Tag Archives: analytics

Using Link Analysis to untangle fraud webs

Posted by Douglas Wood, Editor.

NOTE: This article originally appeared HERE by Jane Antonio. I think it’s a great read…

Link analysis has become an important technique for discovering hidden relationships involved in healthcare fraud. An excellent online source, FierceHealthPayer:AntiFraud, recently spoke to Vincent Boyd Bryant about the value of this tool for payer special investigations units.

A former biometric scientist for the U.S. Department of Defense, Bryant has 30 years of experience in law enforcement and intelligence analysis. He’s an internationally-experienced investigations and forensics expert who’s worked for a leading health insurer on government business fraud and abuse cases.

How does interactive link analysis help insurers prevent healthcare fraud? Can you share an example of how the tool works?

Boyd Bryant: Link analysis is most often used to piece together different kinds of data from multiple sources–to identify key players, connections between those players and patterns of behavior frequently missed. It can simplify an understanding of the level of involvement of individuals and criminal organizational hierarchies and can greatly simplify visualizing and communicating the operations of complex criminal enterprises.

One thing criminals do best is hide pots of money in different places. As a small criminal operation becomes successful, it will often expand its revenue streams through associated businesses. Link analysis is about trying to figure out where all those different baskets of revenue may be. Insurers are drowning in a sea of theft. Here’s where link analysis becomes beneficial. Once insurers discover a small basket of money lost to a criminal enterprise, then serious research needs to go into finding out who owns the company, who they’re associated with, what kinds of business they’re doing and if there are claims associated with it.

You may find a clinic, for example, connected to and working near a pharmacy, a medical equipment supplier, a home healthcare services provider and a construction company. Diving into those companies and what they do, you find that they’re serving older patients for whom multiple claims from many providers exist. The construction company may be building wheelchair ramps on homes. And you may find that the providers are claiming payment for dead people. Overall, using this tool requires significant curiosity and a willingness to look beyond the obvious.

Any investigation consists of aggregating facts, generating impressions and creating a theory about what happened. Then you work to confirm or disconfirm your theory. It’s important to have tools that let you take large masses of facts and visualize them in ways that cue you to look closer.

Let’s say you investigate a large medical practice and interview “Doctor Jones.” The day after the interview, you learn through link analysis that he transferred $11 million from his primary bank account to the Cayman Islands. And in looking at Dr. Jones’ phone records, you see he called six people, each of whom was the head of another individual practice on whose board Dr. Jones sits. Now the investigation expands, since the timing of those phone calls was contemporaneous to the money taking flight.

Why are tight clusters of similar entities possible indicators of fraud, waste or abuse?

Bryant: When you find a business engaged in dishonest practices and see its different relationships with providers working out of the same building, this gives rise to reasonable suspicion. The case merits a closer look. Examining claims and talking to members served by those companies will give you an indication of how legitimate the operation is.

What are the advantages of link analysis to payer special investigation units, and how are SIUs using its results?

Bryant:  Link analysis can define relationships through data insurers haven’t always had, data that traditionally belonged to law enforcement.

Link analysis results in a visual reference that can take many forms: It can look like a family tree, an organizational chart or a time line. This reference helps investigators assess large masses of data for clustering and helps them arrive at a conclusion more rapidly.

Using link analysis, an investigator can dump in large amounts of data–such as patient lists from multiple practices–and see who’s serving the same patient. This can identify those who doctor shop for pain medication, for example. Link analysis can chart where this person was and when, showing the total amount of medication prescribed and giving you an idea of how the person is operating.

What types of data does link analysis integrate?

Bryant: Any type of data that can be sorted and tied together can be loaded into the tool. Examples include telephone records, addresses, vehicle information, corporate records that list individuals serving on boards and banking and financial information. Larger supporting documents can be loaded and linked to the charts, making cases easier to present to a jury.

Linked analysis can pull in data from state government agencies, county tax records or police records from state departments of correction and make those available in one bucket. In most cases, this is more efficient than the hours of labor needed to dig up these types of public records through site visits.

Is there anything else payers should know about link analysis that wasn’t covered in the above questions?

Bryant: The critical thing is remembering that you don’t know what you don’t know. If a provider or member is stealing from the plan in what looks like dribs and drabs, insurers may never discover the true extent of the losses. But if–as a part of any fraud allegation that arises–you look at what and who is associated with the subject of the complaint, what started as a $100,000 questionable claims allegation can expose millions of dollars in inappropriate billings spread across different entities.

SEC taking stock of analytics (and a cool use case for stock exchanges)

Image Posted by Douglas Wood, Editor.

I read with interest today an article about the SEC’s use of analytics in the ongoing fight against financial crimes. The link to that article is below. It reminded me of some work I once did with a major stock exchange around insider trading.

As a passionate anti-fraud technologist, I was thrilled with the challenge of helping the stock exchange better recognize cases of illegal insider trading. The results of the work we did was pretty cool.

The stock exchange – as do all exchanges – had a great deal of data at their disposal. They knew, for example, the names of each and every ‘insider’ within every company listed on their exchange. Insiders include senior executives, board members, legal counsel, auditors, and so on.  Basically, everyone who knew – or ought to have known – about an upcoming event that would likely cause a significant change in stock price.

They also knew, of course, the identity of investors who traded profitably prior to the public release of that information. The problem was exposing the hidden relationships that might exist between Insiders and investors.

Here is an actual example… Joe Blow was an associate partner at the independent accounting firm responsible for auditing the quarterly financial results of publicly traded Company A. By definition, Joe is an insider. He knows Company A’s financials. Jane Doe dumped her entire position in Company A mere days ahead of what turned out to be very poor results. The stock plummeted, and Jane was saved from significant losses.  She was seemingly a complete outsider. So, did she somehow know Joe Blow (or any other insider)? Or was she just one lucky gal.

Using link analysis, crime mapping, and behavioral analytics, we set about the challenge of finding out. Here’s what the analytics exposed:

Joe Blow, the insider by way of being employed at Company A’s auditing firm, shared an address with… oh, let’s call him “Rich Quick”. Rich held no positions with Company A whatsoever.  He did, however, own a pet food store with a lovely young lady.  Can you guess her name?  Yep.. Jane Doe. So, the analytics exposed that Company A had an insider relationship with Joe Blow. Joe lived with Rich Quick. Rich owned a business with Jane Doe. Coincidence? Not likely.

Without the ability to draw out hidden links between individuals and organizations, this case may never have been discovered.  It’s like Six Degrees of Kevin Bacon, only with much higher stakes. All of the suspects were investigated and prosecuted.

(Note: All the names in this example are fictitious, but the case is not. If your name happens to be Jane Doe, Joe Blow, Rich Quick… or if you work for an organization called Company A, rest assured that I’m not talking about you.)

Here is the link to the SEC article.  http://fcw.com/articles/2013/09/18/sec-taps-analytics-to-predict-risk.aspx?s=fcwdaily_190913 .