicians rather  than pathologists, was to draw attention to clinicians that, despite
the may incredibly advanced in imaging, serologic testing etc, we stil miss important diagnoses at a not insignificant rate. And, importantly these error rates are not simply an artefact of selection bias.
In terms of why autopsies are so seldom performed, a viewpoint in NEJM from 2008 provides an overview of the various problems. [Shojania KG, Burton EC. The vanishing nonforensic autopsy. The New England journal of medicine 2008;358:873-5.<>

> Kaveh
> Kaveh G. Shojania, MD
> Professor and Vice Chair, Quality & Innovation Department of Medicine
> Director, University of Toronto Centre for Quality Improvement and
> Patient Safety (<>)
Sunnybrook Health Sciences Centre
> Room H468, 2075 Bayview Avenue
> Toronto, Ontario M4N 3M5
> Editor-in-chief, BMJ Quality & Safety
> Shojania KG1, Burton EC, McDonald KM, Goldman L.Changes in rates of
autopsy-detected diagnostic errors over time: a systematic review. JAMA.
2003 Jun 4;289(21):2849-56.

> CONTEXT: Substantial discrepancies exist between clinical diagnoses
> and
findings at autopsy. Autopsy may be used as a tool for quality management
> to analyze diagnostic discrepancies.
> OBJECTIVE:To determine the rate at which autopsies detect important,
clinically missed diagnoses, and the extent to which this rate has changed
> over time.
> DATA SOURCES:A systematic literature search for English-language
> available on MEDLINE from 1966 to April 2002, using the search terms
autopsy, postmortem changes, post-mortem, postmortem, necropsy, and posthumous, identified 45 studies reporting 53 distinct autopsy series meeting prospectively defined criteria. Reference lists were reviewed to identify additional studies, and the final bibliography was distributed to
> experts in the field to identify missing or unpublished studies. STUDY
SELECTION:Included studies reported clinically missed diagnoses involving a primary cause of death (major errors), with the most serious being those likely to have affected patient outcome (class I errors). DATA EXTRACTION:Logistic regression was performed using data from 53 distinct autopsy series over a 40-year period and adjusting for the effects of changes in autopsy rates, country, case mix (general autopsies;
> adult medical; adult intensive care; adult or pediatric surgery;
> general
pediatrics or pediatric inpatients; neonatal or pediatric intensive care;
> and other autopsy), and important methodological features of the
> primary
> DATA SYNTHESIS:Of 53 autopsy series identified, 42 reported major
> errors
and 37 reported class I errors. Twenty-six autopsy series reported both major and class I error rates. The median error rate was 23.5% (range,
4.1%-49.8%) for major errors and 9.0% (range, 0%-20.7%) for class I errors. Analyses of diagnostic error rates adjusting for the effects of case mix, country, and autopsy rate yielded relative decreases per decade
> of 19.4% (95% confidence interval [CI], 1.8%-33.8%) for major errors
> and
33.4% (95% [CI], 8.4%-51.6%) for class I errors. Despite these decreases,
> we estimated that a contemporary US institution (based on autopsy
> rates
ranging from 100% [the extrapolated extreme at which clinical selection is
> eliminated] to 5% [roughly the national average]), could observe a
> major
error rate from 8.4% to 24.4% and a class I error rate from 4.1% to 6.7%.
> CONCLUSION:The possibility that a given autopsy will reveal important
unsuspected diagnoses has decreased over time, but remains sufficiently high that encouraging ongoing use of the autopsy appears warranted.

To unsubscribe from the IMPROVEDX:
or click the following link: IMPROVEDX at LIST.IMPROVEDIAGNOSIS.ORG


For additional information and subscription commands, visit:

http://LIST.IMPROVEDIAGNOSIS.ORG/ (with your password)

Visit the searchable archives or adjust your subscription at:

Moderator: David Meyers, Board Member, Society to Improve Diagnosis in Medicine

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

HTML Version:
URL: <../attachments/20170410/8211e8d9/attachment.html> ATTACHMENT:
Name: image001.png Type: image/png Size: 6043 bytes Desc: image001.png URL: <../attachments/20170410/8211e8d9/attachment.png>

More information about the Test mailing list