Skip to content Skip to navigation

Gains from Predictive Modeling

September 1, 2005
by Steven Coulter
| Reprints
Combined with data mining, the tool yields improved care and reduced costs

Chattanooga-based BlueCross BlueShield of Tennessee (BCBST) is always looking for new ways to use technology to improve delivery and quality of care for its 2 million members and to rein in costs. Applying proven medical predictive modeling techniques to member health and claims data mined from the administrative system is one such use.

Better care management
Seeking a way to enhance our members' continuum of care beyond simple case management, BCBST Director of Health Services Operations Sylvia Sherrill and I launched an initiative in 2001 designed to more fully utilize our care management products without increasing staff. Like nearly one-third of all Blue Cross Blue Shield organizations, BCBST uses Facets, the managed care administrative system from The TriZetto Group, Newport Beach, Calif., as its core claims processing engine.

Our entire commercial care management program and care coordination are set up on the system's medical management modules, giving staff access to consistent, real-time patient data and snapshots of claims histories. When predictive modeling tools are applied, we can make sense of the mountains of clinical information and pinpoint clusters of diagnoses, procedures and patterns of illness.

We can identify members likely to benefit from our care management programs and assign them to specific program acuity levels based on their needs. If a new condition develops in a member, predictive modeling is applied to determine which care coordination program is appropriate. Specialized member services programs on such topics as beta blockers, medication compliance, emergency services assessment and coordination focus on the patient's needs, avoiding disease progression and providing a high level of support. We have created case management templates for some cases, such as cardiology and urology problems.

The ability to mine data, apply highly effective analysis and accurately model care scenarios for more accurate program placement has significantly reduced expenditures of staff time and improved care. And analysis of this data allows us to calculate the return on investment for each program.

Cost savings
At BCBST, care management and disease management are viewed as a continuum. The goal is to identify the sickest members and quickly start them in suitable treatment programs. Referring only members who truly need and can benefit from a care management program has tripled our savings per case.

When we apply predictive modeling to a large population in aggregate, we have an 85 percent accuracy rate in predicting resource consumption. In analysis of individual members, our accuracy rate in predicting who will be high-cost members is 50 percent to 60 percent--a dramatic improvement over the 2 percent when only age and gender data were used. We estimate savings per case manager per month of about $116,000 in 2002 and about $291,000 in 2003.

Before the use of predictive modeling, 20,000 members went into the system for review every year. Now, about 5,000 cases per year are referred to the catastrophic case management department. The remaining 15,000 are more appropriately handled in other programs.

Additional benefits
We have used predictive modeling for more than three years, on data going back four years, and also use it for underwriting, rating and medical management, and to demonstrate cost savings to employer groups and provide accountability for insurance costs.

The TriZetto team has worked closely with our care management, utilization management, and information services departments to optimize its system as a decision management tool. The team designed an interface to consolidate key functions, so instead of having to open and search multiple programs for data, nurses can access complete information on a particular case with one click, without calls to providers for needed data. TriZetto helped us design translation tables that convert real-time claims data into summaries for nurses managing cases. With process engineering, the data is formatted to be usable by other BCBST staff as well.

Cost savings have allowed us to expand our care management programs. We use the system to compare predicted cost against actual cost of a case. And we produce reports assessing employer-group populations to help customers better understand their healthcare costs and estimate self-funded healthcare needs.

BCBST is consistently among the top 10 percent of Blues plans in member services. We believe this is due, in part, to providing staff with the finest IT tools available.

Steven Coulter, M.D., is President, Government Business and Emerging Markets, BlueCross BlueShield of Tennessee, Chattanooga.