Fighting Fraud: Predictive Analytics & Business Rules Make a Powerful Combination

Insurance claims fraud, waste and abuse (FWA) constitutes a staggering financial problem in the U.S. The Coalition Against Insurance Fraud estimates that in 2006 a total of about $80 billion was lost in the U.S. due to insurance fraud. Insurance companies estimate that 10 to 20 percent of all claims today are fraudulent, and less than 20 percent of those fraudulent claims are even detected, let alone denied. At the same time, competitive pressures are requiring insurers to develop faster claims processing cycles, balance operational efficiencies against reduced claim leakage and improve customer retention strategies.

Technology designed to manage FWA has evolved a great deal in recent years, but there is much room for improvement. Incorporating automated business rules into a fraud management program that anticipates certain types of suspicious claim activity based on past fraud, for example, is a start.

This alone is not sufficient, however, because it is rare that even a chronic, predictable fraud problem can be solved simply with one black-and-white rule. Often the rule can cause problems for similar, legitimate claims, which are then delayed, causing customer frustration and leading to increased expenses. Additionally, even when a certain type of fraud is stemmed, another will inevitably arise.

Detecting problematic claims before payment is a payer's strongest protection against FWA -- and also the most efficient approach. But fraud is elusive, and patterns may not be recognized until after many payments have been made. That is where predictive analytics enter the picture. Analytics are a powerful weapon in fighting FWA -- before or after payment, detecting billing errors, and identifying systemic weaknesses and vulnerabilities such as oversights in provider contracts and loopholes in benefit policies.

Why Current Practices Fall Short

Many insurers handle claims processing and fraud identification in a reactive manner, often dependent on the experience of individuals on the processing team. Trends and patterns usually aren't obvious, and subtle exposures usually aren't visible. The approach that's needed is one that combines the power of analytics to detect unknown fraud, both new and unique, along with business rules management to block fraud earlier in the process.

Predictive analytics and business rules have a synergistic relationship. As predictive analytics algorithms identify emerging and increasingly routine sources of waste, they can help insurers develop new rules. The artificial intelligence learns from the new data patterns after the rules have filtered out the known fraud to continue to monitor for new fraud schemes.

In some cases, FWA is not obvious. One payer discovered an issue with an injectable cancer drug only after cross-referencing dosages with patient weight. Each patient would have had to weigh 700 to 800 pounds to merit the amount of the drug they were allegedly receiving. Building intelligence into the system with analytics and business rules management allowed the payer to have an automatic check in place to distinguish reasonable claims from unreasonable ones.

This intelligent approach also helps to manage the problem of false positives, where a claim matches the profile of an anomaly but is actually legitimate. False positives sometimes occur because particular authorizations are different according to various group plans and providers. What is normal or reasonable for one plan may represent FWA for another. With intelligence built into the system, these differences can be accounted for ahead of time and the system can identify a claim as a "known" false positive, avoiding delays in payment for legitimate claims.

Analytics + Rules = Measurable Impact

Payers in the healthcare industry are already embracing this approach. In one case study, a pathology group had been adding a professional services modifier for automated blood tests -- but automated blood tests do not require professional interpretation. An analytic system identified the discrepancy and enabled the payer to implement a rule that professional services modifiers would not be allowed on an entire set of tests, resulting in millions of dollars in savings.

This intelligent approach, with the addition of link analysis (a data analysis technique used to evaluate relationships or connections between nodes), lends itself especially well to the P&C sector. P&C insurers are minimizing loss accumulation by spotting fraud rings and associated claims earlier, and they are increasing recoveries by finding more new fraud cases as well as additional instances of fraud on existing cases. Accordingly, they are able to focus their investigative resources where they produce the greatest financial impact.

Analytic models "learn" from every new claim and improve their performance and ability to identify fraud. Predictive models that instantly analyze claims prior to payment have helped insurers identify as much as 50 percent more fraud, and combining those models with traditional rules-based systems enables carriers to locate aberrations in claims faster and more accurately. The speed at which claims can be processed also increases, as the analytics can easily identify which claims should be paid automatically and which should be reviewed.

FWA is unique for every payer organization -- issues can be contained to one procedure or to one provider. With business rules management technology built in, specific procedure codes for specific providers can be flagged for review while letting the others flow through.

Curbing FWA within the insurance space is a complex process, but combining business rules management and analytics makes for a powerful combination. It creates an intelligent system that is constantly learning, and it manages claims proactively, enabling payers to stop FWA ahead of time rather than chasing losses after the fact.

Russ Schreiber is vice president of the healthcare practice at FICO. He has more than 24 years of experience in the insurance industry.