The Single Disease Payment Based On The GLM And Mixed Gamma Models

GLMs are extension of the classical linear model,it should no longer be limited to normal distribution,but extends to the exponential family.It is possible to ?t property variables and values in a speci?c range.And it’s more practical with the complex nonlinear relationship between variables,relative complexity.Rank sum test is a common nonparametric test method.It’s mainly used to test for di?erences between samples,the advantage is not to consider their distribution and widely used in medical,health and other?elds,it’s simple. As a ?exible and powerful tool for mathematical modeling,?nite mixture model has got a high attention,in the past few decades. Owing to ?exibility and systemic,it can be used to de?ne any complex probability distribution model in practical,so it has been widely applied in the analysis of bimodal or multimodal data.In the paper,we selected medical data in tertiary hospitals of Urumqi, medical reimbursements is independent variable and gender,age,surgery are the dependent variables.The rank sum test and the GLMs are used respectively to eliminate factors and test the di?erences among the classi?cations of various factors, then merge these having no signi?cant di?erences, and compare the results of the two methods;Use the eliminated and merged data to predict costs by the means of GLMs. In this paper,under individual data and grouped data,we estimate the parameters of ?nite gamma models via the EM algorithm. And the stochastic simulation results show that it is good. Then build single disease paymenr model with di?erent reimbursement in di?erent segments, ?t diseases fee distribution functions through mixed gamma model,?nally calculate the criterion.The advantages of a single disease payment of GLM is considerng the di?erences in patients, and then classi?ed, and ?nally to predict settlement costs;While single disease payment based on the mixed gamma distribution is directed against the distribution density function, and di?erent reimbursement in di?erent segments.