Some insurers see their data from new perspective

Other health insurers are using BI tools to open their data warehouses to customers. Blue Cross Blue Shield of Tennessee uses Cognos PowerPlay to allow 80 large employers to access information about their accounts as well as tools to create their own reports via the Internet.

"We can give them information on utilization patterns so they can create proactive health programs to reduce costs," explains Frank Brooks, chief data architect for the insurer, which has almost 3 million members. "Being able to offer them more information than other insurers is a key competitive advantage." And the BI tools allow employers to get those answers themselves and get them much more quickly than they ever could before.

He says the next step will involve offering customers predictive analytics tools so they can do what-if scenarios. "They want to see, for instance, if they open a plant in Dallas, what the impact on their utilization costs would be," Brooks says. They also want to access more benchmarking data to compare their company's utilization and costs to others in their industry and region.

Pattern recognition is key

Highmark, which owns insurance companies affiliated with Blue Cross Blue Shield in Pennsylvania and West Virginia, uses predictive analytic software from SAS to identify under-diagnosed people, because with Medicare HMOs the reimbursement from the federal government depends an accurate diagnosis and coding.

With a regression tree, Highmark's analysts can create propensity models around 40 to 50 different diseases to pinpoint people who appear to have a certain condition based on all their claims and drug information, explains Chris Scheib, manager of data mining and pattern discovery. For instance, a diabetic person might be insulin dependent and have renal failure but doesn't have a diagnosis. If that's the case, they are flagged and a summarized file is sent to a senior product revenue management office, where nurses go over medical reviews.

"Pattern recognition is important to us," adds Shawn McNelis, VP of healthcare informatics research and analysis. "There are a significant number of dollars involved. Last year we generated $8 million in additional revenue for Highmark in finding these under-diagnosed people."

Highmark uses predictive analytics to find over-diagnosed people as well. To avoid errors and audits by the federal government, the insurer runs models to find aberrations from the established norm.

"There are certain conditions where people don't just suddenly get sick, such as renal failure," Scheib says. "Usually there's a litany of claims beforehand; there's a trajectory. If a renal claim just pops up even though the person has been perfectly healthy over the last five years, that's a red flag." Scheib says advanced computing power at the desktop and the introduction of predictive analytics has revolutionized what his department is able to do and how quickly they do it. "With these new tools you can take one analyst and generate the type of research that would not have been possible five or six years ago," he says. "We didn't have the computing power on our desktops to handle the permutations to spin through the models. It would have been a fool's errand to ask a single analyst to go through 2 to 3 million people with a decision tree model looking at five diseases. Now that is completely feasible."

And McNelis says the use of business intelligence tools just reflects the larger strategic goals of the organization.

"There is value embedded in the data, and the degree to which you can exploit that impacts the kind of information managers use for decision making," he says. "We use predictive modeling for many things and by applying these technologies we are generating dollars-and-cents value for the company. If you fail to use it to its full advantage, you're leaving money on the table."