By STEVE TUCKEY, who has written on insurance issues for a decade for several national media outlets.
Investigators can now look to computer modeling to ferret just what is and isn't kosher, thereby making the workers' comp premium audit all that more effective when it comes down to ensuring that the exposure matches the premiums charged.
Russell Schreiber, vice president of Minneapolis-based FICO (formerly Fair Isaac), said the world of predictive analytics can provide numerous clues for carriers looking to detect all kinds of insurance fraud, including comp premium fraud.
For that task, his software conducts weekly analyses of employer payrolls, providing scores on any number factors. The higher the number, the more concern it will raise.
The neural network technology is specifically built to analyze numerous data elements that at first glance would have no connection to each other.
"It finds patterns that are not intuitive," Schreiber said.
The aggregate scores are then presented to the premium auditors and any numbers over 500 (on scale of 1000) will present some possible red flags worthy of investigation with possible reasons for the higher number. It is up to each carrier client to determine the threshold for more intensive audit.
Schreiber said the size of each audit staff relative to the size of the carrier determines the threshold.
"We work with our customers to turn the dials based on the threshold that their audit staff can handle," he said.
"We do not take the human interaction out of the process," he said. "What we do is tee up the things that are most likely to be a problem."
Instead of having to pore over months of payroll analyses, "we point them like lasers to stuff that really look different," Schreiber said.
"It really helps them to be productive in figuring out what to audit," he said.
Dax Craig, president of Denver-based Valen Inc., said predictive analytics can help the carrier know how to execute the premium audit in the right way. "We will help the insurer find the places where the insured owes them more money," he said.
He said that, other than the largest insurers, most companies lack the data volumes necessary to build effective models, and most lack the in-house expertise to conduct effective analysis.
"The software and hardware needed to build predictive models is highly specialized and usually carries a steep price tag," Craig said. "Each of these factors puts predictive analytics out of the reach of most insurers."
For example, a carrier may have 1,000 policies with different mixes of job class codes, but there are two policies with carpenters and managers on them that will look exactly alike to an auditor.
"But through predictive analytics we can tell one of these policies is different from the other. And so we would advise them to send a field auditor out there because what you will find is a bunch of carpenters and not carpenters and managers," Craig said.
Insurance carriers typically have general premium audit rules--such as all companies with roofers or all policies over a certain amount will gain special attention.
"So they will miss a lot of those mismatches, and we can pinpoint those policies that will have a likely mismatch," he said.
October 15, 2009
Copyright 2009© LRP Publications