Property and casualty insurance fraud cost carriers about $29 billion last year, according to the New York-based Insurance Information Institute. The trouble is that's just part of the picture. It's safe to say the number would increase when you count fraud from life and health.It's also safe to say insurance fraud has been with us for a long, long time. Has detection improved? Does detection even matter if the perpetrators aren't prosecuted? Are insurers benefiting from fraud detection and prevention? And, how are they going about it?
According to the Insurance Information Institute, in the mid-1980s the rising price of insurance, particularly auto and health insurance, combined with the growth in fraud committed by organized criminals, prompted insurers to reexamine the issue. Gradually, insurers began to see the benefits of strengthening anti-fraud laws and imposing more stringent enforcement as a means of controlling escalating costs. They found ready allies among those who had been adversely affected by fraud.
Now, special investigative units (SIU) operate within insurance companies. And, according to the Insurance Information Institute, heightened anti-fraud activity, along with growth in funding for fraud-fighting personnel, has resulted in more prosecutions.
Anti-fraud efforts encompass all lines-property/casualty, workers' compensation and health/life. While criminals can commit insurance fraud at many points in an insurance transaction, a surefire way for insurers to lose out is to pay the claim before discovering the fraud, setting up a "pay and chase" situation, says Tom Brennan, director of the SIU of Pittsburgh-based Highmark Inc. "Once the money goes out the door," he says, "you are only going to recoup 20% to 30% of your losses, simply because the individual who has taken advantage of the system don't have the assets.
HOME-GROWN SOLUTION
Though many of the software tools on the market help detect fraud early, Highmark created its own, using Enterprise Miner from Cary, N.C.-based SAS Institute Inc.
Highmark's SIU works closely with its healthcare informatics group, which supports an SAS-based, Highmark-created sleuthing application dubbed FIRST (Financial Investigation Reporting System for Tomorrow).
The informatics group at Highmark builds sophisticated models based on huge stores of claims as well as on customer and provider data. The tool helps investigators uncover and prevent costly crimes, saving the company $11.5 million in 2005, both in claims recovery and cost avoidance, according to Brennan.
Before FIRST, the informatics group manually assembled queries for the SIU by bringing together data into reports they could then use in their investigations. "The idea with FIRST was to establish an application to empower the SIU staff to submit queries on their own. So we used Enterprise Miner and integrated that into a Web-based interface," says Shawn McNelis, vice president of health care informatics, research and analysis at Highmark. "They define and then launch these queries that are assembled behind the scenes, and it's an easy-to-use Web-based interface. So, his folks don't have to know SQL's impact or how to join tables."
By using FIRST, Highmark hopes to move away from having only historical data in the fraud-detection system. Instead, the company plans to use active data warehouse technology for real-time or near-real-time analysis. Highmark would then use the data to build predictive models that can anticipate fraud, essentially stopping crime before it occurs.
"We're trying to leverage the fact that we now load submitted claims on a daily basis into our enterprise data warehouse," says McNelis. Before the company began using FIRST, it took several weeks before claims reached a data repository that employees could use. Adjudicated claims would take even more time. "Now, by being able to partner, we can combine the SIU's subject matter expertise with our ability to leverage real-time or near real-time data and combine our talents to produce these," McNelis says.
Submitted claims are uploaded daily into the data warehouse. The SIU can submit a query via FIRST at any time. That query operates against data as it becomes available in repositories.
"It's a combination of my folks, who can leverage technology and data, and Tom's [SIU] guys, who have the vast subject matter expertise and experience," says McNelis. "We're bringing these two factions together to verify a strong partnership."
Using FIRST also freed up some time for the Highmark investigators. According to Brennan, work that used to take eight hours now takes only minutes, and investigators can now handle a 30% increase in caseload. That translates into productivity gains and personnel savings of $200,000 per quarter, he says.
MONITORING THE WEB SITE
Marj Hutchings, Director of Internet Operations at Esurance, San Francisco, can relate to the need to provide easy-to-use SIU tools. The auto insurer uses technology from San Francisco-based TeaLeaf Technology Inc. to gain real-time visibility into customers' experiences and activities on the Esurance Web site. The software alerts the production support staff when a customer is having difficulty using the Web site. The company also discovered that the system helps fight fraud.
Hutchings provides the example of a person who is in an accident and afterward goes to the Esurance Web site to purchase or change a policy-and then files a claim. "With TeaLeaf, that's something very easy to verify because everything is time-stamped," she says. "So instead of having to pull something from the archive and build a case and a timeline, we can play a mini-movie of a session that's time-stamped to the second to determine what actually happened."
These mini-movies help Esurance's SIU, says Hutchings. "Our SIU relies on the regular old-fashioned sleuthing. We have very highly trained people who have been dealing with insurance fraud for a long time," she says. "When they're given powerful tools it makes it so much easier to accomplish their jobs and identify things, and ultimately resolve the whole issue in a more timely manner."
Oftentimes, once a suspicious claim has been identified, the company needs to collect reports, background and activities for hand off to the SIU.
It's particularly valuable for someone who is non-technical, notes Hutchings, such as SIU associates, who are good at identifying fraud. "We can send them an executable of the customer's actions on the Web site."
The mini-movies can also be used when prosecuting fraud. "Hypothetically, if there was a prosecution in an insurance fraud case and the district attorney needed evidence, we would be more than willing to provide any record that would help the prosecution win a conviction," says Hutchings.
Other technology, in addition to the TeaLeaf information, can help in fraud cases, especially in the first instance.
Technology lets you gather large amounts of data; you can figure out where payments have gone," says Kirk Nahra, a partner at Washington-based law firm Wiley Rein & Fielding LLP. "Depending on what kind of fraud scheme it is, in some situations you may need to link together people who aren't obviously linked. You need to go into your claims records and payment records."
Having fought insurance fraud for almost 20 years, Nahra believes that technology can help confirm facts found through other investigative methods. He offers healthcare fraud as the example.
The doctor adds a CPT code every time he examines a patient. When evaluating claims status, technology provides an easy way to confirm the diagnosis rather than plowing through a million pieces of paper.
Nahra says he believes the technology to help fight fraud is more sophisticated. "What [the industry is] getting now is vendors who promise the ability to weed out these factors based on their artificial intelligence," he says. "A lot of companies are finding that technology is helpful but not sufficient by itself."
PREDICTIVE ANALYTICS
Like most insurers, Erie Indemnity Co., Erie, Pa., hopes to avoid the cost and headache of prosecuting claims fraud. The carrier uses FraudFocus from Alpharetta, Ga.-based ChoicePoint Asset Co. to review claims- from first notice of loss to claim closing-to identify potentially fraudulent claims and bring them to the attention of adjusters and the SIU. FraudFocus combines predictive modeling, text mining, identity matching and red flag alerts.
FraudFocus' predictive modeling relies on statistical analysis of claims data to develop algorithms that can identify claims likely to be fraudulent. Its text mining function sifts through claim diary entries and log notes for fraudulent patterns, allowing Erie to unlock critical information that is typically difficult to use in automated systems.
On the front end the carrier uses predictive modeling, or fraud scoring, says David Rioux, Erie's corporate security manager. "Our claims," he says, "are scored and those that have a greater propensity for fraud are isolated so they can be further reviewed by a claims adjuster [who can] determine if there are any indicators or facts or circumstances that would warrant a further investigation."
Before FraudFocus, Erie used a manual process-and trained claims handlers to recognize fraud indicators or red flags, says Rioux. "We still use that process, but all claims are screened against a fraud-detection model that further isolates claims that require greater scrutiny and review," he says.
Rioux has seen a couple specific improvements since implementing FraudFocus three years ago. Chief among them: Referral time has been drastically reduced. "From the time a claim gets reported until it is recognized that it needs further investigation-that margin of time from loss date to the SIU referral date has closed considerably," he says. "The other thing we realized is that the claims that get flagged by the system seem to be better quality referrals to the SIU department. They are more likely to prove fraud on the ones the system has flagged."
Erie also compares parties to the loss or entities and information about the loss to an internal watch list of known, suspected or questionable claims, some of which comes from NICB. "We're able to compare some active case data against our claims information, and sometimes we're able to intercept claims very early in the process through the analytical and intelligence component."
Rioux says insurers can become more proactive by becoming aware of a suspect claim long before it's paid-even before the claims handler has delved deeply into the claim.
"We've already intercepted a claim within days of it being reported," says Rioux. In many cases, because claim-related information was already developed and sitting in Erie's database, the time and effort in the discovery phase is shortened.
Using a known fraud ring as an example, Rioux says the claimant may not be an Erie policyholder, but if he or she tries to become one, a flag goes up and the system looks for matches in the company's intelligence database.
"We review the reasons and the circumstances, and we're able to get an investigator in on the ground floor," says Rioux. "In some cases-surprisingly within days or weeks-we have a proven case long before it was ever considered to be paid."
As for the future of Erie's fraud detection, the company is beginning to mine unstructured text. "We don't have a data field that captures everything, and a lot of information is being written and being recorded in text format," says Rioux. "Hopefully we'll be able to introduce it in the model and help give the model even greater lift and greater predictability moving forward."
An Industry Pulling Together
Insurance fraud is the second most costly white-collar crime in America-somewhat behind tax evasion, according to the National Insurance Crime Bureau (NICB), Des Plaines, Ill.
But some express optimism concerning the industry's ability to take control of the insurance fraud problem.
"I think it will get better; I don't believe it will ever go away until the insurance industry, with the help of the government, addresses the problem," says Michael Lucarini, a partner at Bermuda-based Accenture.
Lucarini suggests insurers start sharing more data with each other. "They need to share information about every case and every player that was involved in a prosecution, whether it was successful or not," he says. "And the government needs to go after and revoke the licenses of the attorneys and the doctors who are caught in these things and publish the information."
Privacy laws that protect the rights of policyholders and claimants against the release of confidential information may prohibit publishing some information. However, to bring a case to trial successfully, insurers must provide information to prosecutors on policy holders suspected of fraud, according to the NICB.
Immunity laws that allow insurance companies to report information without fear of criminal or civil prosecution now are in force in all states, but not all laws cover insurance fraud specifically or allow reporting of information to law enforcement agencies as well as to state departments of insurance.
Complex cases, involving large scale criminal operations or individuals that repeatedly stage accidents, can be turned over to the NICB, which serves as a liaison between the insurance industry and law enforcement agencies. The NICB has set up a standardized computer program to eliminate duplicate reporting and speed up electronic transmission. NICB also publicizes the arrest and conviction of the perpetrators of insurance fraud to help deter criminal activities.
Insurers can do more, Lucarini reports, offering Accenture's customers as an example. "The ones who are taking a strategic view of [fraud] by looking at their operating model, looking at their business processes and looking at their measurements-not only at their operational management reporting aspect, but how they measure individual performance and reward it-are the ones getting an edge," he says.
Other vendors, such as Jersey City, N.J.-based ISO Properties Inc., provide technology, tools and services to help insurers thwart claims fraud. ISO's ClaimSearch has a database of claims information and a data-mining system. The system furnishes data for researching prior-loss histories, identifying claims patterns and detecting suspect claims. ISO's Internet interface enables users to search the data.
