Product-line evaluation of graduate medical education program costs.

This article reports the results of an innovative application of traditional multivariate approaches to estimating hospital costs in order to support product-line evaluation of graduate medical education (GME) program costs among the clinical departments and teaching facilities of a nationwide, federal multi-institutional system. Department-level data for 1988, 1989, and 1990 were used to estimate a multiple regression model of total costs per disposition for the specialties of medicine, surgery, obstetrics/gynecology, orthopedics, psychiatry, and pediatrics. Systemwide and facility-specific GME program costs per disposition were estimated for each specialty on the basis of dependent variable scores predicted by the regression model. Measures of case-mix intensity, facility bed size, department staff size, clinical specialty, GME status, teaching intensity, operating efficiency, and regional variation each made statistically significant contributions to the explained variance in total costs per disposition, and yielded an adjusted R2 of .701. Estimates of total costs and GME costs per disposition revealed substantial variation among clinical specialties, both systemwide and within specific facilities. The results of these techniques, their usefulness for enhancing executive ability to evaluate costs of GME programs as product lines, and their implications for public policy regarding hospital payments are discussed.