Filing fraudulent insurance claims for health care services has always been an alarmingly easy crime to attempt. And it's becoming easier as criminals get smarter.

In the U.S. alone, nearly $100 billion a year is lost to all forms of medical fraud and abuse. That's almost ten percent of every health care dollar. Reliastar Financial Corp. has a strong interest in preventing fraud because Reliastar provides insurance and health care benefits to hundreds of mid-sized and large employers throughout the U.S.

For years, Reliastar had to depend on astute claims processors or anonymous telephone tips to identify fraudulent claims before they were paid. By 1994 the savings gained each year from this cumbersome and labor-intensive effort were flattening out.

Reliastar was stretched to its limits trying to catch all the crude and simple schemes that criminals devised, let alone the newer and more sophisticated schemes. As a result, many fraudulent claims were undoubtedly going undetected. The company decided that if criminals were going to get smarter, so must Reliastar. With the support of senior management, Reliastar created a program called Integrity Plus in 1994. As a first step, Reliastar's staff of fraud investigators was increased.

More important, Reliastar realized that its investigators could no longer scrutinize the behavior of providers on a claim-by-claim or incident-by-incident basis. If only one example of fraud was found, the provider could always say, "Oh, we have a new clerk in the office." or "That was just a coding error. It won't happen again."

Investigators had to be able to review several claims from a provider and identify patterns of behavior that would alert Reliastar to potential fraud. So in mid-1994, Reliastar began to build a warehouse of claims data for the sole purpose of detecting fraud. Investigators were also trained to use newly developed Fraud and Abuse Management (FAMS) Software, so they could specify and then target suspicious patterns of behavior.

Before Integrity Plus went into operation, Reliastar had identified only 300 health care providers nationwide whose practices seemed questionable. By mid-1997, Integrity Plus had boosted that number to nearly 4,000. As a result, savings from fraud detection were no longer flat from year to year. In the program's first year, savings shot up by 54 percent. By the end of 1996, savings increased another 25 to 30 percent. Through careful training and high technology, Integrity Plus began saving Reliastar several million dollars.

Turning Patterns of Suspicious Behavior into Models

The FAMS software was developed by IBM, who worked with Reliastar to develop many of the behavior patterns encoded in FAMS. For each medical specialty, Reliastar's staff of investigators listed all the questions that FAMS needed to answer. Because Reliastar's investigators are experts in medical and billing procedures in many specialties, they had no trouble coming up with the questions. IBM encoded all the investigators' questions into instructions for the FAMS software to follow. Some questions could apply to any specialty, such as "What percentage of claims filed by the provider were denied because there was no medical evidence that the procedure was required?" Other questions applied only to one specialty.

At the same time, IBM was working with several other FAMS customers to define and encode the behavior patterns these customers would need. Through these efforts, a library of 818 behavior patterns is now available for Reliastar to use.

Once Reliastar had the FAMS software, behavior patterns could be grouped into models for reviewing all claims within a particular medical specialty. Today, Reliastar uses 25 models, while more are being developed. The investigators working on Reliastar's models knew that even within one specialty, physicians' practices often vary widely from one region of the U.S. to another. Therefore, different ranges of values were established for different regions. This way, a model would compare a provider not only to peers in the same field, but also to peers who worked in the same region.

Building a Network for Detecting Fraud

The key to preventing fraud in the future is to identify suspicious patterns of behavior from the past. Therefore, investigators first used the models to examine data on claims Reliastar had already paid. However, reviewing hundreds of thousands of these "retrospective" claims required far more memory and processing power than could be provided by the token-ring network used at Reliastar's headquarters in Minneapolis.

In addition, Reliastar's token-ring network operates with Novell standards, while FAMS requires IBM OS/2 operating standards. For these reasons, Integrity Plus needed its own distinct computer system. Therefore, an OS/2 server with 16 megabytes of memory was dedicated to serve as the FAMS data warehouse. Next, programs were written for transmitting relevant claims data from Reliastar's mainframe system to the FAMS server.

All 19 fraud investigators work from PCs, but only three work directly in FAMS. These three heavy users are the ones who review retrospective claims for each provider, a process that generally takes two weeks. (Before FAMS, less detailed examinations took much longer.)

For every behavior pattern encoded in FAMS, the system established a numerical score from 1 to 1,000. FAMS scores the dollar amount of a provider's claim along this range. For simplicity's sake, Reliastar decided that any score above 500 would be flagged for careful scrutiny. If investigators uncover suspicious behavior patterns in the retrospective claims, reports summarizing the findings are sent to the database on the OS/2 server. That provider is now flagged so that investigators can scrutinize future claims.

Of course, not all the behavior that seems suspicious turns out to be fraudulent. But many of them are, including these flagrant examples:

  • A Los Angeles chiropractor, 20 percent of whose claims were for Sundays and holidays when the office wasn't even open.
  • A podiatrist who charged an average $978 while the median in his region is $481.
  • A dentist who logged 50 percent more visits per patient than the median and performed 20 times the average number of miscellaneous procedures.

At the same time, Reliastar began to identify specialties that needed the most scrutiny: podiatrists, chiropractors, psychiatric service providers, ambulance services, clinical labs and OB/GYN practitioners.

Staying Ahead of Criminals

To date, only the three heavy-using investigators work directly with FAMS. This is because when they need to access word processing and other functions on Reliastar's Novell network, the heavy users must leave FAMS entirely before they can enter the Novell system. This cumbersome procedure slows their progress somewhat.

Fortunately, when IBM releases a Windows 95 version of FAMS in 1998, investigators will no longer have to move in and out of two different networks. Then Relistar will probably dedicate another three investigators to analyzing retrospective claims. Eventually, all 19 investigators may work directly in FAMS, so the staff can continue responding rapidly to clients' needs as the work volume increases over the coming months. To generate reports from the Windows 95 version of FAMS, Reliastar may utilize newer IBM technology for visualization and graphic displays.

Buoyed by the success of Integrity Plus, Reliastar will continue to refine the existing models so that they reflect the latest practices of health care providers. Reliastar also plans to develop new models for the home health care and dental industries, workers' compensation and fraud practiced by claimants.

No matter how much Integrity Plus is enhanced and improved, a small number of health care providers will find more clever and subtle ways to defraud insurance companies. But if Reliastar can continue to develop sophisticated fraud-detection models, such criminal activity will become considerably more risky and much less rewarding.

Register or login for access to this item and much more

All Information Management content is archived after seven days.

Community members receive:
  • All recent and archived articles
  • Conference offers and updates
  • A full menu of enewsletter options
  • Web seminars, white papers, ebooks

Don't have an account? Register for Free Unlimited Access