| This study utilizes the Data Envelopment Analysis (DEA) methodology to evaluate the efficiency of a group of Wisconsin hospitals in providing inpatient routine services. Specifically, DEA's potential usefulness in directly limiting Medicare reimbursements and controlling hospital costs, given the existing Medicare cost report data base, is examined. In this context, two issues were considered of critical importance: the impact of (1) alternative variable sets (e.g., different aggregations, specifications, etc.) and (2) data variations (e.g. resulting from biased cost reporting, data accumulation errors, etc.), upon the DEA efficiency ratings.Our analysis also discloses that since hospitals are evaluated along their best dimensions (variables), ample opportunity exists for institutions to increase their efficiency score through manipulation of selected output/input data. Indeed, the sensitivity analysis results of our study reveal that individual hospitals could be quite successful in artificially increasing their efficiency ratings without actually improving productivity or reducing costs.It is concluded that in light of the existing political and noncooperative atmosphere of Medicare cost reporting, we would question DEA's usefulness as a direct method for limiting Medicare reimbursements to participating hospitals. Wherever incentives for noncooperative behavior can be curbed (e.g., using DEA for self-evaluation by individual institutions), DEA's potential for generating insightful and reliable efficiency assessments would be improved.Upon close examination of the DEA model, we discovered that several of the problems associated with existing hospital efficiency measurement methods are retained and in some instances magnified under DEA. Moreover, it was found that a hospital's DEA efficiency rating is an increasing function of the number of variables considered in the model. Hence, expansion of the variable set (either through disaggregation of existing variables or addition of new factors) produces a general upward trend in efficiency scores with a successively greater proportion of hospitals being scored efficient. In addition, these effects persisted even in the presence of highly correlated variables, implying that correlation criteria are of limited value in identifying redundant variables and contracting the final variable set. |
