Error Rates: Diagnosis--neither numerator nor denominator is known

Graber, Mark Mark.Graber at VA.GOV
Sun Dec 15 15:18:06 UTC 2013

The problem of not being able to count diagnostic errors is one that keeps me up at night.  Ross is correct in describing the fundamental challenges that arise in trying to establish the numerator.  If its true that 'you can't improve what you can't measure', that will leave diagnostic error at the starting gate.

I've reluctantly come to accept the fact that he is correct - we really CAN'T measure every diagnostic error.  Its not like falls or wrong-site surgery, that are more easily tallied.  But .....  I believe it may be possible to establish a quantitative approach that's not as good as being able to count them all, but may suffice for advancing diagnostic science:

 *   If we use a specific definition and a defined methodological approach, we should be able to reproducibly count certain types of errors in certain settings.  Example:  delays in diagnosing colon cancer.  We will be able to measure the delays, even if we can't measure all the missed diagnoses

 *   If we focus on specific steps of the diagnostic process, we should be able to reproducibly count lapses in specific steps.  Example:  how many patients don't have a differential diagnosis listed on their chart;  How many critical alerts aren't addressed in a timely manner.

Our efforts to reduce diagnostic error will be seriously stymied if we can't come up with at some ways to quantify the current state of affairs and see if interventions have any impact, in a quantitative sense.


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

From: Ross Koppel <rkoppel at SAS.UPENN.EDU>
Reply-To: Society to Improve Diagnosis in Medicine <IMPROVEDX at LIST.IMPROVEDIAGNOSIS.ORG>, Ross Koppel <rkoppel at SAS.UPENN.EDU>
Date: Wed, 11 Dec 2013 22:27:48 -0500
Subject: Re: [IMPROVEDX] Error Rates: Diagnosis--neither numerator nor denominator is known

The discussion on measurement of error and error rates is wonderful and
What seems to be missing from much of this conversation, however, is
that reality that we usually have no measures of the numerator -- the
number of errors;  and we have an often slippery measure of the
denominator-- number of diagnoses, number of correct diagnoses
determined by case review and autopsies, number of correct diagnoses
determined by a brighter doctor, number of correct dx when the patient
was discharged (sic), number of med orders, number of med orders
actually administered,  number of opportunities to order, number of
patients, number of meds administered correctly, etc.
Of course, the denominator and the numerator obviously differ by the
measure on which we focus:
Some errors are easy to spot, e.g.,  leaving hemostats in the thoracic
cavity is a classic.
Med prescription errors are very hard to know.  Patients are sick, old,
have 5 comorbidities and are on 13 other meds.  Bad things happen when
we do the right thing, good things happen when we do the wrong thing,
polypharmacy is pandemic, what is ordered may not have been
administered, no one knows the drug-drug-drug interactions among 13
drugs, etc.     Case review is dependent on the right Dx (tautology
alert) and knowing which tests to order, etc.

And then we come to Dx errors. Again, some are easy, but, as
demonstrated powerfully by this group's insightful conversation, many
are profoundly hard to determine...especially in the first set of
iterations.  Some may be impossible. Thus, the  discussion of error
rates makes me cry for a massive dose of methodological caution and even
more humility.

None of this is to say the discussion is neither productive nor
exciting.  Of course not!  But the epistemological issues here are
beyond most of the problems of science.  One of the reasons I so respect
medical thinking is because physicians must deal with so many unknowns
and uncertainties.  My colleague, Renee Fox, noted that one of the first
things we teach a young doctor is to distinguish between what he/she
does not know and what Medicine does not know.  With diagnostic error,
we have both of those factors plus the more vexing problem of what is
knowable and what is knowable within the constraints of a specific
patient's life and setting.

Caveat: I'm not a physician, but I have spent the last 47 years focused
on research methods, measurement of error, and statistics.

Ross Koppel, Ph.D. FACMI
Sociology Dept and Sch. of Medicine
Senior Fellow, LDI, Wharton
University of Pennsylvania, Phila, PA 19104-6299
215 576 8221 C: 215 518 0134

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>

More information about the Test mailing list