DATA ENVELOPMENT ANALYS – EFFICIENCY ANALYSIS ON 17 MIDDLE-SIZED HOSPITALS IN SWEDEN

Abstract

Rising costs of healthcare in most OECD-countries have contributed to a quest for research into the field of healthcare costs and efficiency, something that has not been free of controversies. As a result of that, healthcare financers as well as providers have become more inclined to measure performance and compare themselves to others with the same responsibilities. In Sweden, the Association of Local Authorities and Regions, SKL, has since 2006 measured and compared the costs and production of entities in the healthcare sector in what is called “Öppna Jämförelser”, Open comparisons (SKL.se/oppnajamforelser). However, most of the comparisons have been between county councils.

In this thesis, the entities compared are hospitals. The technique used is Data Envelopment Analysis (DEA), in its different revised forms. The entities, i.e. hospitals are seen as Decision-making units (DMUs) that are compared in order to find out which hospitals that are efficient in relation to other hospitals and which ones that have potential to increase their efficiency levels. One of the aims of the thesis is to find whether the technique is robust and reliable. A revised version of DEA, namely Multiple-criteria DEA (MCDEA), is also used and compared with the classical one.

As hospitals in Sweden all are financed by county councils, their sizes and patient bases differ depending on how many people live in the region or sub-regional area and how many hospitals are active there. Some county councils that had not reported complete data on hospital level to Swedish Association of Local Authorities and Regions (SKL) have no hospitals represented in this study.
Of the 17 hospitals included, results show that the smallest ones face increasing returns to scale while the bigger ones face either constant or decreasing returns to scale. Spearman’s rank correlation tests show correlations between the efficiency ranks of the hospitals and their ranking orders in some other usually used indicators used in healthcare, such as length of stay, patient satisfaction rate, overcrowding and mean DRG-point.

Compared to the classical DEA, MCDEA performs much better and easing the efficiency score limit of 1.00 shows that the efficient hospitals get different scores higher than 1.00 and are thus discernible and rankable. It is concluded that DEA is a reliable and robust technique and the revised version, MCDEA, is better than the classical one.

Popular Abstract

I this master thesis, Data Envelopment Analysis (DEA) is used to compare 17 middle-sized hospitals in Sweden in terms of their efficiency in producing specialized healthcare. DEA is a non-parametric technique in which certain Decision-Making Units (DMUs) are compared to others using the same inputs and producing the same outputs. Finding the DMUs with optimal mixture of inputs and outputs is the main purpose of the method. As in some earlier studies, some deficiencies of the classical DEA model were detected, a revised version, MCDEA, is also used to find out whether that performs better.

The input variables are direct and indirect costs of producing specialized healthcare while the output variables are DRG points, in inpatient and outpatient care respectively. DRG (Diagnosis-Related Groups) is a classification system where procedures and healthcare contacts are grouped based on their resource use.

In terms of consistency and robustness, the DEA technique show satisfactory results. The efficiency scores of the hospitals are quite the same in both input- and output-oriented. The results show that the classical DEA has some shortcomings in terms of low discriminating power because many DEAs are deemed as efficient, some only for having the highest value in an output variable or lowest value in a certain input variable. The results from classical DEA also show unrealistic weight dispersion. MCDEA, on the other hand is efficient in solving the problem of low discriminating power among the Decision-Making Units (DMUs) but fails to solve the problem of unrealistic weight dispersion. One method which has shown promising signs of helping with discriminating power and rankability has been easing the upper limit of 1.00 so that relatively efficient hospitals can get super-efficiency scores.

The comparison analyses indicate that smallest hospitals in the paper face increasing returns to scale while the bigger ones either face constant or decreasing returns to scale. Between 2012 and 2016, the number of hospitals who face decreasing returns to scale, i.e. have larger sizes than optimal, have increased from a few to 7.

In the end, Spearman's rank-correlation indicate that efficiency scores of the hospitals have quite strong correlation, either negative or positive, with some external factors often used in healthcare, such as mean length of stay, overcrowding rate, National Patient Survey and DRG points produced per million SEK.

• author: Javid Gholam, Reza
• supervisor: Ulf Gerdtham
• organization: Department of Economics
• course: NEKP01 20181
• year: 2018
• type: H2 - Master's Degree (Two Years)
• subject: Business and Economics
• keywords: Data Envelopment Analysis, MCDEA, Hospital efficiency, Swedish healthcare system
• language: English
• additional info: The research purpose from the beginning was to include all middle-sized hospitals in Sweden. Due to lack of complete data for some middle-sized hospital, they were not added so the final number of hospitals in this paper is less than all middle-sized hospitals in Sweden. The author hopes that in the future, other researchers will do research on efficiency with all hospitals included. Hopefully this paper can be helpful.
• id: 8945280
• date added to LUP: 2018-07-03 13:36:54
• date last changed: 2018-07-03 13:36:54