Estimates of the annual cost of workers compensation insurance fraud range from $1 billion to $20 billion nationally. This disparity demonstrates the difficulty in pinpointing where, when and how often workers compensation fraud occurs.
Detecting workers compensation fraud is a big job that has gotten bigger. According to the Coalition Against Insurance Fraud, all forms of workers compensation have grown "significantly" since the early 1980s. That includes worker claims fraud and employer premium fraud. But a new tool designed to help detect both types of fraud has proven more than effective, nearly doubling the fraud recovery rate for the Workers Compensation Fund of Utah (WCF-Utah).
Before implementing HNC Insurance Solutions VeriComp Claimant and VeriComp Employee solutions in 1998, WCF-Utahs fraud recovery rate was between $3 million and $4 million. That rate increased to $7.3 million last year.
"To a lot of companies, that might seem like a drop in the bucket," said Bob Short, senior vice president of WCF-Utah. "But [we have] around $120 million in premiums, so when you talk about $7 million that youd otherwise be paying out, youre talking about 5 or 6 percent probably."
A less measurable effect of the new technology is the deterrence factor. Individuals or employers in Utah know they are being watched more closely and are, therefore, less likely to commit fraud. "The significance of this fraud program is that the employers in Utah pay less money for workers compensation insurance because were cutting out the fraud," Short said.
The technology has been adopted by other states as well, including North Dakota and Nevada.
Nabbing the Culprit
Predictive technology works by sifting through claims and "predicting" which are likely to be fraudulent. After searching for any of 31 data elements that might be indicators of fraud, the software then provides a score from one to 1,000 based on the assessment of the data. A higher score indicates a higher probability of fraud.
The data elements examined in VeriComp Claimant are based on statistics derived from legitimate employee claims. They include the length of time it should take for a particular injury to heal; the age and sex of the claimant; the nature of the injury; and how much money the benefit is paying. This data is compared to the norm and anything out of the ordinary is examined more closely.
VeriComp Employer looks at employer information, including size and industry category, the reported payroll information of a company compared to its peers and changes in employer characteristics over time. VeriComp Employer data elements target claims that tend to be suspicious.
Although a high score by the VeriComp system doesnt necessarily mean fraud, it does give adjusters a target on which to focus. This is a far cry from the "old days" when selective random auditing was about the only way to detect fraud.
"A normal adjuster may work on about 150 to 200 claims, conservatively," said Laki Balaji, vice president of property and casualty predictive software of HNC. "On any given day, theyve got probably 100 bills. Its humanly impossible to check every claim, and thats the power of this neural network technology; its able to analyze an innumerable amount of data elements at a given snapshot in time, which the human brain could never do."
Weaving into the System
WCF-Utah has about 28,000 employer policies, 15,000 to 20,000 of which are very small policies, said Short. Previous to the VeriComp systems, the only way to detect fraudulent claims was to do random audits on the companies with the largest policies. There werent enough resources to check on the rest.
"It is a huge improvement in our allocation of resources," Short said. "When we [did] selective random audits, we used to get a recovery or payback on 60 percent of the cases that we audited. Now we get a 90