Mammogeddon 2017: The Conclusion?

From the ABR’s July 19 email:

Some of you are wondering why it has taken so long for the ABR to provide a solution. We apologize for the delay as we know this has been stressful for you. More than 450 candidates were affected by this situation. The cause of the problem was not initially apparent, and it was important for us to have time to investigate, review preliminary scores of all candidates, obtain direction from our board members and some program directors, and devise a solution that was most appropriate for all stakeholders, including you, your program, and your patients.

The ABR board also received input from the breast imaging community, which feels it is imperative for residents to be tested on breast content at some point in the certification process. The board members considered requiring a breast module on the Certifying Exam for those who did not receive the module on the Core Exam. However, all were concerned that more than two years of delay would require you to study again for the breast module.

The board feels strongly that we must administer the content as soon as possible, and that we should not require travel, other expenses, or additional resources, which is aligned with what we have heard from the breast imaging community. Therefore, we decided that we should trust you to take the online module in a setting of your choice. In addition, the breast module has been carefully edited to ensure that all findings are visible without the need for a high resolution monitor.

[…] We will schedule residents who need to take the breast imaging module at specific times on two dates: September 7 and September 18, 2017. You will select your desired start time when you register.

Still missing: what actually happened in Chicago, what the technical glitch was, how they’ve taken steps to prevent this from happening again, how this module is graded, how “hard” it will be to pass, if it’s actually possible to fail, and a finally—what happens if someone actually manages to fail.

It is interesting that you can take it anywhere you want but that you still must take it at specific times—presumably a compromise to prevent cheating/sharing of the exam content without resorting to using an official testing center. The real exam is proctored with a bathroom monitor, but the fabled mammo content is on the honor code. To me, this is highly suggestive of lip service to an apparently deeply hurt mammography community.

And, speaking of testing centers, the ABR recently released the following narrative about why they haven’t been able to disseminate the exam:

These delivery requirements have proven to be insurmountable obstacles for the numerous commercial testing vendors that we’ve engaged over the years. It’s important to remember that the vast majority of these vendors’ clients deliver text-based question exams with little or no multimedia content.

[…]

Just last year, we engaged two prominent commercial testing vendors to explore our goal of delivering the diagnostic radiology initial certification exams at local testing centers. Both vendors were given in-depth details of our exam delivery needs and asked to provide a proposal for our consideration […]

…but neither was interested.

I like that they’ve finally publically responded to these perennial requests.

I imagine these two were Prometric and Pearson VUE, because (despite the claim of “numerous” vendors) there are only a handful of large commercial testing centers around that could possibly furnish the exam. I suppose it’s possible the big two passed in years past. I have no doubt that the ABR’s demands for administration are not worth the time and expense for most vendors to meet given the low exam volume. The follow-up question, however, is whether or not it’s possible to write a Core exam that can be disseminated.

For example, the video portions of the exam are small in number and generally useless outside of cardiac MR (which, if we’re being honest, plays a comically outsized role on the test). The multi-slice scrolling capacity is rarely used and usually only a handful of images anyway. Mammo and radiographs could be selected that do not require high-resolution high-filesize images. The ACR in-service exam, of note, was able to snag a contract and is also image-based.

We are committed to making the initial certification process as facile as possible. While our past efforts have not been successful, we will continue to pursue our goal (and your wish) of delivering diagnostic radiology exams in local commercial testing centers. As we all know, technology is constantly evolving, and perhaps local exam delivery will become more feasible in the future.

I don’t doubt that the exam the ABR created couldn’t be ported to Prometric as is. Shucks, it didn’t even work in Chicago. But couldn’t we have a Core Exam that was functionally equivalent but wasn’t so off-putting? Exams need to be written with the administration in mind from the onset, not just as an afterthought.

Perhaps putting our hopes in the possibility that bandwidth and memory will be so cheap one day that testing companies won’t find our poorly written and conceived exam so unpalatable isn’t the best plan.

The ABR Mammography Saga Continues

This week’s ABR Core exam snafu update:

Dear ABR Candidate,

The ABR board members and staff sincerely apologize for the problems with the diagnostic radiology Core Exam on Thursday, June 8, 2017 at our Chicago Exam Center. We did not start the exam on time, had intermittent interruptions, and we failed to deliver the breast imaging content to many candidates.  Candidates in Tucson were not affected, nor were candidates who took the exam in either center on June 12-13. We were extremely disappointed, and we know you were too. We have closely examined the situation and made changes to prevent another event like this.  In addition, we have developed a preliminary plan for administration of the breast imaging content to the candidates who did not receive it.

Here is information regarding our plans:

— We are on track to release the Core Exam results by the end of July, 2017.
— Candidates who did not receive breast imaging content will get their Core Exam pass/fail/condition result at the same time as those who did receive the breast content.
— Preliminary results for this Core Exam are very similar to results from previous administrations.
— There are no candidates for whom the presence of the breast imaging module was responsible for a pass or fail result. In other words, people who failed did poorly enough in multiple areas that even a stellar performance on the breast module would not have allowed them to pass.
— Candidates who did not receive breast imaging content will be required to pass a separate breast module, which will be distributed online in September 2017. We are finalizing our plans for this — it will not require travel or additional expense for candidates or their programs, and we anticipate that it will take only about an hour to complete.
— Performance on the breast content for those taking the separate module will not affect their Core Exam result; however, these candidates will be required to pass the separate breast content module in order to be eligible to take the Certifying Exam.

Again, we are truly sorry.  We greatly appreciate your patience while we have worked on the solution to this situation.

The email style has improved a bit since last time.

Summary impression:

  • The sections are a farce
  • The “separate breast module” is a meaningless box-checking endeavor
  • If they can disseminate the “breast content,” then they can distribute the whole thing

The sections are a farce

Of course the ABR would claim that the presence or absence of the mammo section had no bearing on anyone’s actual Core exam results. This conclusion was essentially guaranteed by the ABR’s claim/decision in past years that no one has ever conditioned an individual section outside of physics (which has a higher passing threshold). Essentially, the exam grading paradigm has been structured such that the gap between an overall passing performance and an individual section failure is so wide that no one (n > 4500) has ever managed to fail a single section without first doing so poorly on the exam on the whole that they fail the whole thing outright. This, of course, begs the question, why even pretend to grade each section separately if no one can really fail one?

The corollary to this is that the Core exam cannot actually ensure when you’re really competent in an individual section outside of its overall passing rate. It’s been essentially shown that if you can pass the exam in general, there is no meaningful way for you to fail mammography (or anything else) by itself. The ABR cannot by its own grading system guarantee meaningfully adequate performance in an individual area. Because the grading scheme’s details are kept secret, we can never know what percentage is required to condition or fail the exam. We do know that after four administrations of the exam, it is likely nearly impossible. In real life, we know people are not equally good in all sections. It is not hard to imagine that in some cases someone may just barely pass the exam but still truly be pretty terrible in one section. And yet, this has never borne out with a single non-physics conditioning performance.

This is not to say that I think people should be forced to travel across the country again just to take a one-hour section test—because that would be stupid. Preventing this from happening is presumably one of the reasons why the conditioning threshold for individual sections is so low.

Therefore, the breast module is also a meaningless box-checking endeavor

Based on history and the ABR’s admission that breast module performance had no effect on Core Exam passage, whether or not mammography is actually included in the exam or not is irrelevant from any practical standpoint. Any section(s) could be missed and would likely have absolutely no effect on overall exam passage. What the ABR is admitting with this gesture is not that the Core exam can even guarantee satisfactory competency in an individual section (i.e. that you can actually interpret a mammogram), but rather that it is too embarrassing to simply not test an entire region of the body, perhaps particularly so when the majority of examinees did eventually receive the content.

I do wonder a few things:

  1. Where did this decision come from? Was it from the ABR’s own problem-solving toolbox, or is it a reaction to some perceived MQSA deficiency or sub-specialty push back? Originally, the ABR implied the loss of the mammography content “should” have no bearing on MQSA.
  2. Would this be the solution if it was a different section that was missing? Would people be dealing with the same nonsense for the cardiac section (which as we know is meaningless to the majority of practicing radiologists)? Of course, at this point, the answer to this question would be “Absolutely!” So we’ll never really know.
  3. What happens if you “fail” the breast module? The Core Exam result isn’t malleable, so if this is even possible, does one just simply take it again and again until you pass? Is this the radiology equivalent of the online training modules about information security and fire safety that you pretend to read every year?

If you can distribute one section, you can distribute them all

If the ABR carries out its plan to somehow disseminate a single exam section without any cost to the examinees or programs, they are only two solutions that I can readily think of:

  1. Do it at local commercial venues like Pearson VUE or Prometric and pay the fees for all test-takers out of the ABR’s surely overflowing coffers.
  2. Offer a web-based version (like the ABR’s online practice exam) that can be taken at the resident’s institution (presumably with some form of proctoring).

Either way, making this section and releasing it in either form destroys any claims about the ABR’s inability to do this for the exam as a whole.

Again, I don’t want to diss the ABR’s testing center proper. It’s pretty nice, and the rapid/open bathroom break policy is a welcome change compared with the police state supervision of commercial testing centers. But, it’s still not worth forcing people to travel across the country for.

ABR totally botches 2017 Core Exam

This email belies how royally the ABR botched the 2017 Core Exam.

What the ABR should have done is what any accountable organization should do when they mess up.

  • Express regret and acknowledge responsibility
  • Be transparent and describe the mistake
  • Give an action plan and step to correct the problem
  • Ask for forgiveness

Instead, examinees received the lip service version.

“Technical issue” is not a satisfactory explanation for the cause.

“Problems with the display of some questions” is not what happened.

“Those questions will NOT be counted toward your exam results” is a grossly incomplete solution.

So what did happen?

Well, the ABR still hasn’t offered a technical explanation. It would seem there was an issue with mammo module of the exam. If I had to guess, the larger image file sizes in this module probably exceeded a temporary throttling of server they were hosted on and could not be transferred to all stations as the requests timed out.

But who knows? Apparently not the ABR.

The result of whatever happened is that some examinees in Chicago couldn’t start the exam. Some of them waited nervously in the holding room at the hotel room without explanation awaiting the shuttle. Others already at the center just had to sit at their desks wondering when they would be able to start. For two hours. Which of course turns the already long day into a hellishly long one with nerves racked, tummy grumbling, caffeine wearing off, etc.

Once the exam began, some test-takers had the mammo questions. Others did not. And some had them added to the end of the test mid-way through, suddenly increasing their day by another hour. In all cases, the ABR has suggested that “those questions” won’t adversely affect their scores. This presumably means that no one in Chicago will have mammo graded. But then why add it to some people’s tests and not others? Why make someone whose test-day is already two hours delayed stay another hour for questions that won’t count? How are they going to reconcile the fact that there are psychological and fatigue effects from this mistake that have nothing to do with the “display of some questions,” and that some of this could have simply been mitigated by upfront transparency?

In the grand scheme of things, given that nobody has ever conditioned the mammo section, I imagine the ABR feels confident saying that those questions not being graded will not have a meaningful impact on the grading of the examination itself. With around 103 total fails last year, one imagines only a fraction of those would even include mammo. Even the vast majority of people affected are probably nowhere near the failing mark, unfair psychological BS notwithstanding.

A follow-up email on June 14 (almost a week later) said this (emphasis mine):

The ABR sincerely regrets the problems with the administration of the Core Exam in Chicago on Thursday, June 8, 2017. We are taking this matter very seriously and are working hard to identify the sources of the problem and the impact on affected candidates.

We don’t yet have all the information needed to determine how many candidates have been affected and to what extent. Staff worked very hard over the weekend to ensure that the Core exams administered in Chicago and Tucson this week would go smoothly, and we have had no issues.

I want to emphasize that any candidate impacted by last Thursday’s difficulties with the breast imaging content will not have those items counted against their scores. We don’t expect anyone to have problems qualifying for MQSA.

How can you not know who was affected? The nature of this problem should have made it obvious who was affected during the examination itself. What they mean is that—despite getting into the business of test administration—the ABR never anticipated technical difficulties, has no meaningful system in place for troubleshooting or identifying issues, and had no contingency plans formed to deal with this eventuality.

Also missing: acknowledgment of any the issues outlined above outside of the “difficulties with the breast imaging content.”

And: you don’t “expect” problems with MQSA? The MQSA requirements only state that the radiologist be board-certified, not that the boards actually contain mammography. Of course this shouldn’t be a problem. But if you anticipate that there could be an issue, perhaps you should get some clarification before dropping a half-baked position-statement.1

 

Let’s go back to the underlying arguments for how we got here in the first place.

From the ABR FAQ:

Why do I have to go to Chicago or Tucson instead of a local testing center for diagnostic radiology exams?
With the transition to more image-rich exams with advanced item types, the ABR has built two exam centers in Chicago and Tucson to administer all diagnostic radiology exams. At this time, commercial test centers do not have the technology or means available to support these kinds of exams.

More detail from the 2014 Core Exam FAQ & misconceptions presentation:

Why can’t I just go to a PearsonVUE center to take this test?
• Modular content difficult for PV
• PV can’t handle case structure on their software
• PV monitors aren’t calibrated, can’t control lighting
• Aim: to have distributed exam. We are working on system to implement

So, now in 2017, we can firmly debunk these arguments

1. Modular Content

The content is not bizarrely or unique modular. First, this doesn’t really matter (even the very long Step exams are broken up into multiple modules). In years past, the modules for different sections were given in succession (breast, then cardiac, then GI) though lumped seamlessly into one large mega-module as you progress through the day. This year the modules were jumbled and topics jumped around. Thus there are just two days of relatively unmodular content.

2. PV can’t handle case structure on their software

This is only plausible if the ABR’s software is particularly poorly written. The USMLE also has multiple different case structure formats, including videos, images, and interactive fake physical exams, not to mention Step 3’s ludicrous choose-your-own-adventure CCS program. If we need to get rid of the two or three “drag the X” format questions per test in order to do a disseminated exam, I think we can all agree the collective radiology hivemind would acquiesce.

3. PV monitors aren’t calibrated, can’t control lighting

After this year’s difficulties, one can easily argue that there is no point having a “well-calibrated” monitor that can’t even show the carefully curated “Angoff-validated” questions in the first place. I’ll admit, the lighting is nicely dim. As a practical matter, few images are of sufficient quality for the lighting to be a plausible limiting factor. Most of the MR looks photocopied from books published in the 1980s. Residents take the ACR in-service exam in droves every year. The criticism there has always been the exam itself; not the testing software nor the ambiance of the venue.

4. Aim: to have distributed exam. We are working on system to implement

2018 sounds like a great year to start.

 

The costs of the ABR’s exam paradigm are absurd

There are almost 1200 graduating radiology residents every year (1149 took the core in 2016; 91% passed). Every class contributes $640 per person per year for a total of $3 million per graduating class over the course of a four-year residency ($4.6 million total when including the extra two years to take the Certifying Examination). That also means that the ABR rakes in around $750k per class per year and $3 million per year from residents alone. Not to mention the $340/year for every single radiologist in the MOC phase. Or the $3000+ to take subspecialty exams like neuro or VIR.

To reiterate: the class that just took this failed exam gave the ABR on the order of $3,000,000 to take this test. This figure doesn’t include the additional costs for the honor of traveling across the country to spend two days in a hotel to actually take the exam (at least another $500,000 per year).

If you can’t get photos and radio buttons working consistently on an operating budget of millions, then you’re doing it wrong.

 

Having a decent test is an important noninterpretive skill

When the ABR decided to start from scratch and write a new exclusively computer-based exam, they chose to become not just test-writers but test-administrators. No one forced the ABR to write a test that no high-volume testing center could implement. When you take over something this important, you have to do it right, and you should be completely accountable for your performance. Transparency should not be optionable. The way the Core and Certifying exams were created, graded, and handled is a poorly conceived and unnecessarily obfuscated embarrassment (e.g. why does the Certifying exam even exist?).

You don’t just say things like2

we had a mysterious technical difficulty but also we totally fixed it we promise though actually we don’t know what happened or exactly to whom it happened but also don’t worry about those questions they won’t count for anyone because for real we don’t know who had them or didn’t have them or if they had them how pretty they looked so trust us also by the way your annual fee is due.

Since noninterpretive skills are an important part of the Core exam, let’s just say that a 6% failure rate for successful Core exam administrations is a far cry from Six Sigma.3

ACGME reaffirms independent call for radiology is okay

They didn’t actually do that. That is my subjective interpretation as a random person of the language of the current ACGME Common Program Requirements (emphasis mine):

For many aspects of patient care, the supervising physician may be a more advanced resident or fellow. Other portions of care provided by the resident can be adequately supervised by the immediate availability of the supervising faculty member, fellow, or senior resident physician, either on site in the institution or by means of telephonic and/or electronic modalities. Some activities require the physical presence of the supervising faculty member. In some circumstances, supervision may include post-hoc review of resident-delivered care with feedback.

I think imaging has and should continue to fall under “some circumstances.” Until the machines take over, hold-out radiology programs should strive to maintain their status quos of “post-hoc review.” Efforts should absolutely be made to improve that review process and help residents learn and iterate toward improvement, but the last thing we need in the era of increasing mid-level autonomy is to have graduating residents unable to make a call.

The danger (?) of intravenous contrast media

Another study piling on the mounting evidence that at least modern contrast agents put into people’s veins (and not arteries) for CT scans might not be bad for your kidneys after all.

The biggest single center study of EM patients was just published in The Annals of Emergency Medicine, which studied 17,934 patient encounters and compared renal function across 7201 contrast-enhanced scans, 5499 non-con scans, and 5,234 folks with no-CT.

6.8%, 8.9%, and 8.1% were the rates of AKI respectively. As in, folks who received either no contrast or no CT imaging were more likely to have a significant rise in creatinine than people who got contrast. As in, contrast was protective (statistically). Using different cutoff guidelines for AKI, the three were all statistically equivalent.

Practice patterns here still get in the way. Patients with low GFRs are more likely to get fluids prior to receiving contrast, possibly explaining the pseudo-protective effect of contrast. Patients with poor renal function are less likely to get contrast in the first place, reducing the power for evaluating contrast’s effects on those with CKD. However, controlling for baseline GFR didn’t change the story: there wasn’t an increased risk associated with receiving intravenous contrast in this controlled retrospective study regardless of underlying renal disease.

Historically, randomized controlled trials designed to elucidate the true incidence of contrast-induced nephropathy have been perceived as unethical because of the presumption that contrast media administration is a direct cause of acute kidney injury. To date, all controlled studies of contrast-induced nephropathy have been observational, and conclusions from these studies are severely limited by selection bias associated with the clinical decision to administer contrast media.

Maybe with all this mounting evidence it’s time to do an RCT.