GTE Laboratories has built an advanced DM system to evaluate health-care utilization costs for GTE's employees and dependents. Health-KEFIR (KEy FIndings Reporter) pinpoints groups whose costs are likely to increase in the coming year, finding areas where specific intervention strategies are likely to save the most money. The system can slice and dice the data in thousands of different ways -- by business units, age groups, or types of care, for instance. It can consider multidimensional factors for each subcategory of data, such as payments per day per hospital stay within various segments. (See the figure
"The KEFIR System."
)
Health-KEFIR selects only those medical conditions that are "interesting"; in other words, those
for which there is a known procedure that improves health outcomes and decreases costs. An increase in normal pregnancies, for example, will not be flagged, while an increase in premature births will be flagged, since there are standard medical interventions in prenatal care taken to reduce the premature birth rate.
"KEFIR's reports are more comprehensive, generated in less time, and significantly cheaper than comparable medical consultant reports," claims Gregory Piatetsky-Shapiro, principal investigator of the Knowledge Discovery in Databases Project at GTE Laboratories.
KEFIR is written in tcl and C, with a SQL interface to ensure portability. It currently runs on a Sun SparcStation 20 workstation with Informix DBMS. Instead of a paper report of hundreds of pages, KEFIR's output is on the network in hypertext markup language (HTML) and GIF files, accessible by Web client software such as Netscape's Mosaic. Deployment to GTE regional managers across the country is scheduled for late 1995. GTE i
s now evaluating proposals to turn Health-KEFIR into a commercial product. It is also considering applying KEFIR to marketing and customer-analysis tasks.
Southern California Spinal Disorders Hospital in Los Angeles is using IDIS (IntelligenceWare) on PCs to discover subtle factors affecting success and failure in back surgery.
A coach in the U.S. Gymnastics Federation is using IDIS to discover long-term factors that contribute to an athlete's performance. This information will be used in order to treat potential problems early on.
illustration_link (7 Kbytes)
KEFIR zeroes in on "interesting" health-care trends, then uses it
s findings to suggest cost-saving interventions.
Health Care
