Guest post by Scott Schumacher

The government's track record for reducing improper payments - particularly improper payments made through Medicare and Medicaid programs - is not something to write home about.

The Medicaid program within the Department of Health and Human Services (HHS) accounted for $13 billion in improper payments in fiscal year 2007, according to a then-released Government Accountability Office report. An updated report states that figure rose to about $18.6 billion in fiscal 2008.

Together, Medicaid and Medicare account for 50 percent of reported governmentwide improper payments in fiscal year 2008 - $35.8 billion, out of a total of about $72 billion. That figure does not even include the Medicare Prescription Drug Benefit program, through which $61 billion has been identified as "susceptible to significant improper payments," according to the report.

That's a lot of taxpayer dollars.

The report clearly states that agencies are aware of the challenges they face. Yet there does not seem to be movement forward.

Intelligence agencies are using identification and matching-and-linking technologies to uncover and help eliminate threats to our national security; state health information exchanges are using similar technologies to create a single view of each citizen, significantly reducing fraud and abuse at a state level.

These technologies, in use right now, are specifically designed to address precisely the issue HHS faces. Why, then, aren't agencies implementing these technologies to combat improper payments?

Identifying the Problem

To understand the solution, let's first identify the problem.

Let's say a Medicaid recipient regularly makes claims - and is reimbursed - for a particular drug. If the claim states that the person using the drug is a man, and the drug is something dispensed for a woman's health issue, that's fraud.

Let's say that same Medicaid recipient submits two reimbursement claims for the same medication and, regardless of the intent on the patient's part, the government pays the claim twice. That's overpayment.

To their credit, most agencies today are using some kind of technology to help them identify transactions that may represent improper payments. But many of these are older, statistics-based technologies designed to make determinations based on fixed identifiers such as Social Security number, taxpayer identification number, etc. Errors in these identifiers,either deliberate or unintentional, make the exercise null and void.

Contributing to the challenge, citizen/recipient information is too often stored using a systems-based approach.

Information regarding Medicaid recipients more often than not resides in a multitude of disconnected systems - lab systems, physicians' offices, billing applications within hospitals, etc. Because these systems are typically not connected and each data item is associated with individual systems rather than the recipient, there is no way to gain a comprehensive view of the recipient from the perspective of an HHS representative deciding whether or not to pay a claim.

The Master Person Index

Linking and matching technology used by intelligence agencies and state-level health organizations is designed specifically to tackle both of these challenges.

The foundation of this technology is a master person index (MPI).

The MPI identifies the same recipient across disparate systems and links records to create a complete view. It locates and links recipient records across contributing health information systems and allows real time searching from across all participant applications. It scales to support evolving information sharing requirements for an increasing number of participants and is interoperable with other applications at work within various health information exchanges (HIEs).

There are, however, several types of MPIs - and this is where agencies can find the most effective solution.

Some MPIs are based on deterministic (if/then logic) technology, which provides the ability to match records, similar to the statistics-based technologies mentioned above. MPIs based on probabilistic technology take into account a broader complexity of linking information from disparate systems, as well as the possibility of errors and fraud.

This technology, for example, would rate two records from recipients named Albert Einstein with a much greater probability of a match than two records from recipients named John Smith, since the latter is a much more common name and a less probable match.

Probabilistic MPI technology relies on multiple identifiers - not a single social security number, for example - and takes into account errors in names and numbers, purposeful fraud, and even differences in language.

Probabilistic MPI technology integrates a range of disparate data and provides a single, comprehensive view of all the different data associated with a particular person or entity.

Watch and Learn

Implementing this type of solution is not a forklift change. MPI solutions feed this higher-quality data into an agency's existing systems, which then reach a new level of efficiency.

Federal agencies, HHS in particular, can look at other agencies and learn. Intelligence agencies are using this today; state health organizations are using this today; commercial companies such as financial institutions and insurance agencies are using this technology today.

Having the distinction of being the agency with the "largest [improper payment] amount ... reported for a program" is not something HHS can be proud of. The opportunity is there to make significant changes, and save a significant amount of taxpayer dollars. HHS has the opportunity to start making those changes today.

Scott Schumacher, PhD, is a well-known government and commercial expert in complex data analysis. For more than 20 years, Schumacher has been heavily involved in research, development, testing and implementation of complex data algorithm technologies. This experience includes projects commissioned by the Department of Defense, leading advanced surveillance algorithm initiatives.