Research On Medical Cost Pricing From The Perspective Of DRGs

DRGs-PPS is an internationally recognized advanced medical payment method.It has become an important means to curb the unreasonable increase of inpatient expenses and achieve a win-win situation for decision-making,hospitals and patients.What’s more,compared with many developed economies such as the US,the research on DRGs because of its late start,is not mature.For the newly released CHS-DRG grouping plan,it is also necessary to further refine the grouping according to local actual conditions,and develop appropriate medical insurance payment standards.In addition,under the influence of the two-child policy,facing the peak of fertility,facing a series of medical risks such as advanced childbearing and postpartum hemorrhage,hospital delivery will inevitably consume more medical resources.The reasonableness of birth expenses and hospital charges will inevitably affect the level of maternal health care.In this paper,guided by the " China Healthcare Security Diagnosis Related Groups,CHS-DRG",using the homepage data of delivery woman in a certain area,we first use the relevant knowledge of data mining to preprocess the complex original medical data.Then,on the basis of clinical similarity,we use the CHAID decision tree to build the CHS-DRG grouping model.And finally,a medical insurance payment pricing model is constructed based on the DRGs grouping results.Among them,when constructing the medical insurance payment pricing model,the median method is used as the reference value for the enrolled cases.And for the unenrolled cases,a hospitalization expense prediction model was constructed by using RF,XGBoost,and LGBM integrated learning algorithm,and the parameters of the models were optimized by Bayesian optimization method.The CHS-DRGs grouping model constructed in this paper is rather effective,with small intragroup variation,large inter-group variation,and moderate number of groups.The pricing model constructed based on the grouping results gives the DRG groups the cost standard value and the cost upper limit for the enrolled cases,and the results are consistent with the actual situation.For the cost prediction of the unenrolled cases,the stacking model is better than base learners,and its prediction results are relatively stable.The R square of the stacking model is 0.88,RMSE is 0.23,and MAPE is 1.84%.In summary,the research results in this article can provide a reference for payment standards for medical insurance and help promote the improvement and development of DRGs-PPS.