To understand why the future of radiology is in reporting that is both structured and template-based, look to the earliest days of the profession.
One day in 1896, William James Morton, MD, of the New York Post-Graduate Medical School and Hospital took fountain pen to paper and scrawled:
“The X ray shows plainly that there is no stone of an appreciable size in the kidney. I only got the negative today and could not therefore report earlier. The picture is not so strong as I would like, but it is strong enough to differentiate the parts.”
“So you had workflow problems, and a little bit of a hedge, even in 1896. Some things never change,” Curtis P. Langlotz, MD, PhD, professor of radiology and biomedical informatics at Stanford, quips.
Langlotz was one of three presenters of a talk titled, “The RSNA Reporting Initiative: Developing a Library of Best Practices Radiology Report Templates,” at the annual meeting of the Radiological Society of North America on Dec. 2 in Chicago. Specifically, the session presented tools and resources that the initiative makes available to radiologists who want to tap into RSNA’s growing library of best–practice-based report templates.
Langlotz: The actionable narrative
“Even though we use a lot of fancy technology, speech recognition and so on, fundamentally what we produce is a narrative,” Langlotz says. “That really has not changed much.” Nonetheless, reporting methods are going to have to continue to evolve to keep up with changing payment models, Imaging 3.0 and all of the other change drivers in healthcare, he adds.
In the current environment, structured reporting, the method of verbally characterizing diagnostic images in a way that combines a fixed format with consistent organization and standard language—ideally one that incorporates checklists and produces reports that are readable to machines as well as humans—is gaining disciples.
The RSNA Radiology Reporting Initiative got off the ground in 2007, when the ACR invited leaders of radiology subspecialty organizations to a summit on reporting. A consensus emerged: Structured reporting is an optimal method of creating reports—as long as tools exist to generate structured reports just as fast as narrative ones have been created through the years.
Attendees of that ACR summit also agreed, Langlotz says, that one of the best ways to achieve this is to combine structured reporting elements with voice-to-text software that incorporates macros, the one- or two-word prompts that automatically populate a given text field with full sentences or paragraphs frequently spoken in radiology reporting. “There was then this notion that radiology professional organizations should create a repository of exemplary reports,” Langlotz says. “That’s what this RSNA report library is intended to provide.”
Today, RSNA’s best-practices template library—online for free and unrestricted use at RadReport.org—offers radiologists select, macros-based templates largely created by experts, subspecialty radiology societies and leading radiology departments. It also links to an open template library (open.radreport.org) in which radiologists can share their own templates and critique others’ templates using comments and a starred rating system à la Amazon and Yelp.
“Over the past couple of years we have worked on standards that enable templates to be exchanged from libraries and exchanged between systems [and sites],” Langlotz says. “These are raising the bar for the vendors and giving us more capabilities in our reporting systems.”
A new IHE reporting protocol
Langlotz described a new reporting protocol to facilitate such template exchanges by defining templates using an HTML5-based, Internet-friendly format. Called Management of Radiology Report Templates (MRRT), the protocol was developed by Integrating the Healthcare Enterprise (IHE) in 2013 and will soon be the standard profile of the RSNA template library. “This is something that you should consider having a conversation with your reporting vendor about,” Langlotz says. “It will make for a better report and, in fact, make the radiologist even more productive.” (See sidebar)
“To close with an elevator story for the RSNA template library and for structured reporting, what are the benefits?” Langlotz asks, and then rattles off a few. “They can help a practice that wants to get started with structured reporting agree on a consistent format using standard headings and sections. They can help you create reports your referring clinicians prefer. There’s some good literature to show that clinicians prefer a report that’s organized into sections and subsections. There’s a lot of research to show that, if you are using a speech recognition system and you’re starting with a blank page, you’re really much less efficient than someone who has macros prebuilt.”
Langlotz touched on the concept of template “nesting,” the idea being that templates can incorporate levels of modularity that allow them to automatically “fire” inside other templates with alerts, reminders and such. “Then there’s the ability to annotate the template using standard terms, which allows us to link the radiology report to other parts of the electronic medical record.”
He also pointed to CMS’s Physician Quality Reporting System (PQRS) and other incentive programs. “If you have a template that includes the required language in there, that makes it much more likely that your report will be compliant,” he says, enabling the practice to avoid penalties of between 4% and 6% of Medicare payments in 2015. “So there are certainly good benefits in there for financial folks, there are good benefits in there for quality folks and there are good benefits in there just from a radiologist’s perspective.”
Kahn: We’re just getting rolling
Charles Kahn, MD, MS, a radiologist at the University of Pennsylvania and current chair of the RSNA reporting committee, made the second presentation. Among the sub-topics he discussed were milestones reached in the RSNA template library—beginning with some impressive numbers—and openings for collaboration beyond the U.S.
The template library now contains some 268 templates, Kahn reports. “We’ve had more than 106 million views and downloads of templates from RadReport.org,” he says. “We released the very first group of templates at this [RSNA] meeting in 2009, and we didn’t really start counting until about a year later, so it’s really quite remarkable.”
Meanwhile, most of the templates are in English, although a handful is in other languages, including Turkish and Mandarin Chinese. “We have been working with the European Society of Radiology for several years, but now it is joining us in this effort,” Kahn reports. “We are looking at translating the templates into a variety of European languages and bringing them into use there.”
Kahn emphasizes that the templates in the RSNA library are offered as select, best-practices models, not use-these-or-else forms. They have been coded with RadLex terminology, the vocabulary RSNA has developed to facilitate normalized nomenclature throughout the profession and, eventually, the EHR, although a radiologist accessing a template generally won’t see RadLex at work. “It is [embedded] to help improve the structure of the report,” he explains, “but it’s not something that you as a radiologist have to see or deal with in any way.”
One of the most significant advances over the past year has involved DICOM, the standard for medical image information. That is the development of Supplement 155, Kahn says, which takes the information that comes from MRRT-based templates and sends that knowledge from a reporting system to an EHR system. “In other words, Supplement 155 dictates how you get information out of your reporting system and up to your EHR, including the structured and coded content.”
Kahn spent considerable time expanding on Langlotz’s introduction of MRRT, along the way showcasing “T-Rex,” the template editor developed for the RSNA reporting initiative by a Waterloo, Canada-based software solutions vendor. T-Rex, he says, is still under development but very effective. “We are just beginning to start using it, and we think it’s going to be a very useful tool, allowing people to contribute templates in the MRRT profile.”
The combo allows a reporting system to pick up a template at the start of a dictation session. “The information then flows out into an electronic health record, and then there’s a tool that lets you extract the coded information from the electronic health record,” he explains. “It’s the full trip—from the creation of the template, to its use, to extracting the value that it allows you to encode. That’s really where we have sought to go with this [template library] effort.”
Turning to new opportunities involving the template library, Kahn describes collaborations in the works between RSNA and the ACR, Cancer Care Ontario, the European Society of Radiology and the College of American Pathologists.
The latter, for example, has done a good deal of work to create what they call synoptic reporting. “It’s very similar to our efforts in structured reporting, but they’ve put together cancer checklists for specific diagnoses,” Kahn says, adding that each diagnosis has a specific checklist that every pathologist has to fill out as part of the reporting process on that examination. “That information yields a huge database of information that they have about these patients, and we’re looking at ways that we could integrate what we’re doing with what they are doing. We have sort of the radiologic-pathologic correlation of imaging and histology information, so we are moving forward on that.”
Harking back to the reasons for RSNA’s launch of the radiology reporting initiative, he says one driver was the fact that only 2% of the patients who are eligible to participate in clinical trials do so. “There’s this vast, untapped population that could become part of a clinical knowledge base,” he says. “But having some way to capture information about those patients has been challenging if you’re not entering it into some controlled form. Doing so in the span of a normal clinical examination provides an opportunity to add that information into basically a database to have a registry of all the information about these imaging examinations.”
The key parts of the RSNA radiology reporting initiative, Kahn reiterates, are the report template library, the MRRT profile, the DICOM supplement and the T Rex template. “I hope you all have a chance to take a look at them,” he says. “We want to help people get a consistent format, get things that our referring physicians prefer, maximize our efficiency, reduce risk of errors, promote compliance with accreditation and certification requirements and profit from federal performance incentives where possible.”
Heilbrun: The proof is in the rewarded results
Marta Heilbrun, MD, a radiologist at the University of Utah and a co-leader of RSNA’s reporting initiative, provided the final talk, about how her institution has been drawing from the RSNA template library to measurably increase quality—and earn incentives. “The PQRS guidelines are actually driving our template use,” she says. “We’ve created a PQRS-compliant template that auto-launches with the exam.”
Heilbrun says the effort got a push to arrive at that point when administration sent the neuroradiology section a reminder that they weren’t doing so well meeting the PQRS criteria when interpreting carotid vascular studies. The criteria are based on measurement requirements established in the American Symptomatic Carotid Endarterectomy Trial (NASCET).
“The email said, ‘Hey, remember, you need to report any time you do these studies that your reporting is by the NASCET criteria.’ We went from 6% to 12% with that.” Neuro leaders at a follow-up meeting stressed how important further improvement would prove to the section’s bottom line and, soon enough, compliance rose to 60%.
“Then we used a template that auto-launches on every single case—and we jumped up to 90%,” Heilbrun reports. “We have really made it—the structured reporting and the template reporting has made it—so easy for us to get the information into the report. Then again, we may or may not think that this, the NASCET criteria, is critical for being a good radiologist—but it is critical for what we’re being measured on right now.”
Bring crowdsourcing to bear
Heilbrun urges radiologists to contribute to the open template library at open.radreport.org and bring crowdsourcing to bear on structured reporting. “We really want to learn what the RSNA membership does with structured reporting,” she says. “What are your current templates? What’s out there? Are there elements that every single one of us is doing? Can we exchange statements [from other medical societies] in a way that is easy and usable for all of us? Do we then also point to different decision-support tools?”
“I suspect that a lot of us are doing very similar things,” she adds. “Now, we can really start saying that this is valuable, and this is what we do, and this is our service. This is what a radiologist provides.”
Now, with MRRT, the profession has in its hands the ability to exchange critical, coded data. “We’re at this place where, when you work and you create a template and you create a report, you can report mineable data,” Heilbrun says. “We can then build decision-support around that. Right now it’s really hard to figure out what is the percentage of positive and negative studies for a given indication—and to then demonstrate that the radiologist’s interaction with that study is the right thing for helping the patient.”
Heilbrun closed by highlighting the potential for RSNA’s Radiology Reporting Initiative to improve the state of radiology education. Some will object to the structured reporting process, she says, insisting that it will turn residents into “box checkers” who don’t speak their findings.
“Are they not then learning and internalizing the vision that they’re seeing in making that observation into something that is actionable?” Heilbrun asks. “Are we losing that step? Well, I would say that we may be losing some of that step.”
“At the same time,” she continues, “if our reports are measurable, we can actually look at our residents and say, ‘You know, you tend to over-diagnose in this way. Your colleagues and your peers have, for the same set of studies, only this many positive studies, or this many negative studies or this many recommendations.’ From that the faculty can then [make adjustments]. I think we’re actually in a place where this is going to allow us to better teach. We’re going to be able to teach with evidence.”
Poetry or prose?
In the question-and-answer period following the presentations, one radiologist voiced two objections, the first of which has inhibited the adoption of structured reporting to date: “Some people are very strongly opposed to structured reporting because they feel it violates the aesthetics of the report. If you have to choose between aesthetics and efficiency, what message would you give to your radiologists that would induce them to use a structured format reporting?” While it’s great to have a checklist and a structured format, he adds, “then how do you ensure that the key findings are appropriately highlighted?”
“People have written about this, and there’s that tension between prose versus poetry,” Kahn replies. “Most of us are capable of composing doggerel rather than good poetry. Getting people to write well in radiology reports, to convey their thoughts succinctly and convincingly is challenging. Some people are very adept at doing it; some people communicate very well. I tell my residents that what I want is the findings section to be itemized. I tell people that it’s also kind of a journalistic exercise, because it’s the inverted pyramid: If it bleeds, it leads. You talk about the most important things first. You put in the critical findings; you don’t start with the gallstones, you start with the ruptured aortic aneurysm.”
“These are exactly the things we need to do, and that’s where people can knit together the story of what’s happening with the patient in a way that’s useful and informative to our referring colleagues,” Kahn concludes. “I don’t think those have to be necessarily in opposition. I think you can have both ways of approaching the data.”