Why do we know so little about insurance fraud? Why, after so many years of effort to control workers' comp fraud and abuse by claimants, medical providers and employers, do we know so little about how we're doing? The articles I read, the speeches I hear at conferences, amount to "lots of noise but no one coming downstairs," as the Chinese saying goes. No one seems to share data about fraud and abuse trends.
We could do better. No one really knows if a strategy works. You can't say you are making an impact unless you compare your results with an industry baseline. This has not kept insurers from pressing on. Erie Insurance has been refining a detection system for workers' compensation and other lines. It knows the challenges of workers' comp, in part thanks to its work in other lines where detection is easier.
The National Insurance Crime Bureau released in June a trend report on suspicious claims for property/casualty lines for 25 metropolitan areas and by type of potential fraud. It reported raw counts only. I hope it will soon report incidence rates.
I asked Dennis Jay, who runs the coalition Against Insurance Fraud, what he thinks about the measurement problem. A coalition white paper spells out a clear-eyed, ambitious program to measure fraud. He told me, "Your contention that we know little about insurance fraud is basically correct."
"Like many financial crimes, insurance fraud is complex, difficult to detect and, thus, nearly impossible to quantify," he said. "Research is sorely lacking about many aspects of insurance fraud, including measuring the extent and broadly testing potential solutions. Closed claims studies give us a glimpse into the frequency of red flags."
He went on. "Technology such as predictive modeling is beginning to add to the knowledge base. And the statistical study conducted annually by the coalition provides trend lines about such things as referrals received by fraud bureaus, fraud investigations opened, arrests, convictions and how much restitution courts are ordering. These are all pieces of a three-dimensional puzzle that will never be filled in completely."
I have spoken numerous times with one of the leading providers of fraud detection software for workers' comp claims payers. One of its conference booth people can spot me approaching from 100 feet away. Another supplier of fraud detection software told me they don't have any data on the incidence of fraud or even suspicious claims.
These firms cite two reasons for the absence of trend data. First, each client of these vendors has its own interpretation of what fraud and abuse mean. Second, they say it is not feasible to estimate the extent of the problem because you can't know what you can't know.
This doesn't cut it. If you talk a lot about fraud and abuse, you should be able to use your terms in a way that adequately takes into account disparities of definition. A decade ago, a leading authority on insurance fraud, Malcolm Sparrow, presented a method to estimate the extent of fraud and abuse.
His "License to Steal" is a stop on the road for anyone wanting to research fraud detection methods. He describes how you can make some reasonable estimates by examining known cases within a complete database of all claims. The Insurance Research Council underwrote in the 1990s a major study for auto claims, using the same method. Drawing upon Sparrow's work, the California Department of Insurance will hire a research team to estimate workers' comp medical overpayments and underpayments.
Bring on these kinds of studies, and share what you find!
PETER ROUSMANIERE is a Vermont-based writer and columnist for Risk & Insurance®.
August 1, 2006
Copyright 2006© LRP Publications