The purpose of this study was to explore, from readily available data, the best predictors of medical claims costs at Ball State University (BSU). Multiple regression equations were developed to predict BSU's medical claims costs from selected demographic and health-related measures among 1,799 BSU employees. The predictors were chosen from data previously collected during the three-year study period: July 1, 1995 to June 30, 1998. Regression equations were developed for the entire BSU population and the high-cost population. The linear composite of number of emergency room visits, number of chronic health problems, gender and age predicted 17.86% of the variation in transformed medical claims costs for the entire BSU population. The linear composite of age, number of emergency room visits and gender predicted 9.95% of the variation in transformed medical claims costs for high-cost BSU employees. Logistic regression, performed on the entire BSU population, did not differentiate low and highcost employees well: only 15.0% of high-cost employees were classified correctly. / Fisher Institute for Wellness