TORONTO – A Toronto-based managed services company is proposing the use of biometrics and other tools to combat insurance fraud, which it estimates costs the Canadian economy $10.73 billion a year.
About half of that number comes from health-care fraud alone at $5 billion, with auto insurance fraud following closely behind at $4 billion. Other forms of fraud include debit card, credit card and mortgage fraud, according to MakePlain Corp.
Michael Chettleburgh co-founder and COO of MakePlain recently completed a survey on health-care fraud with Fraudbox Inc. in partnership with the Canadian Health Care Anti-Fraud Association (CHCAA). Chettleburgh, who presented his findings at a media briefing Tuesday, is co-founder and director of research at Fraudbox.
The 2004 Canadian Health Care Fraud Survey surveyed 109 senior health-care insurance professionals and claims processors. The report found that 95 per cent of respondents said they were victims of fraud in 2003 and that 89 per cent said the problem will get worse. Survey participants indicated that anti-fraud measures were not implemented at their sites due to lack of executive commitment, said Chettleburgh.
Auto insurance fraud accounts for $4 billion annually. According to the Insurance Bureau of Canada (IBC), property and casualty insurers paid out $20 billion in claims in 2005 on premiums of $33 billion. Fifteen per cent of premiums paid by the consumer go towards covering the cost of that fraud.
When asked to comment on the state of fraud in the insurance industry, an IBC spokesperson Tuesday said since it’s a member organization it can’t offer its opinion without a consensus from the industry.
The spokesperson, however, did say that IBC has addressed this issue in the past.
“We’ve spoken extensively about the need to deal with (fraud) and we have an extensive history of dealing with insurance fraud in general,” said John Karapita. “Around this particular issue there is no position.”
IBC confirmed that it has held some preliminary meetings to look at ways of dealing with the problem but wouldn’t specify how.
Voice biometrics are a potential way of dealing with fraud. While the technology has been around for some time, its application in a call centre setting has only been possible in the last several years due to advancements in voice and data technology such as VoIP.
“The proliferation of this technology allows us to do these biometrics,” said Gary Saarenvirta, co-founder and CEO of MakePlain demoed his company’s version of voice recognition software that could be embedded in call centre software to help workers authenticate customers.
To use the technology, the company needs to record a baseline voice print using standard security questions such as an individual’s mother’s maiden name. That information is then stored in a database. Saarenvirta said voice recognition technology is about 70 per cent accurate.
MakePlain is also looking into voice risk analysis software that can detect if a person is lying.
While the technology has some benefits, the average cost to implement, manage and maintain the software has a hefty price tag at $1 to $2 million per year. Chettleburgh said using this technology could cut a 15 per cent incidence of fraudulent claims in half, but the industry remains unwilling to take leadership on this front for various reasons.
“Nobody wants to go first,” he said, adding that it’s incumbent on the government to take the lead in terms of employing this type of technology in their organizations.
Aside from biometrics, there are more traditional methods that businesses can use to determine if a person is making a phony claim. Depending on how much information the business has about the client, these methods can vary greatly, said Saarenvirta.
If a business already has existing information about the client, they can analyze fraud claims to generate a set of rules-based tools to identify a set of characteristics common to a phony claim.
Chettleburgh said many insurance companies use this model in conjunction with manual investigations but many insurance company executives won’t tarnish their reputation by admitting to fraud.
“CEOs of insurance companies say, ‘I can’t admit to a $30 to $40 million problem on my watch.’”
Other models include predictive modeling, watch lists and deviation detection or fuzzy logic. The latter is used in cases where the company doesn’t have very much information on the customer. In that situation, the tool will use a math model to put a score on claims. From there, the company can take the top three per cent of claims that are most likely to be fraudulent and send an investigator to follow up on those people, said Saarenvirta.