Measuring cost efficiency in VA Hospital Pharmacies: An applied critique of the methods.

Abstract

This research provides a critical evaluation of the empirical methods available to measure firm-level Farrell cost efficiency measures in a specific empirical environment--Veterans Administration (VA) Hospital Pharmacies. The empirical methods available can be generally divided into two main groups--parametric frontier techniques, and nonparametric frontier techniques. Each alternative method maintains a set of assumptions that may or may not coincide with the underlying data generating process found in a specific empirical environment. The parametrically-based measures restrict the underlying technology to follow a specified functional relationship, but enable the researcher to directly specify variations in the production environment that are not included in historical input and output counts. In contrast, the nonparametric measures are free of parametric functional assumptions, but require that all measured production information be contained in historical input and output data. The parametric-based cost efficiency measures were estimated from a system of conditional short-run input demand equations derived from a "generalized" additive general error cost model. The underlying cost function was specified using a multiple-output generalized-Leontief cost function. The nonparametric cost efficiency estimates were derived from an input cost-based "goodness-of-fit" search algorithm. This search algorithm is generally based upon the revealed preference production theory known as the Weak Axiom of Cost Minimization (WACM). This research finds that empirical estimates of VA hospital pharmacy-specific Farrell cost efficiency measures are a function of the level of output disaggregation, and the method used to incorporate further production information into the analysis. The estimated models demonstrate the importance of controlling for additional production factors outside the direct control of the chief pharmacist. Similar pharmacy-specific Farrell cost efficiency measures were obtained using parametric and nonparametric estimation techniques that did not attempt to control for these factors. The correlations between the pharmacy-specific parametric and nonparametric estimates fell with attempts to control for these additional production factors. The decrease in correlation is attributed to the variation in methods used to incorporate additional production information into the parametric and nonparametric techniques.