The Study On The Cost Of Inpatients Of Hepatobiliary Tumor In Shanghai Based On Data Mining Method

Hepatobiliary malignant tumor is a kind of severe disease which has high morbidity and mortality, can cause excessive consumption of medical resources. On the other hand, the cost of different hepatobiliary tumor inpatients is distinct. There is not a reasonable and feasible cost consultation system for inpatients of hepatobiliary tumor, for current few studies about the cost of single-disease such as hepatobiliary malignant tumor, or simple studies based on small samples, so it is hard to describe the difference of medical resources comsumption among different kinds of hepatobiliary malignant tumor which is medicare advance payment and health economy specialists wanted, so the study on the cost of inpatients of hepatobiliary tumor is necessary.Data-Mining is a kind of new math-methods which have been arisen in recent years.There are many success applied examples in foreign countrys, but few applications in medicare and health-economic fields. The report about the data-mining studies on the cost of hepatobiliary tumor inpatients is lackness. We attempted the data-mining applications on this area, and the results were better than the ones by regular statistics methods, which proved the worthness for data-mining' s application in medicare. The selected cases based on ICD-9, which were collected from 8-years database about 3 great hospitals in Shanghai, included about 30000 hepatobiliary tumor. First, we consulted the methods of health economy and the key parameters in the National Bureau of Statistics of China(NBSC) to emend the correlativeinpatient cost. Second, we eliminated abnormal values, replaced missing values, and then weused data mining software to make some variables to maximize normality and to select the key variables for establishing math models. Last, we established math models by decision tree and neural network, all the data analysis worked on the computer by the data mining software-SAS Enterprise Miner. The neural netwok model is complicated than the decision tree model though it is more effective.We selected the decision tree model about 3-levels and 15-leaves finally, and SAS Enterprise Miner also provided the percentage and up and down limit of controlled-cost for the entire decision tree leaves respectively. The study will be helpful to objectively estimate medical outputs and establish the compensation criterion for the local governments or medicare organizations.The study will also induce normative medical behavior, reasonably using medical resource of medical personal institutin. Moreover, the cost model of hepatobiliary malignant tumor inpatients will be more helpful to set down a standard medical resource consumption of single-disease. \...