When a Medicare administrative contractor found itself under the gun to improve its performance in order to retain its business with the Centers for Medicare & Medicaid Services, it was able to leverage semantic technology to realize its goals.
The audience at last week’s Semantic Technology & Business Conference heard about this case study from management consultancy Blue Slate Solutions, which specializes in improving and transforming operations through business process and analytic transformation and which spearheaded the semantic project for the contractor. “This client was struggling in the medical review process, to decide if claims should be paid,” recounted David S. Read, Blue Slate CTO.
With billions of claims flowing in, it’s impossible to manually medically review each one, as the registered nurses who handle the job have to bring together all the records related to the claim. That’s a time-consuming and costly process to go through when there are no problems with a claim, so the goal is to determine ahead of time which ones actually might have defects and require a more thorough review. The client, before the engagement, had only a 30 percent denial rate, which meant that nurses were spending 70 percent of their time looking at claims where nothing was wrong.
The contractor, according to Blue Slate, also was slow to react to integrate new information into the review process. It took too long to discover if filters set up in the claims system via rules to identify things that could point to problematic claims were actually effective in finding claims to be denied. This was because data traveled from someone’s handwritten notes to people who tried to code that text as definitions into different systems as rules, but the definitions didn’t always agree. There was no structured way to represent the added definition.
It also wasn’t doing well at telling a provider who’d made a mistake what that mistake was that resulted in the claim being denied, so the original defects were repeated when the claim was resubmitted. The problem is there was no easy way to modify the interface the nurse used to record the decision about what was wrong or missing with the claim, and the review details related to it were not collected, either, stumping the providers about how to fix things. Analytics of its data was an issue too, with multiple systems each having parts of the story, but no unified view of it all.
Semantic technology brought much-need flexibility into the picture, providing the flexibility of the data model for looking at different types of claims dealing with different situations and relating different checklists to them; making it possible to aggregate data from multiple systems; and to build triples from the graph by tying together relationships between things like a nurse’s reason for denying a claim and the instructions for the provider to address that, so as to avoid repeat mistakes. And, with the collection of more granular data around the review process, the door opens to better data mining and analytics. No systems were replaced, but the client now has a new component to federate and add agility around the environment.
“It’s easy to add new infrastructure into a semantic environment,” he noted. The system went live in January and at the nine-month mark, results came in about whether it did enough to keep the contractor from losing its contract. Among the improvements that have enabled the client to keep its contract are a 370 percent improvement in hypothesis testing (that is, whether something will be found on a claim) when it needed only a 250 percent improvement. Said Read, “The client drastically changed the definition of its medical review operation through the fact it can federate data and add a lot of detail to it in real-time.”