2 articles of interest - Are more heads better than one?

Bruno, Michael mbruno at PENNSTATEHEALTH.PSU.EDU
Mon Mar 4 02:50:40 UTC 2019


But you make some very good, and important points, as always, Lenny.

Even if we all were agreed that universal double-reading is life-saving and should be done, the reality is that there are simply nowhere near enough radiologists available to even hope to accomplish it.  The most recent estimate I heard was that there are only about 30,000 practicing radiologists in the U.S., and its clear we can't even get around to reading the current large and growing volume of studies that we have now (perhaps as many as a billion per year?) even once, much less twice.  So, as a result, a large number of studies are never actually interpreted formally by a radiologist at all, and we radiologists today are reading studies faster than ever just trying to keep up.

Yes, AI may help with this problem in the future, but as you point out, we could easily get into a credibility spiral when disagreements arise between the AI and the "expert" human, as you point out.  When two (human) expert radiologists disagree, we have a bit of a conundrum on our hands, and we know that those sorts of disagreements actually happen fairly frequently, including in essentially every malpractice litigation case.

For now, at least, humans seem to trump the robots when we disagree, but that may not always be the case.  And the advent of AI raises some other troubling ethical questions--which is the subject of another discussion.

I am very much looking forward to your lecture in May; your topic is near and dear to my heart.  Plus, it doesn't hurt that the meeting will be in Honolulu this year!  It's snowing out my window as I write this...  ugh!
All the best,


Michael A. Bruno, M.D., M.S., 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 pennstatehealth.psu.edu<mailto:mbruno at pennstatehealth.psu.edu> |
*****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: Leonard Berlin <lberlin at LIVE.COM>
Sent: Sunday, March 3, 2019 5:57 PM
Subject: Re: [IMPROVEDX] 2 articles of interest - Are more heads better than one?

Dear Rob (whom I only know by name, but a friend anyway), and my good friends Mike and Kimberly:

Mike's article highlighting double reading refers to interpretations by residents vs interpretations by experienced staff radiologists.  In that  situation the overread makes sense and is probably necessary.  As for addendums placed on interpretations that had been made and now when another radiologist sees something different as mentioned by Kimberly, then the addendum is probably necessary. But that doesn't occur often and is not a routine situation.

Much has been written regarding routine double-reading in the US literature, and it's almost universal that American radiologists have spoken against doing it.  It's time-consuming and expensive to say the least, but consider the following scenario:
Rob orders an abdominal CT on a patient: radiologist 1 concludes that the findings indicate ABC; but a second reader (radiologist 2)  who is routinely called, concludes that the findings are XYZ.  So now a third radiologist-umpire is called to settle the difference: he agrees with rad 1, but disagrees with rad 2.There's no time to call a 4th radiologist, so Rob gets a report that concludes that "by a 2 to 1 margin, the findings are ABC."

I'm sure Rob would be really comfortable and well assured with a conclusion such as that!

On a more serious view, double reading is routinely done in a few European nations, mainly Norway.  Articles written by them indicate that they do find and correct some errors in up to 10% of cases; however, whether the health of a significant number of patients improve as a result of routine double reading is vague at best.

As we all know, volume of radiologic studies, especially CTs and MRIs, has been increasing tremendously.  Radiologists are unfortunately reading faster in order to keep up; taking time for a routine overread for every case is truly impossible, and of questionable value.

AI (artificial intelligence) may bring something new, indeed. However, how about Kimberly, or Mike, or myself, reads a C, then having it checked by a robot.  But the robot disagrees; so another robot is called and it disagrees with the first robot.  What do we do then??

Stay tuned....,....

Best to all,

Leonard Berlin, MD, FACR

PS: Kimberly and Mike, I'll be speaking about discrepancies, errors, and double reading at the ARRS Annual Meeting beginning May 5.  I don't know yet whether the ARRS leadership will have a second radiologist over-lecture  my lecture, and if they decide to do so, what will happen if the second radiologist disagrees with me ?
Perhaps Kimberly or Mike can be the umpire!


To unsubscribe from IMPROVEDX: click the following link:


Visit the searchable archives or adjust your subscription at: http://list.improvediagnosis.org/scripts/wa-IMPDIAG.exe?INDEX<https://urldefense.proofpoint.com/v2/url?u=http-3A__list.improvediagnosis.org_scripts_wa-2DIMPDIAG.exe-3FINDEX&d=DwQGaQ&c=_FmMnDvUH5queZcSmOuBzHZMbp7E7EwtGwv5cxxnTj0&r=XZJky8Jx0OuETXcWpBMhx9j_wSYpSZPDVXdInJ5O9gQ&m=WLe4o5SBZXYPU7lLvvv9oM548Pg-d28cAbdnIX1t5Qg&s=csKsmicU6EIrEl6XtoqeJPqkj2qvHH1pNyIodByaiRk&e=>

Moderator:David Meyers, Board Member, Society for Improving Diagnosis in Medicine

To learn more about SIDM visit:

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

HTML Version:
URL: <../attachments/20190304/4807ae0d/attachment.html>

More information about the Test mailing list