【作者】 赵颖；

【导师】 何钦成；

【作者基本信息】 中国医科大学 ， 社会医学与卫生事业管理， 2007， 硕士

【摘要】 前言本文通过对一所大学附属综合性三级甲等医院15个临床科室连续5年的卫生资源投入、产出及配置情况等方面的研究，筛选出针对临床科室相对效率评价中具有重要性、代表性、可得性等特征的关键效率指标，分别应用了TOPSIS、因子分析、主成分分析以及DEA(Data Envelopment Analysis—数据包络分析)方法进行效率(效益)评价，并对评价结果相关性比较分析，试图在评价指标体系的构建、指标的筛选及DEA方法和其他方法间的比较研究方面做进一步的探索，同时针对如何建立综合医院临床科室考核的奖励机制，提供更为科学、客观、公平、合理的评价依据。资料与方法本研究的对象是某综合性三级甲等医院的15个临床科室，评价中使用的这些科室2001-2005年间人力资源、物力资源、医疗服务质量、科室规模、经费收支及药品收入比例等方面的数据，主要来源于该医院的财务、设备以及人力数据报表及医院效率指标调查数据表。本文的主要研究方法为文献法和多种统计学方法。文献法：本文查阅了大量中、英文文献了解本课题在国内外进展状况并主要采用文献法遴选出医院及临床科室相对效率评价的指标。统计学方法：本文运用聚类分析、相关分析、变异系数等方法分析，筛选用于DEA的评价指标。同时，应用DEA(Data Envelopment Analysis)、TOPSIS法、因子分析法、主成分分析法对科室的效率(效益)进行评价。结果本研究中按照敏感性、重要性、代表性、可获得性等的原则构建了适用于TOPSIS法、因子分析法和主成分分析法的由20项指标组成的评价体系；运用聚类分析、相关分析、变异系数分析方法，对20项指标进行筛选，确定用于DEA评价的5个输入指标和7个输出指标；运用以上四种方法分别对目标科室连续5年的数据进行效率(效益)评价；对评价结果进行相关性分析。具体结果如下：DEA评价结果排序：TOPSIS方法评价结果排序：因子分析方法结果排序为：主成分分析法结果排序为：DEA、TOPSIS法、因子分析法和主成分分析法的相关性分析结果为：*指在双侧0.05的水平上显著相关**指在双侧0.01水平上的显著相关通过DEA的分析结果可以看出同一年度不同科室间的相对效率情况。以2005年为例，如下表所示讨论在以往大多数的研究中，医院中科室相对效率评价指标的选择大都采用主观性方法，很少经过统计学的分析和筛选，难免会影响到评价结果的可信性、真实性和科学性。临床科室相对效率(效益)评价中所采用的评价指标通常会涵盖与科室的规模、质量、效益和效率有关的诸方面，每个评价指标又具有相同或不同的特征，在选择上也因评价目的而异。因此，评价者必须根据评价目的和评价指标的特征确定评价指标体系和相应的评价方法，以从方法学和技术上保证评价结果的科学性、客观性和公正性。本次研究应用了相关系数、聚类分析和变异系数方法并结合评价指标体系的目的性、完整性和评价方法的要求筛选出可适用于TOPSIS、因子分析、主成分分析法或DEA方法的评价指标，这在以往的研究中未见报道。研究中构建了适用于TOPSIS法、因子分析法和主成分分析法的由20项指标组成的评价体系和适用于DEA方法的由12项指标(5项投入指标，7项产出指标)组成的评价体系，这两个评价体系中既有人力、物质、资金的投入指标，也有医疗服务的数量和质量、科室管理运营的效益和效率的产出指标；既有各方面的总量指标，也有考虑科室间规模差异的结构性指标；既有代表经济、经营效益和体现科室发展潜力的指标，也有代表社会效益的指标。特别是将人力成本比总支出同趋势化值、设备投入比总支出和药品收入占业务收入百分比同趋势化值等与医院经营效益或社会效益(指向“看病难、看病贵”这一社会热点问题)有关的指标列入综合评价体系在以往的研究中尚未见报道。本次对75个被评价单元的相对效率(效益)评价属于以终结性评价为主的诊断性评价，考察了DEA方法、TOPSIS、因子分析法和主成分分析方法的应用价值和可实施性。DEA方法是一种非统计性和非参数方法，不需要对数据的随机特征如分布等作出假设，在处理多输入、特别是多输出的综合评价方面的能力具有绝对优势。该方法把每个被评价单位作为一个决策单元(DMU)，通过对各DMU投入和产出比率的综合分析。构造目标函数，确定各DMU的前沿，并根据各DMU与有效生产前沿的距离状况。确定各DMU是否DEA有效及低效率的来源。由于DEA方法避免用主观标准确定效率，以最优化过程确定指标权重，故客观性较强，管理者由此可获得许多有用的管理信息，并以此为依据进行调整，指导实际工作；TOPSIS法作为系统工程中有限方案多目标决策分析的一种常用方法，对数据分布类型、样本含量、指标相关程度和多少无严格的限制，在计算过程中不减少指标个数并能消除各指标不同量纲带来的影响，可根据评价目标对各指标等权视之，其排序结果能定量反映不同被评价对象的优劣程度、直观、可靠；因子分析法是一种将多变量(指标)化简的技术，特别适用于多指标间共线性的综合评价系统。因子分析法通过分解原始变量，从中归纳出潜在的“类别”即“共同因子”，达到对原多个指标降维，消除指标间相关性的影响，简化评价计算过程的目的；主成分分析方法是将原来多个具有一定相关性的指标合成一组相互无关的综合指标来代替原指标。消除原指标的相关性，降低指标维数的方法。在因子分析法和主成份分析法中，各综合因子的权重是根据综合因子的贡献率的大小确定的。这就克服了某些评价方法中人为确定权数的缺陷，使得综合评价结果唯一，而且客观合理。从实现诊断性评价的目标和四种方法的评价模式上考虑，TOPSIS法、主因子分析法和主成分分析法能比较出各被评价单元在同一评价指标Z化值或同一因子得分数量上的差异，却无法获得基于相同投入水平下，特别具有多个投入指标情况下的不同被评价单位各个产出指标相对于效率(效益)最优单位应提高的幅度(增量)，而DEA方法可以给出相对无效单元欲成为相对有效单元在某些投入指标上应减少的量或在某些产出指标上应增加的量，较之TOPSIS法、因子分析法和主成分分析更具有实际的应用价值。本研究所涉及的评价对象、评价指标和原始数据的状况均适合或基本适合这四种方法应用的前提条件，故四种方法的评价结果之间具有非常显著或显著的相关性，以此亦可以互相印证评价结果的科学性、客观性和公正性，同时也能够相互弥补评估方法各自本身的局限。TOPSIS法、因子分析法、主成分分析法和DEA的秩相关分析表明四种评价方法的评价结果趋势基本相同，但DEA的结果与前三种方法结果之间的相关系数远小于前三种方法结果之间的相关系数，存在明显的差异。例如，在TOPSIS、因子分析和主成分分析评价结果中05康复，05皮肤科等科室始终位于年度后五位，而在DEA分析结果中它们却处于相对效率有效单元；05脑外、04肛二等科室在TOPSIS、因子分析和主成分分析评价结果中虽然始终位于年度前五位，但它们在DEA分析结果中却处于相对效率无效单元。四种方法的评价结果间出现上述差异的原因有以下几个方面：1、TOPSIS、因子分析和主成分分析方法突出了规模因素的影响，8个因子或主成分中的前4个与“规模”相关者共占有68.1％和75.6％的权重；2、TOPSIS法、因子分析法、主成分分析法的评价取向属于“多投入-多产出”模式，而对于从技术角度来确定被评价单元的相对效率就略显不足，相反DEA方法的模式是“适当的投入-最大的产出”，着重于对被评价单元相对效率的评价；3、在计算四种方法评价结果间的秩相关系数时，DEA评价结果共37个有效单元(接近被评价单元总数的50％)得分均是1，排序相同，掩盖了有效单元之间的细微区别，可能也是导致DEA与另三种方法评价结果间秩相关系数较小的原因。四种方法中，TOPSIS法、因子分析法和主成分分析法的评价结果主要侧重于从科室效益和因投入较大可能带来的发展潜力的角度反映科室的综合实力。DEA的评价结果能够客观地反映出科室现时的运行效率，同时给出各评价指标实际值与理想值之间的差距(松弛值)，DEA能为管理者指出某些被评价单元效率低下的原因所在，并提示应调整的具体量，较前三种方法更具有现实的指导意义。综合TOPSIS法、因子分析法、主成分分析法和DEA的评价结果，还能够映出科室的发展趋势。以神经内科为例：在TOPSIS法、因子分析法和主成分分析法的评价结果中，神经内科2001-2002年位于年度前二、三、五位，到2003-2005年居于前一、二、三位，表明神经内科的综合效益是逐渐上升的；在DEA评价结果中，神经内科2001-2002年均，位于年度所有无效单元中的倒数第一名，2003年其排名变为倒数第四名，2004年、2005年成为有效单元，最终成为符合“适当的投入-最大的产出”的有效单元。虽然应用的方法不同，但各自的评价结果都表明，无论在综合效益还是相对效率方面神经内科的发展都呈明显上升的趋势。临床科室的建设和发展往往需要一定的周期，很可能在增加投入的初期得不到相匹配的产出。但是，如果科室的发展方向正确，其理想的产出效果通过一段时间的运行最终可以获得。反之，如果经过了必要的周期，目标科室始终处于无效单元，则应及时分析其原因是否源于社会需求萎缩、技术发展定位不当抑或科室带头人能力不足等等，找出症结有针对性的进行调整。本研究提示应用DEA方法结合TOPSIS方法、因子分析法或主成分分析方法，公平、合理地设置评价指标，可以为兼顾科室经济效益、社会效益、运行效率和发展潜力的评价机制的建立做出贡献。四种方法的评价结果结合应用，更能科学、客观地反映临床科室的综合效益、运行效率和总体发展趋势，为加强和改进临床科室的建设乃至建立更加合理的奖励体制提供较为客观、全面、科学的评价依据结论1、构建了与以往研究不同的，涵盖了与医院临床科室规模、质量、效益和效率相关的多指标评价体系；2、应用多种统计学方法可使评价指标的筛选结果更加客观、科学、合理；3、DEA方法较TOPSIS法、因子分析法和主成分分析法更适用于以诊断性评价为主的相对效率(效益)的综合评价，并为管理者指出了某些被评价单元效率低下的原因所在，较后三种方法更具有现实的指导意义；4、DEA方法结合TOPSIS法、因子分析法和主成分分析法的应用更能科学、客观地反映临床科室的综合效益、运行效率和总体发展趋势，为加强和改进临床科室的建设乃至更加合理的奖励体制的建立提供较为公平合理、科学客观的评价依据更多还原

【Abstract】 The purpose of this study is to analyze the health resource inputs, outputs, andthe disposition of the 15 clinical departments（ which from the affiliated hospital ofDalian university ）during last 5 years, and to choose the key efficiency indexes thatpossess the important, representative, attainable characteristics and can be used toevaluate the relative efficiency of the clinic department.The efficiency and benefit ofthe clinical departments were evaluated by the methods of TOPSIS, Factor Analysis,Principal Component Analysis and DEA （Data Envelopment Analysis） respectively,furthermore, the correlations of the results obtained with the different methods wereanalyzed. Construction of evaluating indexes systerm,selection of evaluatingindexes,comparing DEA with other methods and establishing rewards regulation onthe basis of evaluating efficiency and benefit of the clinical departments were probeinto in order to provide more scientific,objective,fair and reasonable evidences forevaluation.Data and MethodsThe objects of this study are 15 clinic departments of a general hospital. Thedate related the human resource, material resource, medical service quality,department scale, service income and expenditure and the ratio of drug income tototal income during 2001 to 2005, were obtained mainly from the annual statistics review of finance,facility and human resource and also from the questionnaire ofgeneral hospitals efficiency and benifit.The main methods for this study are literature method and statistical methods.Iterature method: A lot of research paper available in English and Chinese werereferred to understand the progress in this field. The efficiency and benefitsevaluating indexes of clinical departments were selected by the utilization ofthese literatures.Statistics methods: The method of DEA （Data Envelopment Analysis）, TOPSIS,Factor Analysis, Principal Components Analysis were used to evaluate the efficiencyand benefits of clinical departments. Meanwhile, the cluster analysis, correlationanalysis and coefficient of variation were applied to screen the DEA evaluationindiexes.ResultsThe evaluating system suitable for TOPSIS,Factor Analysis and PrincipalComponents Analysis methods, which was made up with 20 indexes wereconstructed, according to the principal of the sensitive, important, representative,attainable characteristics.Using cluster analysis, correlation analysis and coefficientof variation methods, 8 input and 6 output indexes for the DEA evaluation werechosen from the 20 indexes.The efficiency and benefits of target clinicaldepartments in five consecutive years were evaluated by the methods of TOPSIS,Factor Analysis, Principal Component Analysis and DEA （Data EnvelopmentAnalysis）; Correlation of the results evaluated by four methods were analyzed,Themain results are showed: The results ranking of DEA evaluation:Tablel, The annual DEA evaluating result in different clinical departments（descending） The results ranking of TOPSIS evaluation:The results ranking of Factor analysis: The various sections Principal Component Evaluation Results RankingCorrelation Analysis of TOPSIS, factor analysis, principal component analysis andDEA evaluating results.From the DEA evaluating results, the efficiency of different clinical departments atthe same year can be observed as follow:. DiscussionIn the former evaluative studies on relative efficiency of clinical departmentsless statistics methods were used to analyze and screen indexes, but the subjectivemethods were used in more studies which will affected the credibility, authenticityof the evaluated results. The evaluating indexes used in the relative efficiencyevaluation of Clinical departments usually include some aspects related with thescale, quality, effectiveness and efficiency of clinical department ect. Someevaluating indexes usually possess same or different characteristics, the choice forindexes also are varied according to the purpose of the evaluation. Therefore,evaluators must identify the corresponding evaluating indexes system and methodsin accordance with different purpose to guarantee the objectivity and impartiality ofevaluation methodologically and technically. With the requirements of evaluatingpurpose, indexes systerm integrity and evaluating methods, the coefficient correlation, cluster analysis, coefficient of variation were applied at first time toscreen the suitable evaluating indexes for TOPSIS, Factor Analysis and PrincipalComponent Analysis or DEA methods.A evaluating systerm composed of 20 indexes for TOPSIS, Factor Analytic andPrincipal Components Analytic methods and a evaluating systerm composed of 12indexes（5 input indicators, 7 outputs indicators） for DEA were established. Twoevaluation systems include the input indexes of human resources, material and fundand the output indexes of the quantity and quality of the medical service, the benefitand efficiency of clinic department administration. They not only have themeasurable amount indexes but also the constructive indexes reflecting the scaledifference among the departments; not only the indexes related economic benefit,management efficiency and department’s potential ability to develop, and also theindexes of social effects. Especially, the ratio of human resources cost to totalexpenditure, the ratio of equipment expenditure to total expenditure and the ratio ofdrugs income to total income, which are related to the management efficiency andsocial effects directing to the social problem such as "too difficult to see doctors, andtoo expensive to see doctors", haven’t been reported in the former studies yet.In this study, concering the relative efficiency and benefit among 75 units basedon the results of diagnostic end evaluation, the applicable value and possibility inpractice of DEA were compared with that of TOPSIS, Factor Analysis andPrincipal Components Analysis methods.Data envelopment analysis （DEA） is an approach of non-parametric and non-statistics, it doesn’t require any assumption of data random characteristic likedistribution and has great advantage to handle multiple input and multiple outputmodels. Each unit is evaluated as a decision-making unit （DMU）, through inputs andoutputs ratio analysis to structure objective function and ascertain the front edge ofeach DMU, and according to the distance of each DMU with production frontcondition to determine whether the DMU has effective or not and find the low efficiency origin. Because DEA can judge the efficiency and definite indexes weightsby the process of optimization, DEA can provid more objectivily managementinformation wich can be used as evidences for managers to adjust and guidepractical work. TOPSIS （-Technique for Order Preference by Similarity to IdealSolution） is a method commonly used in limited plan multi- goals decision-makinganalysis. It doesn’t need strict limit to data distribution type, sample content,correlativity or numbers of indexes, and can eliminate the influence of indexeshaving different dimension and give each indexes same weight. Its ranking result canreflect the superior and inferior degree of the evaluated units in direct-viewing,reliable way. Factor Analysis is a statistical data reduction technique used to explainvariability among observed random variables in terms of fewer unobserved randomvariables called factors. It reduces a large number of variables to a smaller numberof factors for modeling purposes, where the large number of variables precludesmodeling all the measures individually. As such, factor analysis is integrated instructural equation modeling （SEM）, helping confirm the latent variables modeledby SEM, finally attain the goal of eliminating influence within relevant indexes andsimplify appraise process. Principal Components Analysis （PCA） is a technique usedto reduce multidimensional data sets to lower dimensions for analysis. It is designedto capture the variance in a dataset in terms of principle components. In effect, it istrying to reduce the dimensionality of the data to summarise the most important （i. e.defining） parts whilst simultaneously filtering out noise. In the Factor Analysis andthe Principal Components Analysis methods, each synthesis factor’s weight isdecided according to the contribution rate. So it can overcome the limitation ofman-made, make the evaluation results exclusive, objective and reasonable.Considered the model of four methods and the target of diagnostic evaluation,TOPSIS, Factor Analysis, Principal Components Analysis methods can reveal thedifference between the score of some indexes and factors, but difficult to know howto adjust the indexes value to improve the relative efficiency base on the same investment. The results of DEA can reflect the indexes ideal value matched withinput and output and the relative efficiency value of each evaluated unit, at sametime, it can point out the amount shoud be augmented or reduceded in the inefficientunits yet. So the results can help the inefficient units improve relative efficiency baseon the scientific evidence. In some significance, the DEA method is more reliable,clear and useful than TOPSIS and Factor Analysis method on the aspect ofinformation that they can supply.The evaluation targets, indexes and original data involved in this study are allor mostly suitable for the prerequisite of these four methods application, therefore,the evaluation results has extremely remarkable or remarkable relevance. Based onthis, it also may mutually verify the evaluation results of the scientific, objective andfair, simultaneously, it can mutually make up their limitation.The rank of correlation analysis for TOPSIS, Factor Analysis, PrincipalComponents Analysis and DEA showed that the tendency of these four evaluationresults is same. But the correlation coefficient between DEA result and other threeresults is far less than the correlation coefficient between the former three results,and this difference is significant. For example, the results of TOPSIS, Factoranalysis and Principal Components Analysis indicate that the departments like 05D₆, 05 D₉ and so on are located yearly latter five all the time, meanwhile, theyactually belong to the effective units in the DEA analysis. Although the departmentslike 05 D₁₂ 04D₈ and so on in TOPSIS, Factor Analysis and Principal ComponentsAnalysis appraisal results are throughout located yearly first five, but they actuallyare effective units in the DEA analysis. The reasons for the above differencesbetween these evaluation results of four methods are in the following areas: 1、TheTOPSIS, Factor Analysis and Principal Components Analysis methods allhighlighted the influence of scale factor. Among eight factors or components thefirst four together hold 68.1% or 75.6% weights with "the scale". 2、The evaluationtropism of TOPSIS, Factor Analysis and Principal Components Analysis methods belong to "more input-more output", which is unsuitable for the evaluation of therelative efficiency from technical aspect. However, the DEA method emphasizes theevaluation of the relative efficiency for evaluated units, whose mode is "properinput-maximal output".3、In calculating the rank correlation coefficient among theresults of these four methods, there are 37 effective units with score 1 in the DEAevaluation results （close to 50% of the total）, and the arrangement is same. So, DEAcan’t show the tiny difference from the efficient units, which lead to fewercorrelation coefficient between DEA result and other three results than that ofbetween the former three results the coefficient correlation of DEA far less than thatof other 3 methods.Among these four methods, TOPSIS, Factor Analysis and PrincipalComponents Analysis results focused primarily on department’s comprehensivestrength, development potential and the greater efficiency by increase investment.DEA method can objectively reflect operating efficiency in the hospital clinicaldepartments, product the disparity between actual value and the ideal value （relaxesvalue） and Simultaneously point out the reasons and adjusting amount for theinefficient units. For hospital managers, the DEA method has more feasible andrealistic guiding significance than other three methods have.The tendency of clinical departments development can be reflected byintegrating with the evaluation results of TOPSIS, Factor Analysis, PrincipalComponent Analysis and DEA （Data Envelopment Analysis） methods. Take D₅（neurology department）as example: in the evaluationl results of TOPSIS, FactorAnalysis and Principal Component Analysis methods, the D₅ is respectively locatedthe yearly first, three and five in 2001-2002 and occupied preceding two, three in2003-2005 year, which indicated its comprehensive benefit has been rising gradually.In the DEA evaluation results, D₅ is located in reciprocal first of all ineffectiveunits during 2001-2002, its place ranked the fourth count down in 2003. In2004-2005, it eventually become an effective unit which meets the requirement of "proper inputs-the maximal output". Although used different methods, theirevaluation results indicate whether the comprehensive benefit or relative efficiencyof D₅ showed a marked increase in its development trend.The construction and development of clinical department often needs thecertain cycle, the matching output cannot be obtained during the initial period ofincreasing investment. But, if the development direction is correct, its ideal deliverswill be allowed to obtain through a certain running period. Whereas, if passedthrough the essential cycle and the target units still remain ineffective, administratorsshould timely analyze the reasons for the shrinking demand, improper direction oftechnology development or insufficient capacity of leaders and other aspects. Somanagers can identify the crux and carry on the adjustment as soon as possible.This study indicates that combining DEA, TOPSIS, Factor Analysis andPrincipal Component Analysis methods can set fair and reasonable evaluationindexes and make a contribution to establish the evaluation mechanism for balancethe economic benefit, social effects, operating efficiency and development potentialof the clinical departments. Applications combining these evaluation resultsof fourmethods can be more scientific and objective to reflect the efficiency and benefit,operational efficiency and overall development trends in clinical departments, andalso can provide more scientific, objective, fair, reasonable basis for establishinghospital clinical departments evaluation and incentives system to strengthen andimprove the clinical departments construction.Conclusions1、It makes up the multi-indexes system including the scale, quality, benefit andefficiency of the hospital clinic department, which is different from the old ones.2、It applies multi-statistical methods to choose the evaluation indexes, whichare more objective, scientific and reasonable.3、Compared with TOPSIS、Factor Analysis and Principal Component Analysismethods, DEA method is more suitable for the diagnostic syntfesis evaluating of the relative efficiency in the hospital clinic departments, and can point out the reasons ofinefficient units. For hospital managers, the DEA method has more feasible andrealistic guiding significance than other three methods have.4、Applications of the DEA method integrating with TOPSIS, Factor Analysisand Principal Component Analysis methods can reflect the efficiency and benefit,operational efficiency and overall development trends of clinical departments inmore scientific and objective way, and also can provide more scientific, objective,fair, reasonable basis for establishing hospital clinical departments evaluation andincentives system in order to strengthen and improve the clinical departmentsconstruction.更多还原