Research On The Application Of Data Mining Technology In Disease Related Groups

Objective: investigate the feasibility and reliability to apply data mining technology in "diseases related groups" practice. And through the analysis of the result, we put forward valuable suggestions for medical insurance cost control in our country.Method: based on numerous documentary reading, our issue uses Coronary Heart Disease as an example to do research on inhospital data from an internal comprehensive hospital. Firstly, through data discretion, we form two different charge classes. Secondly, we mine association rules between patient's condition,prognosis,difficulty of treatment,necessity of treatment and charge class, and use these association rules to filter mainly related attributes. And then, we set these attributes as classified nodes to build decision trees by decision tree algorithms. At last, we get diseases related groups and make variance analysis to evaluate the grouping effects.Result: through data mining technology, we get two plan of diseases related groups, the first one uses "oper", "dis_diag_status" and "admiss_status" as classified nodes, and forms four different groups; the second one uses "oper", "coronarography" and "PCTA" as classified nodes, and forms four different groups too. The two plans all passed variance analysis hypothesis testing. It proves applying data mining technology to build "diseases related groups" is rational.Conclusion: data mining technology can efficiently build rational "diseases related groups". This method can adapt the complexity and variety of disease treatment, and objective,justified define health care benefit unit. It is useful for formulating medical insurance compensation, so as to make up for the deficiency of other cost control methods. Because of the limitation of data acquisition, it should be taken on a larger scale to make the diseases related groups more rational.