Predictive Modeling, Analytics Expanding Fraud-Fighting Toolkit for Regulators
08-11-2010
According to BNA, regulators and law enforcement are expanding their use of new health care fraud-fighting technologies. Predictive modeling and analytics involve datamining large amounts of information and then building models of expected behaviour. In a Medicare setting, for example, individual models can be created from information including health records, pharmacy use data, and prior health care claims, and then compared against incoming claims. When a claim deviates from the expected model, the system can flag it for further review.
The Dartmouth Atlas Project, which has been analyzing the distribution and consumption of medical care in the United States for the past 20 years, has utilized modeling and data-mining software from SAS, a provider of business analytics and business software based in Cary, N.C., to look for discrepancies in Medicare spending. The software can analyze large amounts of health care information, looking for patterns and models in Medicare claims data. Over the course of the Atlas Project, SAS software has helped lead to several hospital fraud investigations after identifying suspicious activity, according to an SAS white paper.
The Department of Health and Human Services has also used SAS software to fight Medicare fraud, waste and abuse, and renewed a contract with SAS for three years in April 2009. SAS software is used throughout HHS, including in the Centers for Medicare & Medicaid Services.
Predictive modeling systems have been in use by the financial services industry for many years, and have been used to prevent credit card fraud in real-time settings. The work of the Recovery Accountability and Transparency Board (RATB) is another example of technology being used by regulators in the fight against health care fraud. The RATB is currently involved in a pilot program with CMS that involves analyzing Medicare transactions with a privately-developed fraud mapping application. The RATB was created by the American Recovery and Reinvestment Act of 2009 to provide transparency for stimulus funding and prevent any associated fraud (14 HFRA 498, 6/16/10).
Douglas Hassebrock, assistant director for investigations at the RATB, told BNA that Medicare has been receiving recovery money, and that they are focussing on the risks posed by the recipients of the funding. The fraud mapping application was developed in conjunction with several private firms, Hassebrock said. Additionally, Hassebrock said that they can even go into an in-depth analysis of the fund recipient. That way, it is possible to check all the risks associated with the principals of a company. A full analysis takes a day to a few days, based on the complexity of the case. The pilot program is not expected to be completed before early September. If CMS likes the results from the pilot program, they might procure their own system, based on the RATB model, Hassebrock said.
Jeff Mudd, a federal sales director with SAS, told BNA that predictive modeling software is capable of handling the massive amount of information associated with CMS transactions. According to Mudd, analytics software can help stem the tide of health care fraud by detecting the fraud and then by focusing subsequent investigations by allocating resources more efficiently, among other steps. Additionally, a new fraud fighting tool, known as social network analysis, is starting to gain interest within the health care community, Julie Malida, a principal for health care fraud with SAS told BNA. ‘‘Social network analysis, which is also called link analysis, involves building statistically significant links between people, then running the links back through predictive modeling and anomaly detection systems,'' Malida said.
According to Malida, organized crime and collusion are growing trends within health care fraud. Link analysis can therefore be used to fully detect the disparate relationships that often occur within crime rings. Lately, there has been increased interest in fraud detection software from Congress, including several bills that specifically call for the adoption of software to help rein in Medicare and Medicaid fraud.
One such bill, H.R. 5546, would require CMS to implement a prepayment review prevention system. Sponsored by Rep. Peter Roskam (R-Ill.), the bill would require the system to analyze CMS claims in near real time, flagging any suspicious activity for further review and investigation. All flagged claims would be prevented from being paid until after review. ‘‘Predictive modeling ‘scores' a claim to identify claims that have a high probability of fraud. A predictive model creates an estimated score on claims using historical data. That estimate is then applied to new claims that are submitted. The predictive model is always evolving, improving and adapting to provider and patient behavior,'' Roskam said in June 15 testimony before a hearing of the House Ways and Means Subcommittee on Health focused on Medicare fraud (14 HFRA 498, 6/16/10).
The AARP endorsed Roskam's bill on July 27, along with two other Medicare fraud-fighting bills. A predictive modeling program for Medicare fraud was added to the Senate version of H.R. 5297, the Small Business Jobs Act (see related item in the Federal News section). The bill was sent to the Senate Committee on Finance on Aug. 5 with instructions to report the bill back to the Senate with the addition of an amendment. The Senate is on recess until Sept. 13.
Reproduced with permission from BNA's Health
Care Fraud Report,14 HFRA 674 (Aug. 11, 2010).
Copyright 2010 by The Bureau of National Affairs,
Inc. (800-372-1033) http://www.bna.com
