-
-
Marketplace
-
Channel Resources
Articles from this Site
-
One brand, one Web site! DM Review is now the home of all the content you're used to at BIReview.com and much more. If you are registered at BIReview.com, you're already registered at DM Review. If not, take just a moment to sign up for all the free services we have for you at the new DMReview.com.
Fraud Fighters
Asked why he robbed banks, depression-era bandit Willie Sutton replied, "because that's where the money is." If Sutton were on the streets today, he'd probably be staking out a doctor's office. Of the $1.7 trillion that flows through the U.S. health care system annually, government and industry experts estimate that at least 3 and as much as 10 percent - up to $170 billion - is lost to fraud.
The anti-fraud business in health care is a growth industry and already deals with the noir elements of a detective potboiler - rackets, offshore crime rings, sex and guns traded for narcotics. Unlike the crime scene investigators on television with their rubber gloves and fingerprint kits, fraud investigators don't start out with a corpse or even corpus dilecti - proof that an injury or loss has been caused by a criminal act. Layer on the fact that some four billion insurance benefit transactions are processed every year, and you begin to see why opportunists - many of whom are "upright citizens" with no prior criminal record - are tempted to try to beat the system.
Unlike property & casualty, where an agent owns a claim as it moves through the system, health care is based on trust, meaning that claims submitted are believed to be true and correct on their face. Not just insurers, but the vast majority of honest providers, policy holders and taxpayers ultimately bear the cost of dishonest providers who work alone or in collusion with patients. The tools of corporate, government and law enforcement teams are a mix of legwork and data gathering, increasingly supported by workflows and analytical tools that try to get ahead of the game. "A prospective approach means you want to see the claims somehow before they are adjudicated," says Joanne Galimi, a research director at Gartner, Inc., "because once the dollars have gone out the door you usually can't get them back." Most current approaches are still based on historical data however, and can at least stanch the bleeding of an ongoing scheme.
Figure 1: Trail of a Case Closed
Chipping Away at Fraud
Highmark Inc. came about from the merger of two Pennsylvania Blue Cross and Blue Shield associations, and today is one of the larger insurers in the U.S., processing about 124 million claims in 2003. Even with 20 full-time agents working in its Special Investigative Unit (SIU), it's difficult to imagine where an investigator would begin to work against that kind of transaction volume.
"In the past, you couldn't do much outside of tips or the really obvious," says Tom Brennan, director of the special investigations unit at Highmark. Over time, because of the magnitude of the situation and his inability to get a clear look at data, Brennan began to work closely with Jack Emes, Highmark's director of decision support in the company's informatics division. The 90 or so programmers and analysts in informatics provide a variety of services such as cost trending, financial forecasting, provider profiling, risk analysis and consulting.
Two people were dedicated to Brennan's SIU unit, first to provide data mining and analysis services when SIU saw telltale signs of fraud. Informatics went a step farther by coming up with some rules-based detection to uncover ongoing fraud that escaped the SIU's attention, such as a doctor who reports working more than 24 hours in a day. "Right now, we are aware of some known scenarios, where we know certain providers might be a problem or situations where fraud might likely be present, and we scan claims on a limited basis for that," Emes says. Another type of fraud detection is pattern recognition, which applies advanced techniques to identify statistical outliers, and helps flag future claims. Since the types of fraud change rapidly, it is hoped this kind of detection can be applied to head off big losses.
Emes is now experimenting in fraud prediction, working with technology partners on decision tree and cluster analysis, as well as advanced statistical tools in the SAS Enterprise Miner product. For all the needs of the market, Emes says there still aren't too many dedicated tools that can form a platform for fraud detection and given disparate infrastructures, insurers will likely cobble together solutions.
In practice, the threats vary and require follow-up, from the occasional medication claim for a pet rabbit all the way up to complex schemes. "We might look at a particular problem with up-coding a visit, and as a result of running the analytics discover we are also being billed for claims that were never rendered," Brennan says. In one investigation, a caregiver's high utilization of a specific steroid shot uncovered a much broader drugs-for-sex operation, a $1.6 million recovery and future cost avoidance of $600,000. The doctor involved is now serving a jail term. "This was a physician that the state bureau of narcotics had information on, but didn't have enough probable cause to do anything until we all worked to develop the case," Brennan says. (See "Outside the Firewall.")
Proof of success arrives in the form of graphs that chart escalating billing behavior, followed by the intervention of SIU, informatics and the attorney general's office, and then a steep drop-off in claims paid. Having the informatics division already in place lowered the cost and risk of building the new investigative division, but the obvious lesson is about putting resources to work where they are most valuable. "All you have to do is look at Tom's [Brennan's] budget and see that he has at least a seven-to-one return in terms of hard money actually coming back to Highmark," Emes says. "When you look at all the dollars out there, and the high fraud rate that has been documented, it's pretty obvious that more work is going to save a lot more money."
Gartner Inc.'s 2003 Payer Application Study found that 45 percent of payer respondents had fraud and abuse tools, up from 37 percent in 2002. Thirty-seven percent of these tools came from business intelligence application vendors, and 22 percent were internally built. Eighty percent of these systems rely on historical rather than predictive intelligence, but these systems are valuable and necessary for storing patterns and discrepancies. Further, payers are reporting demonstrable ROI. "Fraudulent claims are increasing much faster than manual systems can match," Galimi says. "By 2007, we believe health care payers who adopt automated systems will see a return on investment of at least five-to-one."
The high return also indicates the depth of the threat. The growing volume of transactions enabled by EDI and other methods only increase vulnerability, making predictive analysis that much more important in the future. "You have to support analytic tools just because of the volume and the fact that behavior changes every day," Brennan says. "There are individuals out there who can hit you very quickly for a lot of money and be gone."
Jim Ericson is editorial director of DM Review, a SourceMedia publication. You can reach him at Jim.Ericson@sourcemedia.com.
For more information on related topics, visit the following channels:
