Big Data Provides Tool to Shape Future of Radiology Payments

Researchers have interrogated the DRG database to come up with initial targets for help with negotiating bundled payments.

Data is fairly easy to aggregate in large amounts. The problem, according to David A. Rosman, MD, MBA, is making sense of it all. Rosman, a radiologist at Massachusetts General Hospital, made that point at a session he moderated on “The Future of Radiology Payments: Can Analytics Help Radiologists Regain Control,” on December 1, 2014, at the annual meeting of the Radiological Society of North America in Chicago.

Using the Ebola outbreak as an example, Rosman pointed out that cell phone usage in Africa has exploded over the last several years. (Rosman currently is in Rwanda, where he is creating a radiology residency program under a larger grant to build a medical infrastructure in that country.)

“People who can’t afford a meal a day are walking around with cell phones,” he said. “It’s quite stunning.” A huge amount of potentially useful data can be found on these phones if the texts, tweets and Facebook posts communicated through those phones are aggregated.

“All of that data is useless unless you ask the right question,” Rosman says. “But, what if you ask the right question—like how many of these mention the word ‘fever’ and ‘blood’ or ‘Ebola.’ You could focus care, you could know where the disease is creeping up and getting worse, because we can identify exactly what is happening with these cell phone records.”

How can a data analytics approach help radiology as a specialty, particularly as it begins to look at new payment and delivery systems? That’s one of the issues the Harvey L. Neiman Health Policy Institute (HPI) has been evaluating since it was established by the American College of Radiology in 2012.

Can analytics lead the way?

To put that question into context, presenter Richard Duszak, MD, HPI chief medical officer and senior research fellow, refers to a report from the National Commission on Physician Payment Reform (released in March 2013) that begins with this quote: “Our nation cannot control runaway spending without fundamentally changing how physicians are paid.”

“If you believe that our healthcare system is broken and that our payment systems are contributing to that, then you’ll probably agree with that,” Duszak says. “But, if you are a practicing physician, you’ll ask, ‘What does this mean for me?’”

Abdicating how radiologists get paid to people who may or may not know what radiologists do in clinical practice is a frightening proposition, he adds. “It’s important for all of us as practice leaders and practice radiologists to give this a lot of thought and come up with some meaningful approaches to what would happen to [radiologists if they are] pushed into some kind of bundled payment system,” he warns.

While physicians can expect the payment system to evolve from a fee-for-service to a fee-for-value world, how exactly they will get there is unclear. It could involve pay for performance, episodic bundled payments or population health management with large patient populations captured by health systems for long periods of time.

The focus of Duszak’s talk was on episodic bundled payments, which he acknowledges is one of the easier ones to model because there has been some precedents: Under Medicare, hospitals have historically received single bundled payments under the Inpatient Prospective Payment System based on Diagnosis Related Groups (DRGs).

One of the problems with the system, Duszak says, is that it “has created some perverse incentives.” For example, he explains, if a hospital gets paid  $10,000 for a particular admission that should require a five-night hospital stay, the hospital can bring in more money by releasing the patient in four days and flipping the bed to a new patient a day early.

On the other hand, physicians are generally paid on a fee-for-service arrangement based on actual services rendered as reported, using Current Procedural Terminology (CPT) codes. As Duszak points out, critics have suggested that this leads physicians to administer more tests and perform more studies.

“While the hospital may have incentives to get that patient out,” Duszak says, “in a true dollar sense, the more we image patients, the more we get paid. So, we have created these misaligned incentives in our systems.”

Making sense of big data

When attempting to divide a bundled payment among physicians, things get complicated fast, Duszak says. “That’s why a lot of these discussions [about bundled payments] fail, because no one knows how to do this in a fair and meaningful manner,” he says.

With that in mind, the Neiman Health Policy Institute is developing an inpatient imaging information application based on Medicare claims data—a 5% sample of data files from Medicare based on claims from 2009-2011.

What can a health system do with the app? Duszak used as an example the possibility of considering going to a bundled payment for DRG 193, “Simple Pneumonia and Pleurisy with Major Complications and Comorbidities,” in which the app was able to identify 17,000 admissions for that DRG over three years. Multiply that by 20 (based on the 5% sample size) and you have approximately 340,000 admissions. Going further, the app was able to determine that the mean amount of a radiologist’s professional payment for the DRG is $124.

“You need these kinds of pieces of information to get into the microeconomic analyses [of] your own individual institutions,” Duszak says. The problem, however, is the sheer number of DRGs.

“So where do I start?” he asks. “There are more than 700 DRGs, and they’re very heterogeneous with regard to specialist utilization, resource utilization and imaging utilization. How do I as a practice leader negotiate 700 different bundled payments? We’re trying to figure out how to leverage big data to help us make sense of all the noise.”

Prioritization is key

The short answer, Duszak says, is that there will be a need to prioritize.  “One of our more recent efforts has been to develop an evidence-based, prioritized strategic framework for identifying the encounters where your patient bundled-payment  modeling with be most impactful,” he says. “You’re not going to come right out of the gate and do all of [these DRGs] right now. I think you’ll be the exceptional person in this audience a year from now if you’re able to say you successfully negotiated one or two of these as a pilot demonstration.”

Which one should you pick?  What HPI has done, Duszak says, is use CMS data to secure claims data for a 5% random sample of all fee-for-service beneficiaries for 2011. Then HPI used Part A (hospital) claims data to categorize all inpatient encounters using Medicare Severity DRGs, and Part B (physicians services) claims data to identify the radiologist imaging services provided during those inpatient encounters.

Using that data, HPI then frequency ranked DRGs involving radiologist professional services in order to identify those services that were disproportionately associated with inpatient encounters, and also frequency ranked DRGs to identify those that disproportionately contributed to inpatient professional imaging spending.

HPI came up with 618,906 identifiable admission episodes, 430,707 (about 70%) of which involved the use of imaging professional services. Furthermore, that 70% was attributed to nearly all uniquely identifiable DRGs.

“So if you can just identify the 30% [of inpatient admissions] that didn’t involve imaging professional services, and figure out the methodology for that, you could ignore it and start weaning your list,” Duszak said.

Parsing the numbers for guidance

Looking at the numbers more closely, it became clear that a very small number of DRGs accounted for a disproportionately large amount of imaging (see Table), while a much larger number of DRGs accounted for a disproportionately small amount of imaging.

Specifically, just 5.5% of DRGs (41 out of 744) accounted for 50.3% of all hospitalizations involving medical imaging. In addition, these 4 DRGs—470 (Major Joint Replacement without Major Complications and Comorbidities), 871 (Septicemia or Severe Sepsis without Mechanical Ventilation 96+ hours with Major Complications and Comorbidities), 292 (Heart Failure and Shock With Complications and Comorbidities), and 194 (Simple Pneumonia & Pleurisy with Complications and Comorbidities)—accounted for 10 percent of all the inpatient episodes involving imaging.

“So, if you are in a blended system where most of the business is still fee-for-service and you’re doing some pilots in bundled payments, this is where you want to focus,” Duszak points out. “On the specific DRGs where the activity is most relevant.”

On the other hand, about two-thirds (500 out of 744) of DRGs accounted for just 9.6% of all hospitalizations involving medical imaging. “I’m not saying you should ignore these DRGs forever,” Duszak says. “But, if you are in a pilot project situation then these should absolutely be the last 500 you should consider about modeling, because they just don’t matter a whole lot.”

Professional services payments

When the focus turns to imaging professional services payments—the amount of money that radiologists, rather than hospitals, get paid for these imaging services—the results are almost exactly the same. Just 6.5% of DRGs (41 of 744) account for 50% of all inpatient imaging payments to radiologists, while two-thirds of DRGs accounted for just 10.6% of imaging payments.

“Here’s where you follow the money,” Duszak said. “If you pick those 40 DRGs you are looking at half of all the professional inpatient revenue that comes back to your practice. And if you ignore those 500, you are only ignoring 10% of your revenue.

“I’m not saying throw way 10% of your revenue,” he added. “I’m saying that if you are going to jump into bundled payments, this provides a framework for where you go first.”

What practice leaders should be thinking about when approaching bundled payments, Duszak said, are these simple formulas:

  • At the top: 5=50 (5% of DRGs represent 50% of payments and imaging episodes)
  • At the bottom: 500=10 (500 DRGs represents 10% of payments and imaging episodes)

Practice leaders also need to realize that when looking at inpatient imaging services, it’s completely impractical to focus on more than 700 DRGs. “I think that’s one reason why radiologists and other professionals have really gotten stuck when they’ve tried to do this,” Duszak notes. 

However, he concludes, getting adequate historical benchmark data and adopting a more practical approach to prioritizing services to model will facilitate bundled payment modeling for radiology practices.


Table. Top 20 DRGs Most Commonly Involving Medical Imaging

RankDRGDRG Short TitleEpisodes% Total*
1470

Major Joint Replacement w/o MCC

15,155

3.52%

2871

Septicemia or Severe Sepsis w/o MV 96+ Hours w/ MCC

15,053

7.01%

3292

Heart Failure & Shock W Cc

9,299

9.17%

4194

Simple Pneumonia & Pleurisy w/ CC

8,984

11.26%

5392

Esophagitis, Gastroent & Misc Digest Disorders w/o MCC

8,806

13.30%

6291

Heart Failure & Shock w/ MCC

8,456

15.27%

7690

Kidney & Urinary Tract Infections w/o MCC

7,842

17.09%

8190

Chronic Obstructive Pulmonary Disease w/ MCC

6,741

18.65%

9945

Rehabilitation w/ CC/MCC

6,364

20.13%

10683

Renal Failure w/ CC

6,247

21.58%

11193

Simple Pneumonia & Pleurisy w/ CC

5,973

22.97%

12191

Chronic Obstructive Pulmonary Disease w/ CC

5,935

24.34%

13641

Nutritional & Misc Metabolic Disorders w/o MCC

5,577

25.64%

14312

Syncope & Collapse

5,437

26.90%

15065

Intracranial Hemorrhage Or Cerebral Infarction w/ CC

5,071

28.08%

16872

Septicemia or Severe Sepsis w/o MCC 96+ Hours w/o MCC

4,940

29.23%

17189

Pulmonary Edema & Respiratory Failure

4,807

30.34%

18603

Cellulitis w/o MCC

4,753

31.45%

19682

Renal Failure w/ MCC

4,702

32.54%

20313

Chest Pain

4,698

33.63%=

Courtesy of Richard Duszak, MD, Harvey L. Neiman Health Policy Institute. *Cumulative percentage of ranked DRGs     

Michael Bassett is a contributing writer for Radiology Business Journal.

Michael Bassett,

Contributor

Trimed Popup
Trimed Popup