Each year, more insurers appear to be increasing their budgets in implementing sophisticated fraud detection systems; however, the majority of these investments are centered around post-payment fraud detection rather than detecting fraud prior to claims being paid.

San Rafael, Calif.-based technology solutions provider Fair Isaac Inc. found in a recently released survey that over this past year the numbers increased from 37% to 45% of insurers who introduced a higher level of fraud detection into their operations. However, about 80% of these insurers were only able to detect fraud in post-pay mode and not able to detect pre-payment fraud--that is, solutions that detect fraud before a claim check is distributed.

Some insurers have recognized the value of pre-payment fraud solutions and are capitalizing on these investment. Hollywood, Fla.-based VISTA Health Plan, for instance, is deploying technology that relies on predictive modeling to identify fraud early in the process. "Having a pre-payment fraud reduction program is critical because once a claims payment is out the door, you've lost the battle," says William Rushton, an executive with VISTA who oversees the fraud program. "We might only be able to recoup 10% to 15% of a claims payment using what we call retrospective fraud detection--after a check is dispersed. That's not good enough."

In a post-payment scenario, payers such as VISTA must spend money on legal and administrative efforts to recover lost claims payments that occur due to fraud. And, under this methodology, the company does not prevent fraud from occurring in the future because fraud perpetrators realize they can get away with it again, says Rushton.

But with predictive modeling in place to stop fraud at the pre-payment stage, an insurer can stop a $10,000 payment from being made in the first place, but also they have the ability to upset the "fraud pipeline" as criminals are deterred from engaging in such crimes again--at least with that particular insurer.

Rushton says that his company once relied on rules-based technology to combat fraud, which is based on networking with affiliates. But networking is not always powerful enough to prevent crimes. "Rules-based methods are not data-centric like predictive analysis is and equally as important, rules-based methods are not able to provide feedback in real-time as predictive analysis is," says Rushton.

Steve Dwyer