幾年來，台灣經濟的高度成長以及社會的快速發展，使得國民的健康需求跟著提升，而健康需求的引伸性需求──醫療服務需求也隨著大幅增加，醫療事業早已成為國內重要的產業。民國八十四年全民健康保險全面開辦，醫療支付制度的改變，加上不斷增多的醫療院所間日益激烈的競爭，導致醫院的經營遭逢莫大的衝擊和挑戰。因而，如何因應健康保險制度的變遷、制定有效的經營策略，便成為醫院經營團隊最重要的課題。 本研究旨在探討近年來醫院經營環境演變下，大型醫院應採行的策略與發展方向。有關分析乃以醫療產業經濟理論為基礎，以我國健康保險制度之變遷為主軸，配合相關大環境及產業環境與結構分析，然後以一家大型醫學中心為例，針對其組織內部的優勢與劣勢，以及外部環境的機會與威脅，找尋有利的經營策略。資料蒐集除國內外相關文獻外，主要是與個案醫院高層決策者以及醫院幕僚人員進行非結構式的深度訪談，並作深入的分析比較。 研究的結果發現：1.個案醫院應該朝老人醫學領域發展，以建立能夠凸顯其優勢的經營策略。2.個案醫院應採多角化經營策略，同時開發自費及不受健保給付影響的醫療市場。3.個案醫院仍應繼續其差異化服務的競爭策略，以維持固有的競爭優勢。4.個案醫院應將旗下業務部門，依屬性與功能劃分為數個策略事業單位，同時發展一套嚴謹可行的策略規劃程序，以利總體策略之制定。5.醫院應尋求機會爭取制度的鬆綁，讓醫院能夠有更大的彈性重新整合內部資源，以強化醫院的經營體質與增進管理效率。

由於資訊科技與網路的快速發展與普及，打破了時間與空間的限制，促成電子商務蓬勃發展。西藥產業中各行業如醫院、診所、藥局或藥商等雖然已普遍做到電腦化，然而其企業間卻依舊多採傳統方式交易往來。其無效率的效率的處理方式，促使面臨散於各地的資料不容易相互流通，企業無法能夠彼此交流、分享彼此有用的資訊。對此唯有導入電子商務的應用，強化產業上下游間資訊流通，才得以提升整體產業資源有效運用。 如果要以電子商務取代傳統交易方式，其發展的關鍵將不僅只在於促進資訊科技與網路的發展，而是要先瞭解整體西藥產業特性與產品特性，才能規劃出符合產業特性之電子化應用方案。本研究藉由個案訪談方式，來瞭解產業目前電子化現況與其電子化過程中發生的問題，進而嘗試以健保局扮演主導者，推動西藥產業後續電子化。本研究提出健保用藥「包工不包料」方案，配合電子交易市集與倉儲物流技術的應用，由健保局帶領各級醫療院所、藥廠與藥品物流業進行產業電子化。 而健保用藥「包工不包料」方案的提出，對於西藥產業與健保制度而言是一個嶄新的方案，將有助於改善藥價黑洞。在推動此方案同時，應考慮執行相關配套方案，如調整健保給付標準、制訂標準給藥量...等等政策的實施，以免產生可能的弊端。最後考量實際環境本研究提出具體的執行步驟，以供健保局在未來實行「包工不包料」方案的參考。 關鍵詞：全民健康保險制度、西藥產業、產業電子化

本論文對於醫療市場的生產績效研究，係由兩篇獨立的學術研究報告所組成，研究重點在於利用非參數資料包絡分析方法 ( nonparametric DEA approach )，估計醫院的生產技術，以衡量醫院的技術效率及不同年度間之生產力變動，進而分析不同醫院間，生產績效差異的主要原因。本論文所採用的研究方法與探討的主題，不同於國內既有的相關文獻。 第一部分：生產不確定性與醫院效率 本部分主要探討在醫院面對不確定性時的效率評估。一般而言，醫院有兩種生產上的不確定性來源：醫師或醫院的診療結果所導致的生產不確定性；及消費者對醫療服務需求的不確定性 (Arrow, 1963)。當醫院面對生產不確定性時，醫院效率將與廠商如何處理不確定性問題有關，亦即，當廠商事前規劃愈縝密，未來可能的產出失靈水準愈低，則其生產效率表現愈佳。本文利用民國 82 及 83 年（準）醫學中心與（準）區域醫院資料，模擬醫院在面對生產不確定性時，各種可能的產出失靈水準，以chance constrained DEA 模式 (Land, Lovell and Thore, 1993) 估算醫院的隨機技術效率，並與傳統、確定性的DEA模式所得到之結果，做一比較。 Chance constrained DEA模式與傳統DEA模式的不同，在於前者估計出的生產前緣，並不總是包絡所有的樣本點，亦即，允許某廠商之產出超越生產前緣或說允許產出失靈可能性之存在，而後者則否。實證結果發現，在chance constrained DEA模式下，私立醫院的技術效率高於公立醫院，且呈現統計顯著性的差異，但兩者間的差異隨著醫院事前準備程度的提高而縮小；而傳統DEA模式也顯示，私立醫院的技術效率確實顯著地高於公立醫院。此外，若產出失靈水準夠低，則chance constrained DEA模式的效率值與傳統DEA模式的效率值，兩者間的分配會呈現統計顯著性差異。 在面對生產不確定性時，欲提升公立醫院的生產效率，應提高廠商事先規劃的程度，才能與私立醫院之生產效率並駕齊驅。一般而言，廠商事先準備的程度高低，與醫院本身的特性有關，因此，欲改善公立醫院緩衝產能的準備程度，以降低產出失靈水準，有必要進行體制層面的改革，亦即，從進行人事變革、財務之授權與彈性化等方向開始做起，如此應可提高公立醫院的生產效率。 第二部分：全民健康保險制度與醫院生產力變動 全民健保實施後，民眾對醫療服務的可近性提高，醫院間的市場結構改變，因此，醫院生產力與效率的提升，成為眾所關切的焦點。為瞭解醫院在全民健保實施後，資源是否有效配置，本部分利用民國 82 至 86 年醫學中心、區域醫院與地區醫院等大小型醫院資料，以範疇DEA模式估計Malmquist生產力變動指標，並將之分解為技術變動、純技術效率變動、及規模效率變動等三項變動來源。 實證結果發現，從82至86年醫院整體平均效率而言，CRS（VRS）生產技術下的平均效率為 66.00%（74.87%），表示不論大小型醫院，平均而言，皆存在技術不效率的情形。再者，在民國84年，亦即全民健保實施的年度，其效率水準明顯較其他年度為低，其餘年度的效率水準都相對較高，此一結果意謂，政策干擾對於醫院效率表現的影響，是短期性的。另外，小型醫院皆較大型醫院不效率，兩者的效率差異呈現統計顯著性；以權屬別而言，不論是大型醫院或小型醫院中的私立醫院，其生產效率均優於公立醫院，且兩者的效率差異呈現統計顯著性。而透過迴歸分析顯示，全民健保實施、權屬別之虛擬變數、佔床率、平均住院日、及以醫院產出衡量的集中度指標等，是影響醫院生產效率的重要因素。 從Malmquist生產力變動（　et al., 1994）來看，平均而言，82-86年間醫院生產力成長率約在 -3.06 % 左右。就生產力變動來源而言，技術成長率（-2.74 %）與整體效率成長率（-0.33 %）均為負，而技術變動則是阻礙生產力成長的主要原因。此外，若以醫院整體效率變動來源來看，平均而言，整體效率退步是由於規模效率變動所致（-0.74%）。 此外，本文著重在　et al.(1994)、Ray and Desli (1997) 及Grifell and Lovell (1998) 三種定義下的Malmquist生產力變動指標之比較。研究結果發現，Grifell and Lovell (1998) 的一般化Malmquist生產力指數，並沒有正確衡量廠商的生產力變動及其變動來源項。而利用Kruskal-Wallis檢定結果發現，三個模式中的生產力變動差異，並不具統計顯著性，而變動來源項（技術變動與規模效率變動）亦顯示相同的結果。 / This dissertation is focused on the efficiency and productivity studies of hospitals in Taiwan. It includes two independent academic papers. The primary intention is to introduce the newly developed ideas in the measurement of efficiency and productivity, rather than to create new ones. The utilization of these ideas has not, however, been discussion in print. And some of the arguments we used and brought together are new regarding to the literature of hospital efficiency and productivity measurement. Utilizing the non-parametric data envelopment analysis (DEA) approaches, efficiency scores and productivity change indexes were estimated. Efforts were made to explain the difference of productivity performance among individual hospitals. Nevertheless, the methods we used and the economic approach behind them distinguish this study from other empirical studies of the medical market. Part I　 Market Uncertainty and Hospital Efficiency This part of the dissertation is focused on the measurement of efficiency of hospitals, incorporating uncertainty. There are stochastic variations in production relationships for hospitals. Generally speaking, the uncertainty of hospitals comes from two major sources: the natural uncertainty of medical cares; and the uncertainty of demands for medical cares (Arrow, 1963). Given the uncertainty in the medical market, the efficiency of hospitals hinges on how decision-makers deal with it. Undoubtedly, an optimal planning of the output buffers improves the efficiency performance. Using the hospital survey data in 1993 and 1994, and employing the chance constrained DEA model (Land, Lovell and Thore, 1993), the stochastic efficiency indexes of public and private medical centers and regional hospitals were estimated. Compared with deterministic frontier enveloping a given set of sample observations all the time, the chance-constrained frontier envelops them most of the time. That is, the chance constrained DEA allows the possibilities of output failure. Imposing different values of output failure probability, the estimation results were compared with the traditional (deterministic) DEA models. The empirical evidences of the chance constrained DEA model showed that, on average, private hospitals performed significantly better than public hospitals. This result matches with the result of the traditional DEA model. With Mann-Whitney U test, we compared the distributions of efficiency indexes under chance constrained DEA and deterministic DEA models. The test results showed that the difference between these two different models is statistically significant given a higher probability of output failure. These results imply that the nature of risk and the manipulation for risk are different for public and private hospitals. We also find that that the efficiency performance of public hospitals could be improved by the increasing of its reserve capacity. Part II　 National Health Insurance and Hospital Productivity Change In this part of the dissertation, we examine the impact of NHI on hospitals, and trace the sources of hospital productivity growth in Taiwan. To pursue our goal, we employ a data consisting of 157 medical centers, regional hospitals and district hospitals over the period 1993 to 1997, and resort to the Malmquist productivity index to measure total factor productivity change. The index could be decomposed into three components: technical change, pure technical efficiency change and scale efficiency change. The estimation technique used in the study is the deterministic non-parametric DEA approach. The results we find are revealing and suggestive to the public and the government in order to promote and assure the efficient delivery of quality health care. The average efficiency scores are 66.00% (74.87%) for CRS (VRS) technology and it means that there are substantial efficiency losses for the sample hospitals during the study period. The efficiency score of the hospitals as a whole in 1995 (the beginning year of NHI) was much lower than the other 4 years' efficiency scores. A censored Tobit regression analysis is used and identifies that NHI policy, ownership, rate of bed occupancy, average length of stay and the output-specific concentration level were all the significant determinants of technical efficiency. Empirical results indicate that most medical care regions became more output-specific concentrated. Total factor productivity on average deteriorated at an annual rate of -3.1%, and it was dominated by substantial technical regresses at an annual rate of -2.74%. The small hospitals were severely affected by NHI. Furthermore, within large and small hospital groups, the difference in technical change was statistically significant, but the differences in TFP and the associated components between ownership were not. Special attention was paid to compare　 et al.(1994), Ray and Desli (1997) and Grifell and Lovell (1998) approaches to decomposing the Malmquist productivity index. Empirical results indicate that the first 2 approaches yield accurate productivity changes, while GL doesn't. However, they produce almost the same magnitude of average TFP. In addition, no significant differences in the measured technical change and efficiency change were found among the three approaches.