Thanks to new technologies, carriers are unearthing fraud such as "rent-a-patient" schemes and physician upcoding.
In a health care insurance scam that knew no limits in outright audacity, two Los Angeles doctors were indicted (one has so far pleaded guilty) on fraud charges for submitting fraudulent bills and patient records to numerous insurance companies. In this "rent-a-patient" scheme, the doctors hired "marketers" to recruit people with private health insurance to undergo unnecessary surgical procedures in exchange for cash or discounted cosmetic surgery procedures. Those willing to undergo the unneeded procedures were promised between $300 and $1,200. Patients were instructed by recruiters to describe false and exaggerated symptoms that were used to create medical charts used to make the surgical procedures appear to be justified. The doctors reportedly racked up claims totaling more than $2 million before being caught. Procedures performed on the otherwise normally healthy patients included colonoscopy, sinus surgeries and thoracic sympathectomy, commonly called "sweaty palm surgery."
Unearthing such scams requires both vigilance on the part of healthcare payers and an ability to rapidly look through claims data to connect the dots on suspicious activities. Carriers not only face the challenge of detecting and forestalling outrageous schemes such as the rent-a-patient fraud, but also are tasked with detecting the even-more pervasive instances of "soft fraud," in which otherwise legitimate claims information may be fudged or exaggerated.
Although fraud has reached crisis proportions across the industry, the ability to effectively contain it is an area of opportunity for carriers, which face increasing pressure to simultaneously cut costs and improve customer service while meeting regulatory requirements for rapid claims resolution.
TIME IS NOT ON OUR SIDE
When it comes to identifying claims fraud, time is of the essence; carriers may only have a matter of days to detect it before the check needs to be cut and sent. Many carriers are stepping up their anti-fraud efforts, but often lack adequate resources to fully investigate fraudulent claims and prosecute the perpetrators. And, once the money is sent, follow-up investigations and attempts to recover the money are more costly than many companies can afford. The Coalition Against Insurance Fraud, Washington, estimates that claims adjusters and investigators detect only about 20% of the fraud that occurs, and in many cases, payments were already made against these claims.
The ability to recoup fraud losses is greatly diminished at this point, observes Kyle Cheek, director of data analytics for Health Care Service Corp. (HCSC), Chicago, Ill., the parent company of Blue Cross Blue Shield in Illinois, New Mexico, Oklahoma and Texas. "At least 3% of health care dollars are paid to fraudulent claims," he says. "Of the claims that are fraudulent, about 10% are identified, and in 10% of those identified, about 10% of the dollars are recovered. It is difficult to recover dollars after they've gone out the door."
Typically, "payers have only been able to identify fraud after claims have been paid," says Joanne Galimi, research director with Gartner Inc., Stamford, Conn. "They may have some kind of analytic tools to look at the claims data and flag claims that were potential fraud, but from there the flagged claims would be sent to investigative units, and they would have to do it all manually by paper, with no automated processes, to determine if this is indeed fraud. Then it would go through the whole litigation process, and basically, the ability to recoup those dollars is very minimal at that point."
The Insurance Research Council (IRC), Malvern, Pa., estimates that more than one of every three bodily injury claims from car crashes involves fraud. These fraudulent claims add more than $6 billion to the price of premiums every year, the IRC says. The IRC also found that only one of four insurers thoroughly investigates cheating on insurance applications, and even fewer investigate insider fraud from employees and agents.
SOFTER IS NOT BETTER
Even harder to track than outright malicious cases of fraud is the phenomenon of "soft fraud," which may involve instances such as a physician altering information on a procedure so a patient may be eligible to collect, or the padding of expense bills by vendors. A survey of 353 property/casualty companies, conducted by the IRC in conjunction with ISO Properties Inc., Jersey City, N.J., found that between 11 to 30 cents of every claims dollar is lost to soft fraud.
"Most of us know about hard fraud," says John Lucker, principal and national leader of advanced quantitative services with New York-based Deloitte & Touche USA LLP. "Criminal enterprises-in a very calculated way-design ways to deceive and steal from insurance companies. That includes crash rings and workers' compensation fraud, where physicians collude with various parties to create claims or medical treatments."
However, Lucker continues, "Soft fraud is more subtle. For example, people may think that since they've been paying premiums for years, they shouldn't have to pay a $1,000 deductible. So, they embellish or exaggerate their claim."
Typically, claims adjusters and investigators relied on manual review processes or even gut-level hunches to unearth fraud in claims. The problem is that overworked analysts and special investigative units affect turnover, and carriers cannot afford to retain or train claims adjusters.
"There was a point in time when we would use fairly manual rules-oriented processes to identify suspect providers," says HCSC's Cheek.
MODELING TOOLS
Spotting potential fraud patterns-whether "hard" or "soft"-requires sophisticated analysis beyond the capabilities of even the most seasoned adjusters. As a result, many carriers are turning to predictive modeling solutions that can rapidly sift through claims data and flag suspicious activities and patterns, before money goes out the door.
Such technologies have already been deployed successfully within the banking and credit card industries to help spot fraud, says Galimi. She urges insurance carriers-particularly those in the health insurance sector-to adopt fraud detection and prevention technologies. Such tools can be positioned in front of the adjudication system, and "help identify potential fraud in claims, flag them, send them to the unit, and do initial investigative work before those dollars actually get sent out the door," she says.
Such solutions include data mining tools, combined with predictive analytics, that can help detect fraud patterns at points earlier in the claims processing lifecycle than was possible with manual scanning of documents. These tools can be configured to flag potentially fraudulent behaviors such as inflated or repeated payouts, or unusual or suspicious charges from vendors and providers.
Galimi points out that the fraud detection and prevention market now offers new software capabilities. Analysis technologies such as neural networks, analytic tools, business rule engines and detection systems can identify and quantify potential red flag events. Advanced technologies such as predictive modeling and statistical analytics also are being employed to identify and detect fraud. Other technologies include internal database comparisons, internal/external database comparisons, pattern recognition and voice stress analysis. Galimi also sees a growing role for business process management solutions, such as OnBase from Hyland Software Inc., Westlake, Ohio, in this process.
An important aspect of these tools is that they are built into the claims process, and therefore triggered automatically as part of the workflow and case management systems, versus one-off activities, Galimi states. Predictive modeling tools can be built right into daily claims management activities. Predictive modeling technologies, which emphasize an end-to-end lifecycle approach to claims management, can be run against initial claim documents as well as all claims transactions, from initial filing to post-settlement. The sooner the predictive analytics engine interacts with claims data coming from the front office systems, the greater the chances of stopping a payment against a fraudulent claim. Claims flagged as fraudulent can be moved to an investigative process.
Predictive modeling tools build fraud models by scanning past claims data associated with instances of fraud. Both current and incoming claims data are run against this model, looking for and analyzing complex relationships and patterns that are often missed by claims adjusters or investigators. The modeling software also develops fraud scores against individual claims through data mining, predictive model building and testing. If a claim reaches pre-set alert thresholds, a series of actions can be automatically triggered, from recommending additional investigative actions. Since fraud is constantly evolving, the fraud model itself needs to be kept current with new patterns.
80 HOURS A DAY
HCSC employs Enterprise Miner from Cary, N.C.-based SAS Institute Inc. to conduct predictive analysis designed to identify potential fraud among the 300,000 claims that pour in each day. HCSC mines years and terabytes of data relating to claims, providers, members, groups, accounts and products from all three of its Blue Cross and Blue Shield plans.
For example, HCSC's fraud detection analysts might discover that a provider is submitting bills for services that far exceed the hours within a day. Using the SAS tools, Cheek's group can decipher the nuances in the data to see, for example, when a cardiologist is billing for 80 hours of service in one day. That's the hook Cheek needs to turn the case over to HCSC's special investigation team of former federal law enforcement agents, lawyers, claim experts, CPAs and physicians.
At the same time, Cheek's team can now look at the data to determine when several providers are submitting claims using the same provider code, which is an indication that they're part of the same practice. That's an instance when a provider might legitimately bill for 80 hours in a single day, which helps HCSC avoid needless investigations that are costly for both the company and the providers. "We wanted to get past the strictly rules-oriented tools that tend to dive deep, but not in very broad swaths," says Cheek. "We wanted to introduce analytics into our process, so we could canvass more data more quickly and more effectively, so it would not put us in a position where we were investigating someone who should not be investigated."
HCSC's claims investigators use the analytics tools to make two passes at the data. "The very highest level pass that we make through the data is a very agnostic approach to fraud detection," Cheek explains. "Instead of setting up the data mining tools to point at specific types of patterns, we use analytic approaches to model normal behavior, and then we find the entities or providers that don't fit into that definition of 'normal.' After we do that, we take those who most egregiously deviate from normal patterns and start to do some drill downs to identify the reasons that they're such anomalies."
LAWYERS AND CHIROPRACTORS
To save time, National Health Insurance Co., Grand Prairie, Wis., turned to a third-party service to quickly sift through and turn around data on claims. Mary Smith, assistant vice president of oversight and procedures for National Health, estimates that the service has helped save the company about $200,000 over the past year by uncovering inflated or fraudulent claims before payments were made.
As is the case with many carriers, National Health did not have the resources to sift through its claims on a daily basis. "It's very hard for examiners to catch fraud and abuse," says Smith. "We don't have the time to investigate every claim. It's hard for our examiners to be working and producing at the level that we need to stop and look to see if every diagnosis code agrees. They have to deal with production and get many claims out the door so we can stay current in processing our claims."
National Health partnered with HealthCare Insight (HCI) Salt Lake City, Utah, to provide a second level of checking of its claims data. "All claims go through our complete adjudication process," Smith explains. "Within our system, we have edits built in. Both our base system and an add-on application look for fraud and abuse. After the claims go through our adjudication process, but before we cut a check, we send our claim file to HCI. They run it through their system, and their system flags any claims that are suspect of fraud or abuse." The process takes less than 24 hours, she adds.
Examples of fraud the HCI analysis helps National Health spot include soft fraud, such as assistant surgeons being billed at the same rates as surgeons, or physicians upcoding their bills to help a claimant cover a treatment not covered by their policy. Hard fraud schemes National Health has uncovered include procedures being conducted on patients that are not appropriate for their age, or collusion between providers for unnecessary procedures. "We've found lawyers colluding with chiropractors. They will get a police report, and they'll send the person involved with the accident to a specific chiropractor. The chiropractor will order MRIs on everyone, even though they're not injured."
Soft fraud instances usually are difficult to detect, but can be spotted through "statistical signals" that the tools can elevate to the attention of claims adjusters, Lucker says. "You're typically not predicting soft fraud, you're predicting a propensity for it within a group of claims. There are statistical signals embedded in the claim, or in the claimants, that may indicate a possibility of some soft fraud possibility."
Typically, to address instances of soft fraud, Lucker advises segmenting claims or claimants into categories. "Then you triage those claims and build operational processes to handle those claims. It's very expensive to handle 100,000 claims individually, one at a time. So you want to segment them, and find the 500 or 1,000 claims that have a statistical propensity to be problematic. Then you build an unobtrusive process to handle those claims differently."
This may consist of a follow-up phone call asking for additional information, versus a formal investigation, he adds.
Simple adjustments to processes to change behavior may head off a great deal of soft fraud as well, Lucker says. For example, asking for receipts with all claims, or even simply adding a message in customer service calls that "this call may be monitored for quality purposes" may head off claimants that were considering fudging on a claim. "Embedded in that message is a subtlety which says, 'if you are not truthful and you get caught, your words could be used against you,'" Lucker explains. "A person who may have an intention to exaggerate or embellish may think twice about doing so."
Joe McKendrick is an author and consultant specializing in information technology, based in Doylestown, Pa.
