Massive Medicaid Fraud Busts Offer Blueprint for States

On June 22, the U.S. Department of Justice announced the largest Medicaid fraud bust in history. The National Health Care Fraud Takedown included 301 defendants charged, $900 million in false billings, 61 medical professionals and 29 doctors, across 36 states.  The massive investigation was led by the Medicare Fraud Strike Force, a joint effort between the Department of Justice and the Department of Health and Human Services.

The bust also highlights the promise of collaboration between states and the federal government. The work of approximately 20 state Medicaid Fraud Control Units (MFCUs) was cited by U.S. Attorney General Loretta Lynch as being critical to the takedown. According to AG Lynch, this reflects “the close partnership between state and federal authorities in combatting health care fraud – a partnership that we will continue to strengthen in the days ahead.”    

There is tremendous potential to crack down on health care fraud across the U.S. if federal and state agencies continue to work together. States are primarily responsible for policing fraud in the Medicaid program through MFCUs. What can they learn from the recent massive takedown, as well as from other busts around the country?

The great thing is, this could be done without sharing data, which is often prevented by law, policy or insufficient technology. The existing CMS led State Program Integrity Support & Assistance channels between the U.S. Center for Medicaid Services (CMS), State Medicaid Agencies and MFCUs could be used to share best practices from the Medicare Fraud Strike Force’s complex fraud busts in a trusted clearinghouse or consortium. This information would include attributes from various schemes and recommend how State Medicaid Management Information Systems can benefit from proven techniques. 

For example, there are instances in the National Healthcare Fraud Takedown whereby individuals colluded with one another through various referral schemes within clinics, pharmacies or even within certain areas of care. 

For example, investigators in New York uncovered more than $86 million in suspected fraudulent physical and occupational therapy claims to Medicare and Medicaid. Five individuals were charged for allegedly filling a network of Brooklyn clinics with patients by paying bribes and kickbacks. The clinics were controlled by the accused, and once there, patients were subjected to medically unnecessary therapy.

While this specific use case took place in an Eastern District of New York, there is much that can be leveraged by other states to extend the value of what was learned, and support the ability to do a similar analysis.

Finding the Relevant Data to Crack Down on Similar Schemes

In the example above, the data establishes relationships between individuals (five individuals), areas of care (physical and occupational therapy), locations (Brooklyn clinics) and what was ultimately identified as “medically unnecessary therapy.”  Even if one has rules or predictive models that are designed to detect “medically unnecessary therapy,” it would be useful to look at the various schemes that were identified and consider ways to create identifiers or data attributes/flags that would identify these behaviors and could be shared without necessarily sharing actual data. 

A recommended best practice would be to examine scenarios of known fraud and ask, what data preparation steps can we take to create new variables that might give each state the ability to detect the same scenarios? 

We could use claims data from the New York example to create a baseline of information, and use “peer group analysis” to identify outliers worthy of further investigation. Some things we might incorporate to create the baseline include:

• Patient populations served by specialty (i.e., physical, occupational therapy, other)
• Patient population by location (i.e., clinic, pharmacy, other)
• Diagnostic codes and billing amounts (what was billed and why) 
• A time element to further break down data by day of week, month or year 

Some of the abnormal things that could prompt an alert, when compared to a peer group, include:

• Average payment per beneficiary in the top 90th percentile

• High number of unusually large payments within a provider or clinic’s patient population.
• Average clinical severity of beneficiaries in the top 90th percentile. For example, a provider billing for a high number of urgent care procedures.
• Average treatment time in the top 90th percentile, indicating possible padding of hours

Comparing behaviors across the provider population can detect who is possibly doing something outside of the norms of their peers. A preponderance of suspicious events would help an analyst prioritize risk and focus their analysis on “the big picture,” as opposed to a single alert.

How Can States Improve Data to Prevent Fraud?

To perform effective peer group analysis, better data preparation is needed. It may behoove states to use federal funds to expand Medicaid Management Information Systems, and have that expansion take the form of better data preparation efforts and the integration of more data elements.

This is where groups like the Medicare Fraud Strike Force come in. They could compile and share such a list of proven data preparation recommendations in a trusted clearinghouse. 

One example of a secure clearinghouse that is up and running is the Financial Services Information Sharing and Analysis Center (FS-ISAC). This industry forum’s purpose is to promote “collaboration on critical security threats facing the global financial services sector.”  This clearinghouse is used by global financial industries who want to collaborate on cyber and physical threat intelligence analysis and sharing.

In addition, since 2009 the U.S. Department of Justice, along with U.S. Attorney’s Offices around the country have tallied up an impressive $29.9 billion in recovered money through the False Claims Act, with more than $18.3 billion of that amount recovered in cases involving fraud against federal health care programs.

Sharing data preparation elements from successful fraud fighting efforts would not just lead to more busts. It may finally get us to the point where, across the country, we move the detection of improper payments away from a pay-and-chase model to one where fraudsters are identified before payments are made.