technique measures efficiency of hospital relative to other hospitals in the sample. Such benchmarking approach can help identify the best practices and provide potential target for the relatively inefficient hospitals, which cannot be identified by ratio and regression analysis. Third, DEA technique can avoid the risk of misspecification of production function, which is one major drawback of regression analysis, because it is a nonparametric technique that requires no specification of production function. Fourth, the technique also provides flexibility in data selection, i.e., different measuring units for inputs and outputs such as dollars, beds, discharges, and average length of stay. Finally, DEA technique does not assume behavioral assumptions like cost minimizing or profit maximizing behavior, and this is especially relevant to public hospitals that are not profit maximizer. DEA technique considers observations in the sample as production units and assesses how efficiently each hospital uses inputs to produce outputs relative to others in the sample. Considered as a production unit, the production function of hospital can be explained by the Input-Control-Output-Mechanism (ICOM) model of business process as illustrated in Fig. 1. This model has been used for hospitals by Lee and Menon (2000) to explain efficiency in hospital production function in the context of DEA. According to the ICOM model, a production process is a mechanism in which people and/or equipment use inputs to produce outputs. Controls are operational control such as process design, procedures, and rules that have control over the production operation. From this framework, inefficiency of hospital operation can be caused by poor production mechanism or weak control over the production process. For a given set of hospitals being evaluated, DEA assumes that all the hospitals have the same production frontier and all efficiency variation results from differences in managerial performance. The efficiency measurement of DEA technique can be illustrated by using the case of three hospitals producing a single output (y) from two inputs, x1 and x2, as shown in Fig. 2. The production frontier is
Fig. 1. ICOM model.
B. Watcharasriroj, J.C.S. Tang / Journal of High Technology Management Research 15 (2004) 1–166
defined by the hospitals that can use minimum inputs to produce a given output. This is illustrated by Hospitals A and B in this example. These two hospitals are considered efficient or exhibit best practice. On the other hand, Hospital C, which does not lie on the frontier, is the hospital that uses