No Such Thing as Big Data in Health Care

December 4, 2014

Small-to-medium sized data in health care will help radiologists push for their piece of the reimbursement pie.

When it comes to big data, health care doesn’t really have any. And, for radiology, that’s a good thing. Small and medium data will work just fine – especially for testing and designing new reimbursement models, according to speakers at this year’s Radiological Society of North America (RSNA) meeting.

Industry experts at this year’s RSNA say the data hospitals and health care systems already have can help providers identify ways to maximize their influence in the design of any future payment models.

“We’re currently in the lowest life form of payment policy. We get paid for events – it’s a transactional delivery system,” said Richard Duszak, MD, vice chair for health policy and practice, department of radiology and imaging sciences, Emory University School of Medicine. “Increasingly, we’re moving to models where we’ll be paid by encounters and engagements.”

The question, he said, is how those models will be designed to ensure radiologists receive appropriate reimbursement for services rendered in a correctly incentivized way. To date, there’s no clear-cut answer, but there are steps radiologists can take – armed with small-to-medium data – to ensure their seat at the decision table.

Analyze Top Services
One of the most overlooked opportunities for making an impact is in analyzing claims data, said Duszak, who also serves as the Neiman Health Policy Institute (NHPI) chief medical officer. Delving into this information will bolster radiology’s position as health care moves toward new payment and delivery systems.

Based on Center for Medicare & Medicaid (CMS) statistics, 5.5% (41) of the more than 700 diagnoses-related groups (DRGs) account for 50.3% of hospitalizations that involve medical imaging. Individual hospitals and practices should concentrate on the top few DRGs to pilot test any bundled payment system, he said.

“In blended systems where we’re doing both fee-for-service and pilots of bundled systems, focus on the DRGs where the activity is most relevant,” Duszak said. “If you want to start somewhere that will evolve into a bundled payment, agree to bundled payments with the DRGs with the most admissions.”

To augment those efforts, the NHPI is also beta-testing a new app that would provide the same data analysis for all DRGs. Using 5% of CMS claims data from 2009 to 2011, the app, when ready, will allow providers to see average reimbursement rates for specific diagnoses and compare themselves to others.

The long-term goal, he said, is to create reimbursement benchmarks that will inform payment model discussions and decisions. In the short-term, these types of analyses will give radiologists the data they need to push for the appropriate amount of reimbursement.

Maximizing the Radiology Report
One of the most straight-forward ways radiologists can contribute to the switch from volume-based to value-based care is to improve their reports, said Woojin Kim, MD, a musculoskeletal radiologist at the Hospital of the University of Pennsylvania.

“Radiology reports are our main product,” he said. “And, data is becoming more and more important in the era where patients are getting our reports through their web portals.”

As a first step, he said, hospitals and practices should review their reports to identify any errors and pass along results to providers. He tried this tactic and informed providers of the mistakes that were slipping by. Within a month, he saw a 48% drop in errors.

Natural language processing (NLP) tools are also a fail-safe for identifying errors of laterality, gender, view, age, or contrast. The most effective NLP tools, he said, can also identify calls for follow-up visits, as well as alert providers when pathology reports are available. That way, radiologists will know if their diagnoses were correct or what they can learn from any mistakes.

The next step forward in controlling costs and increasing productivity, he said, is identifying ways to more broadly implement prescriptive analytics, a strategy that anticipates what, when, and why something will happen. Many physicians already do this. They identify and monitor symptoms, helping them prevent or minimize negative outcomes.

Ultimately, he said, you should choose an analytics or business intelligence tool that can manipulate your data to make it decipherable and actionable.

“Look for a business intelligence solution – whether you build it or buy it – that allows you to ask questions,” Kim said. “No business intelligence solution will have all the dashboard solutions you’ll ever need, but it must be interactive. Make sure you have a solution that allows you to do things yourself.”