Measuring efficiency in acute care hospitals: an application of data envelopment analysis.

In this article, the authors attempted to demonstrate how DEA can be useful to hospital administrators and health care planners. They used actual data collected by the American Hospital Association through its Monitrend Data Service. Since these were national data, they are presented here for illustrative purposes only. The efficiency with which a hospital operates may well depend upon the local or regional labor market, the competition among health care providers in that market, and the demographics of the service area. The choice of variables was dictated by reasonableness and availability of data. Given the routine collection of case mix data by DRG since 1984, the use of a different set of output variables for any future studies would be quite appropriate. Additionally, if DEA were to be used, a consensus concerning relevant controllable and non-controllable input variables would need to be achieved. There are more technical caveats of which the reader should be aware. 1) The efficiency scores are all relative and are based on the performance of the other hospitals being compared; nothing can be said about the absolute efficiency of a given hospital. However, the relative ratings are conservative in that the approach "bends over backwards" to give the individual hospital the benefit of the doubt in terms of the relative importance of the various outputs and inputs utilized. The approach maintains equity in that any weights chosen for a given hospital must be feasible for all of the other hospitals. 2. The ratings assume a causal impact of the inputs on the outputs. In addition, it is possible that inclusion of additional inputs and outputs could modify the relative scores and/or help explain the differences. However, based on the factors available, any unit rated inefficient is inferior in a very real and demonstrable sense. 3. DEA is based on the generalized notion of convexity which assumes that the performance arrived at by taking any linear weighted combination of other hospitals' inputs and outputs represents a feasible and achievable technology. The general frontier surface is approximated by piecewise-linear segments with the result that observed differences in efficiency cannot be explained away as differences in economies of scale. 4. The inefficiency score and the resource conservation potentials are based on a unit's so-called contraction path, i.e., all of the controllable inputs are required to be reduced by the same factor.(ABSTRACT TRUNCATED AT 400 WORDS)