Examine the broader context to identify fraud

February 1, 2013

Predictive analysis can link suspicious behavior back to fraudsters before claims are paid out

FRAUD REMAINS a stubborn and growing problem for the U.S. healthcare industry, raising costs for patients and cutting sharply into margins for insurance payers. Of the more than $2.5 trillion spent on healthcare in the United States annually, some $60 billion to $250 billion is lost to fraud, waste and abuse.

To date, much of the effort to combat fraud has focused on the laborious process of trying to recover money from false claims by investigating suspicious claims after they have been paid. On average, the “pay-and-chase” approach takes one to two years, and in some cases much longer to recoup payments.

A more effective method is to identify false claims before they’re paid. To do this, healthcare payers are now supplementing their business rules with sophisticated predictive analytics and link analysis to detect fraud, waste and abuse.

Most healthcare fraud is committed by organized crime groups and a small number of healthcare providers, according to the National Health Care Anti-Fraud Assn.
In fact, in May 2012, the Medicare Fraud Strike Force identified one of the biggest Medicare frauds ever. The services, totaling $452 million as documented on counterfeit medical charts, were billed to Medicare but never actually performed.

The fraud ring involved 107 doctors, nurses and social workers. Among the suspects arrested were the owners of two community mental health centers in Baton Rouge, La., who recruited patients-including vulnerable seniors and drug addicts-to participate in scams that involved submitting claims for phony patient services. Without the bust, U.S. taxpayers likely would have lost much more in the years ahead.

Since its inception in March 2007, the Strike Force has charged more than 1,330 defendants who have billed Medicare for more than $4 billion in false medical charges.
Last February saw yet another significant insurance fraud crackdown. This scam involved 10 doctors, nine separate clinics and 105 corporations in New York, all attempting to steal as much as $279 million over five years from private insurance companies. To handle the reimbursement transactions for the phony medical treatments, the fraud ring set up three billing locations and created and ran nine clinics in the Bronx, Brooklyn and Queens.

The clinics provided unnecessary medical treatments including physical therapy and radiology services. The fraudsters were cashing in on New York State’s no-fault insurance law, which allows drivers and passengers involved in bodily-injury accidents to collect up to $50,000 per person, regardless of fault.  

While the examples are sobering, the real challenge is to prevent healthcare fraud.

Predictive analytics help to combat fraud by identifying patterns in claims that might point to fraudulent activity and by tracking payers’ transactional and relationship data to uncover wider instances of fraud. Predictive analytics enable payers to discover unknown types of fraud, identify new schemes and recognize networks of fraud.

Using this type of capability, health insurer Highmark, for example, uncovered 20% more cases of actual fraud, resulting in a substantial increase in loss avoidance.

Payers often start with a rules-based approach that flags claims that fall outside certain parameters. The first step is to identify potentially fraudulent patterns and then develop the rules to flag them when claims are being processed. The rules could include instances such as specialist providers who submit claims using a particular code a certain number of times per month, or charges for services outside their areas of expertise.

Predictive analytics can advance the rules-based approach, making it more dynamic and effective, identifying more fraud and creating a line of defense against unknown schemes that rules do not catch. As predictive analytic models identify emerging types of fraud, payers can leverage them to develop new rules.

The intelligence in the predictive analytic system then “learns” from the new rule patterns and builds increasingly more sophisticated models. The most effective models not only highlight claims with the highest likelihood of fraud but also describe the reasons each claim looks suspicious, so claims can be assessed with maximum efficiency.

Enhance with Link Analysis
Fraud rings are a major concern for payers. When a reviewer is examining a single claim, it is helpful to see the larger picture. This is where link analysis comes in. Link analysis-a data analysis technique that examines relationships between organizations, people and transactions or between providers, members and claims-has been gaining popularity in recent years as a means of enhancing fraud investigations. It actually has general applicability for any organization that wants to better understand its customer relationships and consider the impact of both formal and informal networks of people, groups, organizations and events.

In combatting insurance fraud, link analysis works by ferreting out related claims or providers that might not always appear to be related in an obvious manner.

For instance, in the case of a pharmacy that appears to be a pill mill-that is, a facility that routinely prescribes or dispenses controlled substances outside the scope of prevailing standards of practice-the use of link analysis combined with predictive analytics might show that the pharmacy is not, in fact, suspect. There may be a pool of providers issuing an unusually high number of prescriptions that are simply being filled at the pharmacy.  

When viewed individually, each claim looks legitimate. But when viewed in a broader context, the questionable nature of the claims becomes clear.

Predictive analytics can deliver significant savings for healthcare payers. Insurers deploying analytics as part of their anti-fraud efforts have seen reductions in fraudulent claim losses of 20% to 50% and loss adjustment expenses of 20% to 25%. Predictive analytics and link analysis help payers detect more fraud, prioritize claims by likelihood of fraud, reduce false positives by more accurately identifying real fraud, and improve customer satisfaction by streamlining the payment of legitimate claims.

By making it easier to stop fraudulent claims before they are paid, analytics helps to cut costs, not only for insurance payers, but ultimately for patients too.