Fraud and abuse detection software enters a new age
Fraud and abuse detection software enters a new age
System is expected to recover millions in Texas
You’ve heard it all before, no doubt. Providers often are advised to stay abreast of the tools the government is using to fight fraud, waste, and abuse.
But providers should take special note of one state’s solution to the problem. Texas recently unveiled its weapon against fraud and abuse: the Texas Medicaid Fraud and Abuse Detection System (MFADS). By the year 2000, the state-of-the-art MFADS is expected to recover about $14 million a year for the $8 billion Texas Medicaid program.
The strength of MFADS is that it compares information from databases that reside in different agencies, says Aurora LeBrun, associate commissioner for the state’s office of investigations and enforcement in Austin. "Previously, you could only look at one program at a time; now you can look at the relationship among programs."
For example, officials can compare data for a patient in a nursing facility program with data in the acute care program to see if someone is billing for this patient for home health services or acute care services that do not correlate with the patient’s status as a nursing facility program recipient.
"Not only do we use the Medicaid claims or encounter data from the Medicaid Management Information System (MMIS), but we include other things that typically would enhance detection efforts such as regulatory information, licensing information from state licensing boards for professionals, certification dates, and ownership information if it is available within the state agencies," explains Diane Davis, fraud and abuse systems and products manager with EDS’ state health care unit in Austin.
EDS, an information technology services provider, was selected by the Texas Health and Human Services Commission to create and implement MFADS.
HNC Software in San Diego and Intelligent Technologies Corp. in Austin are subcontractors to EDS and are providing additional services for the Texas project. Silicon Graphics in Mountain View, CA, provided the platform, an Origin2000 server.
In addition, the MFADS database contains a variety of information that is critical to detection, such as claims history, provider and recipient data, procedure code information and data tables, and will store electronic files such as medical policy and claims adjudication criteria for user reference.
"It’s a fairly comprehensive collection of data," Davis says.
A two-pronged approach
MFADs uses these two forms of detection, LeBrun says:
1. Neural network model.
This model identifies suspicious patterns or risks given a multitude of different elements, Davis says.
The model reviews about 300 to 500 different elements or components and aggregates all of the information, she explains. "Then it [formulates] different types of analyses for the investigator to review. That can be combined with specific fraud-scheme queries that are pointed to a certain outcome."
For example, a neural network model can look at physicians who have an unusual number of frequency of hospital admissions for certain diagnosis types or certain age groups. The model will evaluate whether the admissions could be a potential risk for fraud or abuse — whether they fall within a normal range and are valid or whether they fall completely out of the norm for what has previously been designated for that provider or peer group. Models will identify changes in relationships of data and will search for what goes undetected from viewing data claim by claim.
"[The models] look at providers based on the type of claims they file rather than how they enrolled themselves in the program, and the models evaluate and profile them accordingly," Davis says.
The Texas investigations office has built about four neural network models, LeBrun says. One model resulted in 56 cases, another 22. "Right now we have cases that have been assigned for investigation from the models. Two are serious, and one could be a potential criminal referral."
2. Fraud detection algorithm.
Algorithms are moderate to complex queries on the data, Davis says. "The algorithms make links within the data and produce a result."
"When we have knowledge of a pattern or a scheme of fraud, waste, or abuse, we develop an algorithm that seeks those patterns from the data in the system," LeBrun says.
The algorithm goes through the data quickly and produces a list of suspects. "The list shows individual instances where a provider has been found to be involved in that practice or that scheme," she explains. "We take [the providers on the list], and we test and validate them. We do research to make sure that everything is correct. Then we rank the providers that pass the tests and validation in order [of the most serious]. The top cases are assigned for field investigation."
The Texas investigations office is now using 34 algorithms. "We run some more frequently than others," LeBrun says. "Some are more productive; some are more reliable. And the pattern of fraud and abuse changes so we have to discard some of the old algorithms and build new ones."
LeBrun has algorithms showing $8 million, $7 million, and $6 million each in recoverable payments. However, she says she is not sure she will recover every dollar.
"I have some that are 100% [recoverable]. I have one right now that we are finalizing that is more than $500,000, and that’s 100% [of the money owed]." LeBrun says she expects to recover less on the cases from some of the other algorithms. "Some are more complex than others."
LeBrun says she can’t send field investigators to examine all of the cases. "We had a list of 400. We ranked them and decided that out of those 400, 125 were worth investigating."
Although the remaining cases are not investigated, they do remain in the system if the investigators have the time and the resources later to look at them.
"We are required by statute to prioritize and pursue those cases that have the greatest potential for recovery and the greatest amount of money," LeBrun says. "That’s where I put my resources."
Subscribe Now for Access
You have reached your article limit for the month. We hope you found our articles both enjoyable and insightful. For information on new subscriptions, product trials, alternative billing arrangements or group and site discounts please call 800-688-2421. We look forward to having you as a long-term member of the Relias Media community.