Reimbursing hospitals on the basis of treated cases, as in the New Jersey diagnosis-related groups (DRG) experiment, is equivalent to a centrally set pricing scheme, with all of its inherent difficulties. In addition to the problems of appropriate case definition, it is not obvious how hospitals should be classified to form reference groups for cost determination. Because empirically derived cost schedules are based on observed treatment patterns and resource use, they reflect variations in clinical appropriateness and quality and in resource use efficiency that characterize the system from which the data are drawn. If case-based schemes are to incorporate desirable performance incentives, they must be much better specified and take into account the complexity of hospital behavior. This article identifies the basic components of case-based systems of hospital reimbursement and discusses the analytic and empiric problems involved in their design.