Measuring rad performance: ‘SWIM’ now or sink later

Picture this. You’re in charge of medical imaging at a practice employing numerous radiologists across several sites. Your boss knocks on your office door. Comes in and asks you for a quick snapshot of practice-wide performance. Wants to know report turnaround times by radiologist and by site. Says, “That’s really all I want. It’s not a big deal, right?”

“Not a big deal at all,” you reply, figuring all you need to do is look at timestamps to find out when images were ready for reading and when they were reported. Subtract the difference and presto. You’ll have that info for your boss.

A short while later, you dig in. You begin to build a simple spreadsheet, figuring you’ll use session numbers to keep individual exams straight and to check which radiologist read and reported each one. Then you go to collect and integrate the data.

Only then do you realize that the data are locked up in silos. Each of your sites has its own software, systems and processes for capturing and dispensing the info you need. Your little performance-check project isn’t going to be so easy after all. In fact it’s looking to become the source of a serious headache.  

With that scenario, Marc Kohli, MD, launched into a forward-looking discussion at last fall’s RSNA meeting in Chicago.

“What can we do to make all this easier?” he asked before answering his own question by describing a blossoming project led by SIIM, the Society for Imaging Informatics in Medicine. “They got together with some IT folks and some really smart physicians and created SWIM—the SIIM Workflow Initiative In Medicine.”

Normalizing nomenclature

SWIM has since been integrated with RadLex, the vocabulary RSNA has developed for use throughout the radiology profession, added Kohli, a radiologist at Indiana University Health and a member of RSNA’s Radiology Informatics Committee.

Toward the top of his talk, which was titled “Measuring Your Department with the SWIM Lexicon,” Kohli explained that SWIM combines two highly useful components. The first is the lexicon, a controlled vocabulary subset that defines exactly what a workflow step means. The second is a mechanism for HTTP data exchange, which allows users to use a browser to retrieve data germane to given workflow steps from a SWIM-enabled RIS or a SWIM-enabled PACs.

“One of the nice things about SWIM being incorporated into RadLex is that you get all of the benefits of an ontology,” said Kohli. “So you get things like synonyms, relationships to other terms and a way to understand relationships between terms programmatically.”

It’s actually easier shown than described. Go to bioportal.bioontology.org and you can see, for example, how SWIM works interchangeably with RID-45924—the preferred numeric name for the Final report approved workflow step—and RID45812, or Reported.

“The ontology gives you that framework and that understanding of how the different terms might fit together,” said Kohli. “If you had a SWIM-enabled RIS, with an easy RIS-based HTTP operation, you could specify an accession number and it will give you back all of the information that pertains to that exam in a nifty little JSON format. You can also ask things like, ‘Tell me all of the exam info between this date and this date.’ You can ask for a date range and get back a whole bunch of exams instead of just one exam.”

And wouldn’t that impress the boss?

Lexicon of laudability

In a perfect world, all RIS and PACS systems would be SWIM-enabled, said Kohli. Such widespread adoption would keep seekers of measurement data from having to do normalization for workflow terms, as that would be done automatically. It’s a world worth working toward, Kohli suggested. 

“This is one of those things where we need your help to ask the vendors to implement this when they’re building new tools,” he said. “You can’t buy one of these things today, so why am I standing up here talking to you about it?” His answer, in so many words: to not only encourage such engagement with vendors but also to urge anyone building a data warehouse in-house to download and integrate the SWIM lexicon into that data warehouse.

“Then, when we have analytical platforms that can pull things from something that’s SWIM-enabled, you will already be ready to go,” he said. 

Kohli then pointed out that he’d begun his presentation with a turnaround-time story because that’s the most basic case to consider when measuring the performance of a radiology practice. He showed how he went through the SWIM lexicon looking for terms that might pertain to something more complicated. When he keyed in CT scanner utilization, up came a thick stew of words, terms and general verbal confusion.

“You can imagine that this list would mushroom into a huge number of unmanageable terms,” he said. “It will be very difficult to just find out how well you’re utilizing your CT scanner.

Kohli’s bottom line: “As our complexity skyrockets—as the complexity of the questions we are trying to ask skyrockets—we’re going to need this standard even more than we need it today.”

“Why should you care?” he asked the audience. “You just want to make more RVUs, good case/next case. Why should you care about this? Is it just because Dr. Kohli at RSNA said you should do the right thing and this is a good thing to do?”

He then made his case, emphasizing that no actionable analytics are possible without good, clean data—and the SWIM lexicon can make for a short, highly cost-effective path to good, clean data.

Then too, as radiology shifts to new payment models, he said, it will need to quantify the value it provides the healthcare system—from the time imaging was first considered until results were delivered and discussed with patients and referring physicians—or else.

“You can apply the SWIM lexicon to make it easier to measure some of the output that we have and some of the value that we provide,” he said.

“I hope you realize that what we’re talking about would have far-reaching implications for the entire imaging chain,” concluded Kohli. “If you want to do things like predictive analytics and business intelligence and reporting, you really need to get this piece nailed down.”