The purpose of this study was to compare health status outcomes in three alternative long-term care settings in the Veterans Administration : 1) home care; 2) community-based nursing home care; and 3) hospital-based nursing home care. Patients were measured on a behavioral index of health status, at two points in time: when transferred from the acute care hospital to one of the three treatment programs (pretest and three months later (posttest). Since patients could not be randomly distributed to programs, two methods were employed to control for potential sample selection bias: the choice of a nonequivalent control group design, and multivariate analytic techniques. First, within each program type, patients were randomly selected from both a hospital that offered only that program as a long-term care alternative and from a hospital that provided all three treatment settings. Second, multiple regression analysis was used to control for pretest differences among patients. Patients placed in the home care program displayed the greatest mean improvement in functional health status, holding all other variables constant. This treatment effect was not uniform, however; patients showed differential rates of improvement across the three programs, based upon both initial health status and prognosis.