- 學位論文
以DEA 方法與演化式類神經網路探討醫療品質與技術效率間之關係
Investigating the Relationship between Medical Quality and Technical Efficiency by DEA and Hybrid Neural Network
摘要
台灣實施全民健保以後，醫療市場結構驟變，醫院經營日益競爭。因此各醫院無不亟思改善其經營體質，提升醫療品質以追求卓越之道，進而強化其競爭力。 本研究分為兩個部份， 首先我們使用資料包絡法(Data Envelopment Analysis，DEA)探討民國93 年度台灣地區69 家區域醫院相對效率。並分別計算各決策單位(Decision Making Unit，DMU)其技術效率及純技術效率，並藉由差額變數分析，具體探討各醫院無法達到相對效率之主要原因，提供相對無效率的DMU 其投入及產出項改進方向與幅度。第二部分，我們使用演化式類神經網路 (Hybrid Neural Network，HNN)進行評估，採用演化式類神經網路進行效率評估，由於此模式運用類神經網路之非線性特性，並結合各效率衡量法的優點，因此經由此模式所建構的效率前緣，乃為更客觀的真實效率前緣。 本研究結果發現: 1.品質變數的加入會明顯影響效率值及差額變數，因此，在效率衡量時，應加入品質變項，以使效率評估更加完整2.使用演化式類神經網路計算效率值，其平均值比DEA 方法相對來的高3.兩者方法具有顯著性差異，代表方法不同，可能造成效率結果的不同4.使用HNN 將可以將可以進一步解決DEA無法處理有效率單位排序的問題。 經由本研究之分析結果，不僅可瞭解醫院之相對競爭力，以提供醫院改善的參考，並可提供醫療主管機關研擬政策之參考依據。
並列摘要
After Taiwan implemented the National Medical Insurance, its Medicare market structure experienced some changes. Competition among hospitals have also become stiffer. As a result, hospitals have left no stones unturned inupgrading their operations, the quality of medical treatment and services offered also pursued the path to excellence. This new development increased the competition in the field. In addition,The healthcarequality is the important lesson to managers,consumers,and public healthorganization inrecent years. In this study we use the non-parametric Data Envelopment Analysis (DEA) and Hybrid Neural Network(HNN) methods to measure the impact of joining quality index on the production efficiency of regional hospitals. First, we use a DEA approach to measure relative efficiency scores on a sample of 69 Taiwan’s regional hospitals. We use individually the model of CCR and BCC to assess the Overall Efficiency and Technical Efficiency. Also, we use the Slack Variable Analysis to confer the main reasons why the hospitals can’t reach the relative efficiency and offer the way to improve about the inefficient input and output of DMU . Second,we use neural network to measure relative efficiency scores. Most of the scholars have used the efficiency analysis to analyze management efficiency in the past, however, the limitation of this theory makes the evaluation results deviated. Here we use hybrid neural network to measure relative efficiency. This function applies a non-linear character of neuralnetwork and then link up advantages of other efficiency measurement methods, hence, the efficiency frontier which is produced by this function is much more objective. Our results include the following (1) It’s robust method of measuring efficiency and slack variable of hospitals to intro duce quality index. (2) the average efficient value which was calculated by HNN was higher than DEA. (3) two method is significantly different in efficiency value. (4) Use HNN will solve DEA model is unable to compare its efficiency rank. The results of this study can provide an understanding of the competitive advantage of these hospitals, as well as useful information for future upgrade purposes.