Feds take (baby) steps to fight fraud with analytics

Government agencies have begun, tentatively, to mine their vast stores of data for the public good.

1 2 3 Page 2
Page 2 of 3

The ROC staff uses link analysis tools to uncover "non-obvious relationships" that may warrant further investigation. The principals of one debarred company, for example, may show up as executives under a different legal entity.

Because the analytics are focused on preventing fraud, rather than detecting it after the fact, it's hard to quantify results, says Wood. But as of the end of November 2012, the stimulus funds lost to fraud totaled only $27.8 million of the $283 billion.

ID'ing Medicare fraud

Preventing fraudulent payments for healthcare is the focus of a project at the Centers for Medicare and Medicaid Services (CMS). Prompted by a provision in the Small Business Act of 2010, the CMS in June 2011 established a predictive analytics system designed to prevent Medicare fraud. The CMS Fraud Prevention System screens 4.5 million Medicare claims each day, according to CMS officials.

The Fraud Prevention System works with an Automated Provider Screening (APS) system, which cross-checks the provider's information against government data sets such as license credentials, criminal records and Social Security data. The goal is to ensure each provider has a valid license to practice in that state, has not been fined or censured and is not using a stolen Social Security number, among other criteria.

CMS also geo-codes the information so it can match provider addresses with actual physical locations to see if there is an appropriate office or facility at that address, says David Nelson, until recently the director of the data analytics and control group at CMS.

"We look for indicators of ineligibility such as an address that turns out to be a campground, a provider that is using an invalid ID or a provider that has been sanctioned," he explains.

According to a CMS progress report (pdf) published in December 2012, during its first year of operation the Fraud Prevention System generated leads on more than 500 new fraud investigations, uncovered new information for another 500 ongoing investigations and stopped, prevented or identified some $115 million in improper payments.

In one example given in the report, the predictive algorithm flagged a provider as high risk, and upon investigation CMS found that 80% of that provider's Medicare reimbursements were for "highly suspicious activity" and that it was providing many more services per beneficiary than comparable providers. Once CMS confirmed that beneficiaries had not received these services, it suspended payments to the provider.

CMS says that the system is already showing a positive ROI, generating about $3 for every $1 spent.

(Article continues on next page)

1 2 3 Page 2
Page 2 of 3
Bing’s AI chatbot came to work for me. I had to fire it.
Shop Tech Products at Amazon