Study On Diagnosis Related Groups Of Diabetic And The Forecast Of Its Hospitalization Expenses

Diabetes is one of the most common chronic non-infect diseases.With the rapid development of Chinese economy and the aging of the population,the incidence of diabetes in China is rising continuously.Excessive medical expenses for diabetes have caused a heavy economic burden on families and society.Therefore this paper conducts research on the diagnosis related groups(DRGs)of diabetes and their hospitalization expense.The research results have theoretical significance for the application of integrated learning algorithms in medical data,and also have practical significance for improving and promoting the DRGs payment system for diabetes in China.This study takes AR-DRG in Australia as a reference,and then use the homepage data of the patients diagnosed with diabetes in a major hospital in Beijing,a diabetes DRGs grouping model was constructed based on clinical similarity,based on the results of DRGs grouping,the hospitalization expense prediction model was constructed by using the ensemble learning method.The specific research is described as follows.Firstly,this paper uses data mining related knowledge to preprocess complex raw medical data,including data integration,data specification,data transformation,and data cleansing.The processed hospitalization expense data of diabetes patients were constructed into a database,which laid the foundation for subsequent DRGs grouping and hospitalization expenses prediction,and ensured the accuracy and scientificity of grouping and hospitalization expense prediction.Secondly,taking the hospitalized diabetic patients as the research object,the classification model of diabetes complications was proposed.The multivariate linear regression and other statistical methods were used to determine the factors influencing the hospitalization expense of diabetes,and the characteristic variables were obtained.Then the feature variable is used as the classification node to group the DRGs by using the CHAID decision tree model,and 10 DRG sub-groups are obtained,and the corresponding reimbursement standards and cost ranges are formulated.The grouping effect was evaluated by indicators such as variance reduction(RIV)and coefficient of variation(CV).The results showed that the cost homogeneity within the group was high and the difference between groups was large,and the grouping scheme obtained was reasonable.Finally,the results of the grouping of diabetic patients with DRGs obtained in the previous section are taken as new characteristic variables and raw data to form a new data set.The hospitalization expense prediction model was constructed by using the random forest and XGBoost integrated learning algorithm,and the parameters of the two models were optimized by grid search method.In order to improve the accuracy of hospitalization forecasting,this paper proposes a new Stacking model,which combines random forest and XGBoost as a base learner to further improve the prediction accuracy of the model.Finally,this paper conducts an empirical study on the diabetes model for the Stacking model,comparison with commonly used regression prediction models shows that: The Stacking model constructed in this paper has good prediction accuracy.