The Study On The Case-Mix Model Of The Cost Of Hospital Stay For Stroke Inpatients

Since the foreign countries got some effect in controlling health cost by adopting Prospective Payment System based on Diagnosis Related Groups, feasibility studies in our country about how to control health cost according to disease types shows that it is necessary to build our DRG-PPS based on the situation of our country, according to disease type as the main factor and other factors related disease types, but as an effective instrument for output-price, whether the case mix can provide a satisfactory prediction of the variance of costs in stroke inpatients , No research has been found about case- mix in providing the quantitive prediction to inpatients costs in China so far. For this reason, we adopt the decision tree to form the case- mix in stroke inpatients and built two kinds of case-mixmodels about costs of stroke inpatients to evaluate whether the case-mix can provide a satisfactory prediction of the variance of costs in stroke inpatients. This work aims at providing statistic references to control costs in stroke inpatients and to exert DRGs payment pay in medical reform. The results are as follows:1. We applied decision tree model and found that nodes of importance ranked ahead respectively are surgery the degree of care and infection inter-hospital by which the case-mix is suitable with our stroke inpatients and fit medical institutions and the third payers.2. The results of multivariate regression analysis show that using case-mix formed in this research, the variance explained in the costs amounts to 17% for inpatients. The variance explained in total costs can substantially and significantly increases when add other factors to case mix, especially days of stay. This result do suggest us to discuss the feasibility of treating days of stay as grouping nodes in forming case-mix of stroke inpatients, which awaits further studies.3. In this work we also pick out better case-mix recursive model which alse show that the effect of case-mix to costs isless than that of LOS, and which also reveal the internal relations of DRG groups and all other variables for costs in stroke inpatients. The regression model is more suitable for prediction and the recursive model fit to explain costs of inpatients.