Clinical analytics for radiology practices, in its current form, is defined by nothing so much as its limitations, according to Benjamin Strong, MD, CMO of Virtual Radiologic (vRad). “Practices don’t have the kinds of actionable insights they should, in terms of study characteristics, study mix, referring patterns, workflow, RVUs, and so on,” Strong says. “It’s something with which practices are struggling, if they’re even aware that they don’t have the analytics they need. We were managing blind during the good old days, and that continued for so long that many people don’t even miss analytics.”
With the current economic environment (as well as the federal push toward new payment and care-delivery models), however, practices are working on thinner margins than ever before—and they need data to substantiate everything from operational efficiency to outcomes. “However the changes we’re seeing now wind up manifesting themselves in the future, there is no question that clinical analytics will contribute positively to participation in any kind of organization that is based on cost containment and improved outcomes,” Strong says. “This represents a dramatic departure from how radiology practices have functioned in the past. You can’t get paid for quality if you can’t prove it.”
Laying the Groundwork
At vRad, Strong has been part of an effort to lay the groundwork for better clinical analytics. “We’ve definitely suffered, over the years, from the fact that our platform, like those in use by many practices, was not built for extracting actionable insights,” he notes. “Our system was built to aggregate studies from more than 2,000 different locations each night and get the study to the right radiologist in a timely fashion. Anything we really wanted to analyze required a manual data pull, and half the time, you’d realize the data didn’t answer the question you’d wanted them to, and then you’d be back at square one.”
To address the issue, going forward, the company is rolling out a new approach to normalizing the studies that it receives. The patent-pending innovation includes a component similar to a vehicle identification number (VIN), in which separate digits in the number assigned to a car stand for discrete pieces of information about it (the engine type, the plant where it was built, the model, and so on). The format is readable to the naked eye and will be used to orchestrate study movement; this VIN of radiology is just one of 23 separate study attributes included in the new system. Strong says, “The normalized data are accessible via intuitive extraction layer, so we can easily produce actionable insights.”
The 23 discrete study attributes, coupled with natural language processing (NLP), make possible unparalleled access to information about radiology outcomes, Strong says. “If you can review radiology reports for context, then that further enables you to analyze things such as the ordering behaviors of referring clinicians and the percent positivity of any given site,” he notes. “You can determine whether the report is positive or negative for a pertinent finding—which was something only a radiologist could do before, and you’re not going to pay a radiologist to do that. NLP allows you to pull reports for pertinent pathology, and now, you have information that could affect the practice of radiology—and even medicine.”
Powerful Potential
As applied to quality assurance (QA), analytics can further refine radiologists’ workflow for maximum efficiency and optimal outcomes, Strong says. “Your traditional quality oversight in a typical, smaller radiology practice varies, and objectivity is sometimes affected by the nature of the partnership model and size of group,” he says. “When you can use analytics and text mining on the QA database, you open up incredible vistas of quality investigation.”
This innovation could redefine how radiologists practice by putting increasing focus on the viability of the report: “Reports are about more than just interpretive accuracy,” Strong notes. “With this kind of analytics capability, we’ll be able to see the relationship between the structure and thoroughness of a report—including factors such as vague language or excessive use of disclaimers—and the presence or absence of discrepancies.”
Clinical analytics produced by radiology practices also holds the potential to change the behavior of referring clinicians, over time. “With data converted to actionable insights, you can really start to analyze the potential effect of ordering a certain study, in a certain clinical presentation, on patient care and outcomes,” Strong says. “The information has the potential to alter the ordering behaviors of physicians, based on the expected contribution of any given study to the desired outcome. It is a different approach from decision support.”
With appropriateness and utilization of imaging more in the spotlight than ever before, that information could prove invaluable to practices in demonstrating their value under new payment mechanisms, such as accountable-care organizations (ACOs). “Resource utilization, expected percent positivity, and outcomes are all going to be things that ACOs—as well as any structured-care system—are going to be interested in,” Strong says. “In terms of radiology practices’ contributions to new payment and delivery systems, we see incredible potential. Being viewed as a cost center in an ACO is a bad thing. Analytics will give radiology a seat at the table. ”
Cat Vasko is editor of Medical Imaging Review and associate editor of Radiology Business Journal.