Highmark maximizes Medicare revenues with SAS® Enterprise Miner

Health insurance companies that participate in the federal Medicare Advantage program are reimbursed based on the level of illness in the population they cover, which is measured using diagnosis codes. But what happens when a patient has an illness that a hospital or physician’s office doesn’t code properly? Or a member does not seek services for that condition in a particular year? Insurers risk losing thousands of dollars per case. With SAS Enterprise Miner, Highmark dramatically reduces that risk.
  
Highmark relies on SAS Enterprise Miner to build decision trees that map likely outcomes a patient faces based on measures such as symptoms, health history and demographics. Called the CMS Reimbursement Model, the application looks for patients who are missing appropriate diagnoses for up to 27 condition hierarchies and prioritizes them based on risk. An accurate diagnosis for each patient is critical because the reimbursement received from the Centers for Medicare & Medicaid Services (CMS) is contingent upon the diagnosis.

"For example, if we find that someone looks like they have Parkinson’s disease or chronic obstructive pulmonary disease, but the patient’s claims are missing the specific diagnosis code tied to the disorder, that will potentially cost us a lot of money," explains Brian Day, Highmark's Director of Advanced Analytics. "Using SAS, we can identify which cases have the highest probability of reimbursement potential. Our revenue management area would then submit a change and receive proper reimbursement."

Although Highmark could theoretically review every chart for missed diagnoses, it is important to effectively whittle down the number of cases that might require resubmission because the actual process is costly and time-consuming. “With SAS we don’t have to do as many chart reviews and that improves the bottom line,’’ Day says. “This program identifies millions of dollars in previously undetected medical risk each year.”

Here's how it works: The decision trees build predictive models that comb through reimbursement data looking for treatments and prescriptions that suggest a population of patients that have an undiagnosed condition, such as diabetes. In the diabetes example, some patients might have been diagnosed with diabetes in the past, but for unknown reasons their diagnosis is no longer on their records - even though they are insulin dependent or have other markers of the disease. By summing up the parts of the 27 disease hierarchy trees - each with hundreds of variables - patients receive a comprehensive, severity-weighted “possible diagnosis gap” score.

“We can be strategic in the way we identify underlying conditions,’’ Day says.
 
Choosing an in-house approach, rather than outsourcing
Day says there are numerous vendors that Highmark could outsource to in order to perform this type of analysis. “But they are just giving you lists, it’s all black box stuff,’’ Day says, explaining that the vendors don’t share how they find potential undiagnosed patients. “You can’t repeat it in-house because it’s proprietary. It is also very expensive,’’ Day says.

“With the capabilities we have from SAS, there is no reason to go with an outside vendor,’’ Day says. News about the work in Day’s group has spread. Highmark’s division in charge of its Medicare Advantage program now wants to move its condition management process internally, using SAS as a base for the analytics.

With SAS®, Highmark can also:
Mine data for marketing. The insurer searches the databases for members who have recently left a company that purchased insurance through Highmark. Using various attributes on file, Highmark creates a list of individuals that are good candidates for purchasing individual coverage. “We look at information like family structure and type of employment they’re in to come up with a mail campaign strategy,’’ Day says.

Develop sophisticated data mining skills easily. Day’s group recently transitioned to SAS®9, a process that went very smoothly. “The SAS experts have been nothing less than super. And the training helps my staff use SAS immediately.” Day himself had not data mined prior to using SAS, yet found it easy to pick up. “The gist of the process was very explainable. It’s just so nice and straightforward to work with.

“SAS helps us do some very sophisticated processes. If we didn’t have SAS we couldn’t come up with some of the answers we’ve gotten.’’

Copyright © SAS Institute Inc. All Rights Reserved.