摘要：Data Envelopment Analysis (DEA) is a non-parametric linear programming-based technique used for evaluating the relative efficiency of homogenous operating entities on the basis of empirical data on their inputs and outputs. It is suitable in cases where other approaches do not provide satisfactory results. The aim of this paper is its application in measuring the efficiency of 63 hospitals in the Croatian health care system on the basis of two inputs and two outputs. Comparison of the results of basic models having constant and variable returns to scale has shown the relevance of BCC model, and input-oriented model has been chosen. DEA identifi ed efficient hospitals as benchmark members and those inefficient to be analysed as candidates for reorganization. Sources and amounts of relative inefficiency, which were identified in each input and output, establish guidelines for needed improvements. As an extension of the basic model, we introduced categorical one that takes into account different levels of services provided by hospitals and isolate the impact of their unequal position in efficiency results. Analysis showed no significant differences between results which leads to conclusion that the level of services has no great impact on hospital performance if it is evaluated through here selected inputs and outputs.
