Highmark, a provider of health insurance to about 4.6 million Pennsylvanians primarily through employer-sponsored benefit plans, walks a high wire when it comes to pricing its premiums. If it sets prices too high, it may lose customers to lower-cost competitors. But if prices are too low, the company's revenue erodes.

The group tasked with this balancing act is Highmark's actuarial department. Actuaries analyze incoming health care claims and make pricing recommendations based on forecasts of future trends. For instance, if the actuaries see a rise in claims for CT scans, they may recommend adjusting the premium to cover these costs. Of course, anyone who's been soaked in a rain shower when a weather report predicted sunshine knows forecasters don't always get it right. For Highmark's actuaries, a bad call results in a different kind of soaking--missing a 1% change in a trend can mean losing millions of dollars.

The $12.4 billion-a-year company believed timelier and richer data would lead to more accurate forecasts, so Highmark's actuarial and IT departments teamed up to create the Actuarial Trend System. ATS is built on an enterprise data warehouse using technology from Teradata, which ingests subscriber information and medical and drug claims from physicians, pharmacies, and hospitals. These claims come from a variety of application and database platforms.

The data warehouse provides two major benefits: It makes claims data available more quickly, and it reduces the amount of time actuaries have to spend assembling data for analysis. The data warehouse is fed daily by Highmark's automated claims process. The claims system automatically approves or denies a high percentage of the claims, says CIO Tom Tabor. "It minimizes errors and speeds up the process in payments going out," he says. The rest of the claims are sent to adjusters to resolve.

With more claims decided earlier, actuaries have fresher data to work with. Under the old system, the actuaries would wait until all the claims were manually resolved before they could be analyzed. This meant actuaries were working with claims data that was 2 or 2-1/2 months old, giving the company less time to identify and react to trends. The data warehouse also normalizes all the data coming in different formats into a single framework to make it easier for the actuaries to analyze the information. Before the data warehouse was built, the actuaries would have to write database queries against several different legacy database applications just to assemble the data to be analyzed. The sheer volume of records--about 1.6 billion--meant that actuaries were spending as many as 10 business days simply assembling claims data for analysis.

	CIO Tabor sees brighter (actuarial) forecasts Photo by Erica Berger

"It was so much information that we had to do separate extracts in pieces because we couldn't pull everything at once," says Dave Berry, actuarial director at Highmark. "Now I can ask a question of the system and get an answer in a day or two." Having access to data sooner lets actuaries spot trends earlier, Berry says. And this means actuaries spend more time on analysis, which is what they should be doing, says Tabor. Highmark estimates that in addition to getting better forecasts, the Actuarial Trend System saves $200,000 a year by eliminating time spent on database queries.

Highmark's IT and actuarial groups both understood that ATS was a top priority, but IT pros don't always speak the same professional language as the constituencies they serve. This was certainly the case with ATS. "Sometimes there were challenges to understand the requirements of the actuarial folks and translating it to what IT needed," says Del Gealy, director of financial systems at Highmark.

A key group in the actuarial organization, the Actuarial Infrastructure Support Group, built a bridge between the two departments. This group of business analysts knew enough about IT and the actuarial profession to help each department communicate its requirements to the other. "It was great because I could speak actuarial talk and not have to talk about relational databases," Berry says.