How social network analytics can combat fraud
Government health care fraud is big business. According to the federal website, PaymentAccuracy, the government distributed nearly $65 billion in improper health care-related payments, and government agencies have been hard at work trying to put a stop to it — but with limited results.
Even though $1.7 billion was spent to combat fraud, waste and abuse in 2010, less than 10 percent of fraudulent dollars were recovered. This shouldn’t be surprising. Once the money is out the door and in the hands of criminals, it’s almost impossible to get back.
Time to Rethink “Pay and Chase”
This “pay and chase” model, however, is often the only anti-fraud weapon in an agency’s arsenal. The ideal situation is to thoroughly review every medical claim — but when patients need immediate treatment, there’s little time for a traditional investigation, and less money to fund it.
In fact, Medicare conducts reviews on less than three percent of medical claims before sending out checks. These realities are distressing enough, but what makes it worse is the disconcerting rise of organized crime. Take the case of south Florida, a popular retirement region that is now ground zero for organized crime. Losses to healthcare now measure in the hundreds of millions of dollars, much of it siphoned to crime rings based in countries across central and South America.
Smelling blood in the water, criminals have descended on the healthcare sector. Representing just a minority of claimants, criminals apply their traditional skills to extract tremendous sums from the vulnerable medical ecosystem. Constructing sophisticated fraud rings and phony businesses, criminals set up durable medical equipment shops, stage faked auto accidents and collect money from private insurance and government healthcare programs.
Some even dispense controlled prescription drugs in so-called “pill mills,” lucrative schemes that line the pockets of the criminal class while leaving in their wake hundreds of overdoses and drug-related deaths. Stopping these crimes isn’t just a matter of protecting an agency’s limited resources, but protecting the public welfare.
Traditionally, agencies left this work to law enforcement, and law enforcement in turn waited for tips and other pieces of information to trickle in. If enough data accrued, agents would begin to map the various connections and interdependencies of the criminal enterprise.
This time-consuming process depends on strong relationships between health care programs and law enforcement, ample resources to devote to an investigation and a reliable stream of tips. Unfortunately, the stars rarely align, making health care fraud a lucrative and low-risk target for organized crime. Technology is finally beginning to change that.
Stopping Fraud on the Front-end
Using big data and what’s called “social network analytics,” healthcare agencies can now reverse the traditional investigative process. Rather than starting with one suspect individual and then building out the network of contacts and business relationships, the new process begins with data. Lots of data — as much as 50 terabytes.
Here’s how it works: a health care agency delivers a list of providers and beneficiaries to its data analytics partner. Using this list alone and an understanding of the agency’s mission, the partner cross-checks providers and beneficiaries with its public records database. These records are essential: billions of public records on everything from business licenses to residential addresses enable the partner to make connections that are otherwise invisible to the healthcare agency.
[Not merely lost: A look at what happens to stolen medical records.]
Take the recent case we worked on wherein a state was suffering from a high incidence of Medicaid fraud for instance. When we were provided with the beneficiary and provider lists, we ran a quick query: how many people were both providers and beneficiaries? The answer was about 13,000 individuals. While this may raise a few eyebrows, it is not in itself an indication of fraud. Nurse aids and other medical staff may be listed as providers and still meet the requirements to be on Medicaid.
We went a step deeper and asked: of these 13,000, who has a wealth profile that is inconsistent with the average Medicaid recipient? This query returned 130 individuals. But this still wasn’t enough, because when we dug a little deeper, there were legitimate reasons for the majority of these individuals to be both wealthy providers and recipients of Medicaid. That left just two individuals who were suspect.
Next, we analyzed their immediate family and business connections within two degrees of association, looking for suspicious relationships. The result: a network of family groups with shared address and relationships, all very wealthy, all on Medicaid, and all serving on the boards of Medicaid providers.
Over 20 individuals were registered to the same address, with rampant fraud expanding far outside the scope of Medicaid. The discovery represented a major win for the agency’s anti-fraud efforts. It also illustrates a number of important points about the value of social network analytics:
- Time: While investigations of this nature used to take months, new technologies and large data sets enable agencies to uncover the connections in a fraud scheme in days or even hours.
- Coverage: Social network analytics looks across the full spectrum of an agency’s data, searching out fraud across the entire provider and beneficiary network and even across multiple jurisdictions or states.
- Proactive: Rather than waiting for a criminal informant or anonymous tip, agencies can root out fraud before it further exploits the system.
- Impact: The fraud uncovered through social network analytics is worth uncovering. This kind of fraud isn’t being perpetrated by doctors trying to scam the system, but organized criminals laundering money or using it to fund drugs and crime. Although it may represent a minority of fraud, the sheer size of these schemes costs agencies millions of dollars a year.
The analytics are only as good as the data
Enthusiasm for social network analytics is high, but it’s important for agencies to remember a few key lessons about the limits of technology. First, analytics are only as good as the data being analyzed. For agencies to be successful, they must have access to data outside of their sphere of knowledge.
For example, most agencies would not be able to make the connection between a doctor and the individuals who shared the same address with him 10 years earlier — even if they are all part of a pill mill scheme distributing prescription drugs. This is why public records data is the critical ingredient in fraud detection and prevention.
It’s also important for agencies to recognize that not all links and connections are indications of fraud. As we saw with the example above, there are a host of legitimate reasons for an individual to be both a provider and a beneficiary.
Although technology can make these connections, only people with an eye toward value can discern the significance. That doesn’t mean that these connections are useless: just that they may be the first step in an extended process of refining the query.
The new threat demands new tools to fight it
Today, health care is estimated to represent close to 18 percent of the gross domestic product of the United States. As this trend continues, agencies will see more and more instances of complex fraud perpetrated by organized crime.
The fraud won’t be easy to catch through traditional means, and it won’t be quick to flag with rules-based systems; it will represent the sophisticated efforts of a criminal enterprise.
The new threat demands new tools to fight it, including the rich technologies that underlie social network analytics. But, like every other effort in health care, the tools will need to be supplemented with good data, intelligent analytics, and a firm understanding of the values that drive each agency’s mission.
Though there is no silver bullet to fight fraud networks, the networks themselves may just be the silver lining.