LIVE FROM RSNA: At Massachusetts General, an Assertive Push into CDS for Radiologists

Dec. 1, 2015
Tarik Alkasab, M.D., Ph.D. is helping to lead his fellow radiologists at Massachusetts General Hospital forward in a highly innovative initiative around clinical decision support in a radiology-specific context

At Massachusetts General Hospital in Boston, Tarik Alkasab, M.D., Ph.D., service chief for Informatics /IT and Operations, and a practicing radiologist, is helping to lead an important initiative around establishing and expanding a menu of clinical decision support offerings intended to help radiologists at the very busy hospital, which performs nearly one million radiological studies a year, and whose medical staff includes 200 radiologists—about 120 attendings, about 30 residents, and about 50 fellows. Alkasab spoke with HCI Editor-in-Chief Mark Hagland on Tuesday at the McCormick Place Convention Center in Chicago, during the RSNA Conference, sponsored by the Radiological Society of North America. He and his colleagues are using a solution created by the Burlington, Mass.-based Nuance Communications, in collaboration with the National Decision Support Company (NDSC).

As the press release, issued on Monday, Nov. 30, notes, “By leveraging ACR Select™, nationally-recognized Appropriate Use Criteria (AUC) published by the American College of Radiology and exclusively licensed by NDSC, physicians are guided to order the most clinically-appropriate exam for a patient while working within their native electronic health record (EHR) workflow. ACR Select, which is widely deployed and integrated with major EHRs, is used by physicians nearly one million times per month when seeking clinically-validated imaging guidance when ordering exams. This helps healthcare providers comply with federal law that will require physicians to consult qualified decision support criteria when ordering Medicare outpatient advanced imaging exams.”

Tarik Alkasab, M.D., Ph. in the Nuance Communications booth at RSNA15

Below are excerpts from Hagland’s interview with Dr. Alkasab.

Let’s begin by talking about the clinical and organizational context around your and your colleagues’ work in this area.

Yes, certainly. We have a staff of about 200 radiologists altogether, and do nearly a million studies a year. In that context, creating high-quality, low-variability processes is a challenge. We have so many modality machines. And obviously, we’re managing a number of external imaging sites as well. I spend, and the leadership spends, an enormous amount of time thinking about variability—how to manage variability, how to reduce variability.

What kinds of variability are involved?

It’s everything—setting people up to do the right study is complicated in its own right, and we’ve been one of the pioneers in creating decision support for ordering physicians. We also have systems in place to make sure that we do a particular type of study in a consistent way. Beyond that, the reporting process, in terms of how the radiologist handles the process, is next. One of the things that comes up all the time for a radiologist is seeing a nodule in a lung or a patient’s adrenal gland, in a CT, or possibly an MR. The question for the radiologist is—usually, three’s a follow-up imaging study that a radiologist would recommend. It’s unlikely that these things are significant, but if you found out a couple of months later that the nodule is growing, that would be an important finding.

And you are hoping that radiologists who follow the guidelines will become more consistent clinically?

Yes, we want to help the radiologists following the guidelines. There are dozens now. And because the imaging studies we do see so much, there are all kinds of clinical scenarios that might take place from a particular type of study. Looking at abdomen and chest CTs for a day, you’ll probably come across 5 or 6 pulmonary nodules, 1 or 2 adrenal nodules, something in a pancreas, something in a liver, and they should be saying something consistent in what they’re saying and in what a colleague might say tomorrow, reading those same CTs.

The thing is, as physicians, you weren’t trained to necessarily be consistent with standardized norms, correct? Historically, traditional medical school training has focused on physician judgment at the point of diagnosis, in purely individual situations.

Yes.  It is still important for radiologists to carefully consider the clinical context and the individuality of patients. And one of the most valuable things a radiologist can ask is the “what else can it be?” question. That’s one of the most valuable things we do. But the things that we’re talking about tend to be sort of routine, incidental things, and there tends to be a routine way of handling these things. And if guidelines apply, it’s best for the radiologist to use the guidelines; it’s simpler for everyone to understand. And legally, it creates kind of a safe harbor for them, by following a clear standard of care, which is an advantage. And as things get more consistent, it’s easier for the people to have to do the follow-up to have confidence in the recommendations, as opposed to off-the-cuff recommendations, which can introduce ambiguity.

Has there been any resistance among the radiologists at Mass General to becoming more standardized in their practice patterns?

It’s an interesting thing: most radiologists would prefer to follow the standards, but they want to have the freedom to recognize and apply their own special insights into what’s going on with the patient and the specifics they see in the particular situation. So as we apply tools to help them, that we don’t make those tools restrictive or compromise that ability of radiologists to apply their specialized knowledge.

So what have you done?

So what we’ve done is, where radiologists do their work is where they’re creating the radiology report based on the images. So we’ve created a tool that sits inside the dictation system and report creation system they use, which at MGH is Nuance’s PowerScribe 360. And so that’s where radiologists live during the day. They have their PACS with the images in window, and then the report-writing system in another window. So we’ve embedded a tool for when the radiologist sees one of these incidental things for which there’s a guideline, the tool will pop up and give them a help. It’s called “Clinical Guidance” in PowerScribe 360. It implements “ACR Assist,” as the ACR calls it. It’s a concept. This is based on what they recommend. At the ACR level, it gets to the level of saying, when we create content, here’s the format we should use to do that. And that’s part of what Nuance has put together, is a format that allows essentially plug-ins for these clinical guidance modules. The first version was released last year with six modules and we’re demo-ing at least five more this year at RSNA, and at the end of next year, there will be a dozen more. There are dozens in the pipeline at ACR and we’re doing it at MGH. But there will be a standard format for putting these modules out, and eventually… Right now, it’s sort of a closed system, but eventually, there will be an architecture with which people will be able to bring into their own system. And the introduction of the cloud to support the Nuance solution makes this simpler, because new guidelines can simply be pushed out into the solution. And that’s obviously a big win. When the ACR updates a guideline, you don’t have to go out and create a new thing, then. You’ll just upgrade the software.

For those of us who are non-clinicians, can you provide a few examples?

One of the big ones would be recommending follow-up for pulmonary nodules. Using the tool, radiologists will do that a consistent way; they’ll make consistent recommendations in line with national guidelines for follow-up. And as an added benefit, the presentation will be consistent. Another one that’s important is adrenal nodules—renal masses, either cystic or solid. Incidental liver lesions seen on Adrenal, renal, liver, and ovarian cysts, those are four categories.

When did you go live with the initiative?

At MGH, we developed a prototype we started using in 2012, and then Nuance released their first product that incorporated this concept, just this past year, PowerScribe 3.0 came out with it in April. And the new version issued two weeks ago 3.5, is going to have more, and more advanced features. More content and more advanced features.

In a broader context, this initiative is a part of the shift towards transparent, accountable, more standardized care delivery, correct?

Yes, I think this is an important part of that. It’s an opportunity for radiologists, for certain specific things that come up in their practice, to have a standard that they can demonstrate they’ve applied a national standard, that when they make recommendations, they’re making recommendations out of national guidelines. Or recognizing that according to the guidelines, no explicit thing needed to be made. And it brings that level of transparency. And I think radiologists should be seeking to do that as much as possible, to have it clear that when they diagnose or categories something, they know it’s based on something clear.

Radiologists probably do practice more defensively than other specialists, in the medical-legal context, perhaps?

Yes, I would say so. And we need to be able to justify our clinical decisions. As in, for example, a situation involving a concussion.

And in the context of clinical informatics development, this is an example of a meatier, deeper plunge into the weeds of clinical decision support for a specific specialty.

Yes, this is the manifestation of that in the context of radiology. A lot of the guidelines and clinical decision support being provided for other specialists fits within an electronic medical record system or ordering system, but that’s not where the radiologist lives; we had to bring it to where radiologists live. But also, this fits into the context of bringing them usable clinical data and bringing it into the patient’s clinical data stream, to feed other streams. Clinicians are creating structured data that can be used downstream to feed a downstream decision support system. For example, if a radiologist reports a benign lesion and that feeds into the EHR, the fact that the patient had that benign lesion, is now incorporate for the future, for other radiologists and other clinicians to see it in the future. And that may be a clue for something that comes up later on. For example, a radiologist may see a nodule in a patient’s adrenal gland. And their first concern is, is this cancer, or not? A certain percentage will be a metastasis from a patient’s lung cancer. But if you look at it and see that it has certain properties that are benign, you can note it. Or six months later, if it hasn’t grown, it’s probably not a cancer. But some of the adrenal things that are benign that a radiologist sees, they can be hormone-secreting benign things. So if you had some diagnostic conundrum—not an imaging conundrum, but why does this patient have high blood pressure or something, created by one of these adenomas, then it’s already locked down in the structured data, and is more available for someone else to recognize later, and that is an additional benefit to this kind of initiative.