Use of Partial Least Squares - Structural Equation Modeling for Identifying the Most Important Variables via Application of Data Envelopment Analysis

Abstract

Background: Since the introduction of data envelopment analysis (DEA) in 1978, this method has become one of the best performance evaluation tools. Use of DEA has many advantages, as it can accommodate multiple inputs and outputs. However, there are some constraints concerning the number of variables. For instance, selective variables are commonly used to evaluate the performance of hospitals, based on the researcher’s personal preferences. Objectives: In application of traditional DEA models in hospital comparisons, variables are always selected based on the analyst’s opinion regarding the significance or availability of data. On the other hand, there is a need to reduce the dimensions of variables. The main goal of this study was to reduce the number of input and output variables in the evaluation of military hospitals of Iran. Methods: In this study, the Delphi technique was applied, as well as partial least squares - structural equation modeling (PLS - SEM) method. Results: Acceptable goodness of fit was established for validity of the measurement model. The test of validity was accepted for the structural model in this study. Conclusions: In the first step, the number of variables was reduced to 29, using the Delphi technique. In the second step, using SEM - PLS, the number of variables was reduced to half.