Estimates of annual fraud losses in the healthcare insurance sector range from $60 billion to $250 billion. The Insurance Information Institute estimates that property and casualty insurance fraud accounts for another $30 billion in annual losses.
Fortunately, the emergence of big data, combined with predictive analytics and link analysis, offers insurers a powerful weapon for fighting fraud. While many insurers have yet to leverage big data, forward-looking carriers stand to reap enormous benefits by embracing it.
Predictive analytics help combat insurance fraud by identifying patterns in claims that are indicative of fraud. Analytic models analyze transactional and relationship data, enabling insurers to uncover formerly unknown types of fraud, identify ongoing fraud schemes and discover fraud networks.
[4 Ways Big Data is Changing Claims]
It would seem obvious that more data would only be helpful in this effort. Big data, however, sometimes presents too many potential paths and overwhelms an insurer's ability to sift through the really meaningful data. To address this challenge, insurers must move toward greater automation and sharpen the filters they apply to their data.
It is impossible to manually examine the terabytes (even petabytes) of data that insurers can now collect and store. For example, a health insurer may want to examine the time patients take between visits to an orthopedic specialist to determine if the time interval is predictive of fraud. But what interval should the insurer attempt to correlate with fraud? Two weeks? Three months? It is not possible to manually study every time interval listed on millions of claims to discover new fraud markers.
Without the proper filtering, even highly automated insurers find themselves chasing dead ends. Insurers must invest -- either in-house or through outsourcing -- in analytic expertise to filter big data. Insurers need to know what questions to ask in order to focus their antifraud efforts. Only by asking the right questions can insurers get meaningful answers and avoid the distractions and noise in big data.
Link Analysis: Stretching Big Data Wide
When a reviewer is examining an insurance claim, it is helpful to see that claim in the context of the bigger picture. This is where link analysis comes in. Link analysis is a data- analysis technique that examines relationships among organizations, people and transactions.
Link analysis works by ferreting out related claims that may not appear to be related. For instance, a suspicious auto body shop may be handling an unusually high number of accident repairs. Link analysis might show that the body shop is not the culprit -- or at least, not the only culprit. There may be a pool of crooked attorneys, "victims" and vehicle owners who take vehicles to the same shop. When viewed individually, each claim may look legitimate. But when viewed in a broader context, the fraudulent pattern becomes clear.
Unlike predictive analytics, there is no concern about data overload for link analysis. In fact, link analysis is a data-hungry process that is bolstered by big data's broad reach. The key to link analysis is identifying relationships across as many sources as possible. In the future, those sources may even include GPS devices and social networks. More data yields more information about more relationships, which helps insurers recognize potential collusion.
When big data is added to predictive analytics and link analysis, insurers are able to detect more fraud, reduce false positives by accurately identifying real fraud, and improve customer satisfaction by streamlining the payment of legitimate claims. Big data is a potential game changer that could save insurers billions of dollars.
While the industry's track record of embracing new technology is mixed, the potential payoff from big data is extremely compelling. But to reap the benefits of big data, insurers must do more than build larger databases; they must invest in the technology and expertise needed to apply big data in an efficient manner.
Scott Horwitz is a senior director in the insurance practice at FICO. He is an accomplished mathematician with more than 23 years of experience working with the insurance industry.