Some insurers see their data from new perspective

Or they might find there are segments of their business that can be automated. Others use analytics to look at claims and get better at fraud detection.

MetLife gets a jump on fraudsters

Grange Insurance isn't the only insurer to undergo a transformation in the way it views its data.

Five years ago, investigators at MetLife Auto & Home had to rely on hot-line calls from disgruntled spouses or the eagle eyes of claims handlers to spot possible fraudulent activities.

"There was very little beyond that," says John Sargent, director of the MetLife subsidiary's Special Investigation Unit (SIU). "And even though we were aggressive and felt we had a pretty decent program, we knew we were missing a lot of things we should have caught."

Convinced that predictive analytics could help, MetLife began working in 2002 with vendor Computer Sciences Corp. to develop a product called Fraud Evaluator to help flag suspicious claims. The product combines predictive models, MetLife claims data and industry data sources to help underwriters flag suspicious claims.

"It sweeps all claims through a system that compares them to see how close they are to years of prior claims we know were fraudulent," Sargent explains.

For instance, early on in its use of the product, MetLife identified suspicious patterns of activity between medical clinics and durable medical supply companies and was able to stop payments to suspected fraud rings. This year, MetLife added to its arsenal a predictive analytics engine from SPSS called PredictiveClaims, which analyzes all claims entering the system against risk profiles and external fraud databases. The system will either approve a claim for processing or flag it for investigation.

Sargent adds that since MetLife began using Fraud Evaluator, the SIU has seen a twofold increase in the number of investigations it runs. Meanwhile, the number of days it takes for that claim to get from the adjuster to investigators has been cut in half. That also translates into closing claims faster, Sargent says.

Despite the obvious competitive advantage the business intelligence tools have given MetLife investigators, Sargent knows quite a few insurers aren't yet using predictive analytics and he attributes it in part to data quality issues. Many are still using "green screen" mainframe systems where data isn't easy to sort.

"We had our own data protocol issues when we first started," he says. For instance, until investigators explained the use of the analytics software, claims employees hadn't been putting information such as the name of a person's doctor in the appropriate field in the database because they hadn't seen a need to do so.

Sargent says that the fraud detection process still requires talented adjusters and investigators, but the technology has made a huge difference in the first step of the process. "Without something like this to prompt us," he says, "we were still missing a lot."

Health insurers get predictive

Like property and casualty insurers, health insurers have their own complexities to deal with when it comes to enterprise views of their data. Besides dealing with claims, they have to handle diagnostic and procedural coding as well as data from providers, employers and members. "A provider code might be seven digits in one system and eight in another," says Rick Ingraham, healthcare industry principal at SAS. In moving to business intelligence, the challenge to health insurers is how to create the metadata on which to do reporting and analytics, he adds.

The insurers turning to SAS are currently most interested in doing predictive modeling of patient use of the newer high-deductible health plans and health savings accounts and to plot which new products to offer around the Medicare Part D prescription drug program.

"Insurers are recognizing that their core business is shifting to information," Ingraham says. They are becoming leaders in the healthcare arena in recognizing patterns of care and using evidence-based medicine to push for improvements.

Some health insurers that have cleaned up their data are using analytics to find ways to cut costs through patient interventions.

"They use predictive modeling to not only identify people at risk but also measure outcomes in chronic disease management," says Diane Laurent, senior VP at DxCG, which counts 24 Blue Cross plans among its customers. Health plans use DxCG's predictive modeling software to improve their pricing ability and to provide employers information about the health status of employees and cost drivers, she adds.

For instance, in 2003 Capital Health Plan started using DxCG's software to identify patients with multiple chronic diseases and created a Center for Chronic Care, which offers comprehensive care, including mental health services, to that population. Making sure patients are taking their medications and seeing providers is already improving outcomes and driving down costs considerably, the health maintenance organization says.