Constructing a Hospital Efficiency Measurement Model Using DEA, DEA-ANN, CART and Malmquist Approaches

• Main Authors: Chien-Wen Chen, 陳建文
• Other Authors: Rong-Ho Lin
• Format: Others
• Language: zh-TW
• Published: 2008
• Online Access: http://ndltd.ncl.edu.tw/handle/zqwnh7

碩士 === 國立臺北科技大學 === 商業自動化與管理研究所 === 96 === The constant changes in health insurance payment method and the implementation of the new hospital accreditation system have triggered drastic impacts and challenges on hospital management in Taiwan. Hospitals are required to reduce medical costs and upgrade healthcare quality so as to enhance competitiveness and ensure sustainable development. Hospital efficiency measurement not only offers an understanding of the management effectiveness of a hospital but alos provide decision-makers with valuable references for optimal utilization of medical rsources. The study is divided into four parts. The first part adopts Data Envelopment Anaysis (DEA) to measure the relative efficiency of the hospitals that passed accreditation in Taiwan during the three years from 2004 to 2006. For hospitals whose efficiency needs to be strengthened, Slack Variable Analysis and Returns to Scale analysis are performed to understand the scale of and directions for improvement. In the second part, the DEA-ANN model is used to obtain a more objective nd reasonable efficiency score as the model integrates the non-linearity feature of neural network with the advantages of efficiency measurement method. The third part applies CART (Classification and Regression Tree) to establish relevant rules for measuring hospital efficiency and to build up the hospital efficiency measurement model. In the fourth part, the Malmquist productivity index are consulted to analyze the changes in the healthcare industry and to understand the changes in technical efficiency, technology, and productivity of the sample hospitals. The results show that: (1). The efficiency scores of primary, secondary and tertiary hospitals report a significant different only in 2004; there was no significant difference in 2005 and 2006. (2). The DEA-ANN model reaches an average efficiency score higher than the one obtained by the DEA model and a lower standard deviation, suggesting that the DEA-ANN model is able to reduce the underestimation of decision-making units frequently associated with the DEA approach. It can further rectify the inability of DEA to rank the efficienct decision-making units. (3). In terms of the indicators of hospital efficiency, the CART-generated classification tree reports the highest accuracy rate in 2006; this is also applicable to 2004 and 2005. The numbers of physicians, medical specialists, nurses and outpatients (and ER patients) appear to be the indicators most compatible with measuring hospital efficiency. (4). By the Malmquist productivity index, the average hospital productivity drops 3.7%. For the sources of changes in productivity, technical growth (-4.1%) and growth in technological efficiency indicate that technical/technological changes are major hindrance to the increase in productivity.