Calculation Of Indicator’s Weight Of Public Hospital Regulation Based On BP Neural Network

ObjectiveTo comprehend the situation and problems of the medol and index system of publichospital reguration in China by policy analysis and field survey; To select the suitable indexbased on multiple governance by literature review, expert-consatation and interview; Toexplain BP (short for the back proragation) nueral network’s application in the calculationof index’s weight; And to calculate the weight of public hospital reguration index with thedata from field survey; Finaly, to build up the index system and to provide support andreference for the reguration of public hospital in China.Methods①Informatic methods. Offical website of departments of health administration andhealth care were visited, and the data base including CNKI, WanFang Data and VIP Datawere searched to acquire the latest public hospital data, hospital supervision policy andrelated study, so as to analyze the charactistic and problems of the medol and index systemof public hospital reguration in China.②Social Research methods.Data of118public hospitals in Zhuhai, Shenzhen,Zhuzhou, Zunyi and Guangzhou was collected by field survey or mail-survey.30person inknow from hospital administrative departments were interviewed.③Statistic methods. The basic information and each index of sample hospitals weredescriptive analysed and the hypothesis testing was done with SPSS.④Operational research methods. Data evelope analysis(DEA) was done to calculatethe efficiency of sample hospitals with DPS software. And BP network model was built tocalculate the weifhts of index with SPSS software. Results①The the medol of public hospital reguration in China had gone through threedifferent stage, the problems currently is the confusion of sponsor and supervisor, and toomany depertments to regurate the public hospital to make it poor efficient. Meanwhile,there is no a comprehensive index system to regurate public hospital. We put forwards7index, including the percentage of medical income within the business income, non-debtratio and so on.②The allocation of resource and health service output in different level hospitals weresignificantly different, while the the data of above index of them were not. The relativeefficiency of each sample hospital was calculated with DEA.③The theory and implementation of BP nueral network model were introduced tocalculate the different level hospital’ regulation indes’s weight respectively, including thepercentage of medical income among operation income (0.177，0.037)、none-indebt ration（0.100，0.220）、the percentage of nurse among nurse and doctor(0.004，0.258)、thepercentage of salary and welfare spending among total spending(0.099，0.084)、rate ofhospital beds utilization(0.521，0.037)、return on assets(0.047，0.342)、the percentageof operation income amongoperation income and spending(0.052，0.022) thus establishingthe index system. Empirical test showed that the index system was of high reliability andvalidity.Discussions①The study use BP neural network to calculate the weight of indicators, which canminimizing the influence of human aspect, and make it accorded with rearity, increase itsreliability and validity.②The study used new method to train the network, which greatly reduce the requireof the number of training samples.③The study transform the indicators perspectively according their real meanings,andmake them qualified for the BP neural network. ④The network in the study was single-hidden layer backforward neural network, it iseasy to calculate and simple for managers to use.Suggestion①The current public hospital reguration model was hard to regurate public hospitalefficiently. Public hospital management system should be reformed, and the committee forpublic hospital supervision should be buidt up. Meanwhile, the information system betweenhospital and government should be improve, so as to provide quick and real timeinformation for supervision.②It was feasible and reliable to use BP nueral network model to calculate the index’sweight. However, attentions should be paid to many details such as data pre-processing,variable selection and network designing. Keep in mind that the scale and quality oftraining sample have big impact on the result.Innovations and shortageCompared to other index system, this one has its innovation points. It is based onmultiple governance, the index was relative number and the weight were calculated moreobjectively.Due to the limit of time and money, the number of sample hospital is still not satisfied,and indicators were still needed to be perfected.