CMO/CIO forecast for 2016: New breakthroughs in radiologist workflow optimization

With the possible exception of diffusion-tensor MRI for concussion in athletes, no clinical imaging procedure stands a chance of bumping information technology—with its attendant ripple effects on regulatory compliance and business performance—off radiology’s figurative front page in 2016.

For starters, watch for the continued general blossoming and fine-tuning of medical data normalization, natural language processing, auto-formatting and deep machine learning.

Targeted analytics of big data will guide the implementation of ICD-10 decision-making, and universal formatting will enable everything from identifying internal inconsistencies in reports to flagging omitted language that’s required for reimbursement or PQRS reporting.

Not only will all of these IT advances keep getting better, they’ll also intelligently integrate into radiologists’ normal daily workflows. 

At least, that’s what you’ll see if you keep an eye on vRad in 2016.

“Within healthcare, everyone’s biggest challenge has been, and will continue to be, integrating software systems developed outside our industry into a radiology workflow in a practical fashion that has immediate effects—here, now, today,” vRad’s Chief Medical Officer, Benjamin W. Strong, MD, told Medical Imaging Review in a recent review-and-outlook interview.

Strong stresses that the nation’s largest provider of radiology services is pursuing the continued development of the aforementioned technologies “from a very pragmatic standpoint.” He adds: “How does each technology help us serve our clients right now? And how can we use it to help us serve them better in the future?”

The digitalization of the profession at MEDNAX affiliate vRad, which employs nearly 400 radiologists, isn’t always about efficiency, adds vRad CIO Shannon Werb. “It’s also about making sure that we create the right environment for patient-facing physicians to provide the best quality care they can,” he says.

Phenomenal flexibility

As U.S. healthcare continues transitioning from volume-based to value-based reimbursement in 2016 and beyond—most notably with government initiatives dialing up pressure on providers to demonstrate quality and safety—Strong and Werb are of one mind on how to help those patient-facing physicians stay ahead of the curve: Equip vRad’s radiologists with technological tools that all but self-deploy.

“Reducing our turnaround times, improving our patient care and increasing the satisfaction of our clients—across the board, that’s always what we are looking to do,” says Strong. “It’s a case of our rising tide lifting all their boats.”

To give an example of a specific application in action to this effect, Strong points to vRad’s implementation of a patent-pending customized structured report formatting program entitled rScriptor by Scriptor Software. Highly customized for vRad’s unique and patented workflow, each vRad report is created in real-time using line-by-line dictation and follows a consistent template that includes Exam, Technique, Comparison, Findings and Impression. This he describes as “universal” because it’s used across the massive practice, and with big buy-in from the rads, at that. 

The custom formatting program “uses natural language processing for the precise identification of phrases,” he says. “Recommendations for a lung nodule, grading of trauma injuries, categorizing osteochondrosis defects—all of those things are built in so as to give real-time warnings to the radiologist (upon dictation).”

“The flexibility is astonishing,” he adds. “We can build in financial or reimbursement requirements, reporting accuracy and thoroughness requirements—really just about anything where we need the radiologist’s constant vigilance and constantly current knowledge of best practices to be applied.”

Strong says implementation of the vRad customized rScriptor formatting program is directly creditable for a major fiscal achievement in 2015: It lowered the practice’s own reimbursement refusals by significant double digits.

Big data is only the beginning

Looking back on the technology-enabled breakthroughs that contributed to vRad’s current market position, Strong and Werb note that the big gains began taking shape in 2013, when the practice recognized that it needed to work with normalized datasets.

“We apply normalization much more broadly now than then,” Werb says. “But we started at the procedure level, and that enabled us to make headway toward ICD-10. Once we had an underlying common set of procedures that we can build capabilities around, we then could get into things like laterality and contrast and other standardized metadata.”

Werb says an underlying common theme is the criticality of such metadata to vRad’s optimization and automation of deliverable services.  

He also says that moving forward with the more advanced applications is only possible because of vRad’s volume of readily available data: over 5 million studies read annually.

Then again, it’s also a matter of doing the right things with the big data.

“Data is essential, but in order to leverage deep learning or artificial intelligence in images, you must have, for example, labeled data,” Werb said. “It might be tempting to say, ‘The radiologists reading the studies can label the data.’ But until you have a normalized data set—along with standard procedures and a formatting program that is structuring the output and leveraging natural language processing—you can’t effectively and confidently create a highly structured and highly labeled data set to go take the next step of actually training the technology to, for example, automatically recognize critical conditions.”

Not surprisingly, the technology developments and fine-tunings don’t rise in separate, discrete projects. Instead, they augment and extend one another.

“We started with data normalization, then we built upon it every step of the way,” Werb says. “And we have plans to continue building on it in 2016.”

Practices under pressure

As for diffusion-tensor MRI for athletic head trauma, Strong says it’s the only diagnostic imaging procedure to register as on the rise in vRad’s aggregated volume tabulations—and the increase has been decidedly modest. However, the practice’s overall volume is up significantly.

Based on the gains, vRad is projecting double digit growth overall, with about equal lifts from new and existing clients, in 2016.

“An increase of that size is unlikely to result from simple shifts of volume between on-site and teleradiology,” he says. “It suggests an actual increase in utilization nationwide.”

Strong further predicts that current reimbursement rates will hold for another year or two.

“We are definitely seeing diminishing pressure on reimbursement,” he says. “There were big stepwise decreases in reimbursement for radiology services over the last several years, and, over the last one to two years, we have really seen that begin to plateau.”

That’s the good news for the profession as a whole. The not-so-good news is the fiscal squeeze hospitals have been putting on smaller, midsize and even relatively large practices.

“I get more and more calls about consolidations, about loss of hospital contracts, about a hospital desiring contract renegotiation and hospitals insisting on employed status for radiologists, which no radiologist wants,” he says. “That has really increased in frequency over the last year, and I get the sense that the pace is accelerating as we head toward and into 2016.”

CDS is delayed, but why wait?

Turning back to information technology—or, more to the point, to challenges that may slow its transformation of the practice of radiology—Strong and Werb state that vRad will continue to invest heavily in applications that will prove game-changers.

“We are already working on deep learning and image-based analytics, where we are training the computer to visually recognize critical conditions and move the study up the worklist,” Werb says. “That will be an important part of our investments in 2016. The challenge is that this is all very new. A lot of people are talking about these things, both in healthcare and outside of it, but nobody is really doing them.”

It’s a challenge he and Strong believe vRad can meet by building on the momentum and capabilities the practice has already developed.

Werb singles out clinical decision support as a particular area they will continue working hard to perfect, even though Medicare has delayed its CDS mandate originally slated to take effect in January 2017.

“CDS will play on top of a lot of the metadata work that we have already been completing with data normalization and ICD-10,” Werb says. “We’re doing a lot of the CDS work up front so we can get our customers ready.”

Strong says it’s no secret that integration of exciting IT advances with existing radiology workflows has been, and will remain, a challenge for most radiology practices. 

“You can’t be adding log-ins and applications to the radiologist’s workflow,” he says. “Too many people are adding applications that aren’t thoroughly integrated with the workflow. The technology ends up becoming a distraction or a detriment rather than a true contributor.”

vRad’s software engineers have ensured that technology enhances rather than detracts from radiologist workflow. “They make sure we build everything with extensive integration into our workflow,” Strong says. “In fact, we don’t set out on the development of technology until we know how it will be integrated and how it will affect our workflow.  Our innovations are geared to maximizing the amount of time our physicians can keep their eyes on the images – without unnecessary administrative distractions.  It’s about letting doctors be doctors.”

Technology, there for the tapping

As 2016 hurtles into clearer view, radiology faces challenges of various kinds—and most of them can be met with technology, Strong maintains.

Going forward, the radiologist’s ultimate goal is “providing a quality service on the front end and then, on the back end, objectively quantifying that provision of quality service,” he says. “Technology can and should be used for both. Data needs to be normalized. Study reports need to be formatted. Study report content needs to be mined using natural language processing and similar tools. That is how you ultimately provide a quality service and document it.”

Is the profession of radiology as a whole up to the challenge?

“I have a great deal of optimism over the potential of practices that adopt the technology-enabled model we’ve been talking about,” Strong replies. “Our profession’s hope needs to arise  from our understanding that delivering the best possible care—meeting fast turnaround times while providing accurate interpretations, specialty interpretations and demonstrable overall quality—now means using technology both to deliver quality patient care and to report on it in a structured  way.”