RMIS-Web Risk & Insurance RMIS Article

"Technology Sidebar: The Case of the $800,000 Man"

January 1999

The case of the $800,000 man begins with a motor vehicle injury resulting in a concussion and neck strain. A 45-year-old man, who had been with his employer about three months, filed a workers' compensation claim. After the initial report of injury, claims arrived for a series of consultations, visits to a number of physicians, multiple MRIs, a CAT scan, and finally, a request for permanent total disability.

The Workers' Compensation Fund of Utah, which insured the employer, faced the possibility of paying close to a million dollars for this single claim. Accordingly, reserves of $800,000 were set aside.

An astute adjuster may have noticed the unusually high level of activity on the claim and that expenses were growing;signals that the claimant might be exaggerating his injury. Instead of leaving the responsibility of detection solely to a busy employee, the Workers' Compensation Fund employed a new predictive software solution, which almost immediately identified this claim as one to watch.

Developed by Costa Mesa, Calif.-based HNC Insurance Solutions, the software pinpoints suspicious claims not based on red flags that it has been programmed with, but based on experience. The software uses a neural network that is built from historical data and is able to learn to recognize which inputs have a significant effect on the prediction. The neural network automatically discovers not only additive linear relationships among data, but also iterative, nonlinear relationships.

The model identifies and analyzes hundreds of variables from the initial claims report, payment transactions and medical services details that may be indicative of fraud. The neural network is then "trained" based on historical data which allows it to recognize and weigh activity indicative of fraud. For each claim, it then compiles a score of 1 to 1000, with higher scores indicating a higher likelihood of fraud and lists the reasons why this claim should be examined.

The case of the "$800,000 Man," as he came to be known, was quickly spotted as a potentially fraudulent case when the score reached 500 in December 1997, a few months after the accident. The high score for this particular claim was based on a high rate of payments and activity, in particular for diagnostic radiology. Within a few months, the duration and cost of the claim had become excessive, compared with similar claims.

The case was monitored and investigated through the spring of 1998, at which point it was finally terminated. Through the use of predictive technology, an $800,000 claim was avoided.