Research And Implementation On The Prediction Model Of Medical Expenditure

With the development of universal health insurance in China,the growth rate of medical expenditure is accelerating.On the one hand,the demand for medical services and the acceleration of the aging process have forced the accelerated growth of medical expenditure.On the other hand,we face the dilemma of declining in economic growth,and the fiscal revenue is difficult to catch up with the growing speed of the medical expenditure.Therefore,the payment pressure of health insurance will be heavier in the future.So it is necessary to forecast the medical expenditure in order to improve the service efficiency of insurance funds.Now Diagnosis Related Groups(DRGs)have been widely recognized and effectively applied internationally.China is conducting a pilot application of DRGs,and has developed the Chinese version of DRGs based on the experience of foreign DRGs.The most representative one is Bei Jing Diagnosis Related Groups(BJ-DRGs).And the Classification and Regression Trees(CART)are also suitable for carring medical cost prediction because of its low requirements on capacity of training set,fast speed of classification and the high accuracy of prediction.We use BJ-DRGs and the algorithm of classification regression tree to predict and analyze medical expenditure to provide some reference for China to deal with aging and the health insurance payment reform.The main work of the article is as follows:Firstly,we construct a data set for medical cost forecasting.Since there are no open available data sets for medical cost prediction in China,we use the front page medical records of the two hospitals in Shenzhen as the data source in this paper.In constructing the data set of medical expenditure prediction,for making the data meets the requirements of data analysis and data mining,this paper applies the methods of data analysis and data mining to carry out necessary data preprocessing,including data cleaning,data integration,data reduction and data transformation.The paper tries to make sure the process is normative,the research results are in line with the scientific property and provide certain reference for the current medical reform.Secondly,based on the medical cost forecast data and set in-depth study of the classification regression tree and the BJ-DRGs grouping theory,we use the Python tool to construct the medical cost prediction model via the BJ-DRGs grouping principle and the classification regression tree algorithm and calculate the parameters of the model separately.Finally,using the K-fold cross validation to test the rationality and accuracy of the grouping under the two medical cost prediction models,and the differences between the two prediction models were presented in the test results,based on the classification regression model and the BJ-DRGs model.We compare the grouping and cost prediction effects of the two models from the grouping number,accuracy rate of classification,and forecasting evaluation index of medical expenditure.We finally get the conclusion that the prediction model of classification regression is better than the BJ-DRGs prediction model in the prediction of comprehensive grouping and medical expenditure.