Measuring and Determining Efficiency in Long-Term Care: Non-Parametric, Semi-Parametric and Parametric Approaches

Measuring and Determining Efficiency in Long-Term Care: Non-Parametric, Semi-Parametric and Parametric Approaches


Economics Seminar by D. Dineen (Limerick) - Ricardo (B31)
June 2018, Monday 18 (10:30 am)

 

****

Abstract :In particular, this paper explores different two-stage semi-parametric methods to determine technical efficiency in long-term care provision in Ireland.  Technical efficiency (TE) scores in the health sector literature are often estimated using either a non-parametric conventional Data Envelopment Analysis (DEA) or the fully parametric Stochastic Frontier Analysis (SFA) approach.  While the SFA technique accounts for noise in the data and also allows for an unbiased estimation of the factors affecting efficiency, this method requires specification of the functional form of a production or cost function.  On the other hand, DEA does not impose these restrictions.  However, the estimation of the efficiency determinants using the conventional DEA scores in the second stage may lead to biased estimates of both the TE scores and the efficiency determining variables.  This paper tries to fill the gap between the conventional DEA and SFA approaches and applies alternative semi-parametric methods.  The methods considered do not impose any functional form restrictions on the data, and they also account for random sampling variability and reduce the bias in both the estimated TE scores and the estimated efficiency determinants.  In this paper, we model TE in terms of an input distance function which allows us to estimate input-oriented technical efficiencies by investigating how much the input vector can be proportionally reduced while holding the output vector fixed.  Output is measured as total patient days, while inputs are measured as medical staff, non-medical staff and the number of beds in long-term care units. Furthermore, we use several efficiency determining variables which are divided into objective quality characteristics and conventional determinants such as size, ownership, location, chain, case mix and the age of the nursing home facilities.  Additionally, we split the sample into different groups: private and public nursing homes; chain and non-chain private units; and nursing homes located in rural and urban areas.  We find notable differences in the results between the conventional DEA and the semi-parametric methods.  We also derive some conclusive findings with respect to the estimated TE scores and the efficiency determining variables for the long-term care sector in Ireland.

***