Analysis On Hospitalization Expense Structure And Influencing Factors Of Insured Inpatients With Primary Liver Cancer

Objective:The hospitalization expense and basic information of primary liver cancer inpatients with medical insurance in Tianjin from2008to2011were collected, to understand the trends of hospitalization expense. Gray Correlation Analysis and Degree of Structure Variation Analysis were used to analyze the internal structure of the hospitalization expense in four years, in order to know the relationship among expenses components and hospitalization expense. Firstly, the single factor analysis was used to screen the meaningful factors that affected the hospitalization expense. Then, to analyze the factors influencing hospitalization expense by the multivariable linear regression method and BP neural network model. What’s more, we compared the applicability of two models in the analysis of hospitalization expense. The study expects to offer a reference on related problem in medical research. We also put forward some corresponding suggestions for relevant government departments, in order to adjust the structure of hospitalization expense reasonably and control the growth of medical care expenditure as well as perfect the medical insurance system’ reformation.Methods:All information of6525primary liver cancer inpatients (ICD-10cord is C22.901) were collected from Tianjin urban employees’medical insurance system from January1st,2008to December31st,2011. Firstly, the basic situation in our data was described. Then, the Gray Correlation Analysis and Degree of Structure Variation (DSV) Analysis were been used to analyzs the internal structure of hospitalization expense by the Microsoft Excel2010softwere. And the factors influencing hospitalization expense were analyzed by multivariable linear regression model with SPSS20.0and BP neural network model with SAS9.13Enterprise Miner. Finally, we compared the difference between the two methods’results.Results:From2008to2011, the hospitalization expense of primary liver cancer inpatients in Tianjin has displayed the trend of rising. Drug expense, the main proportion in the hospitalization expense, accounted for47.36%of all. The followings were the expense of inspection (22.62%), treatment (12.58%) and medical materials (11.19%), the all above accounted for about93.75%. Gray Correlation Analysis found that the correlation of drug expense was the first, followed by inspection and treatment expense. Degree of Structure Variation Analysis showed that the contribution rate of DSV on drug expense is the first, followed by inspection and medical materials expense, and the cumulative contribution rate of all above accounted was about91.50%. The results of univariate factor analysis showed that ten variables of the age, personnel sorts, length of stay, history of hepatitis, cirrhosis or not, complication or not, surgery or not, transfusion or not, grade of hospital, with or without other diseases were correlated factors of hospitalization expenses (P<0.05). By multinomial linear regression analysis, the major influential factors as follows: length of stay (β’=0.501), surgery or not (β’=0.245) and grade of hospital (β’=0.220). The sensitivity analysis result of BP neural network model showed that length of stay (0.217), grade of hospital (0.140) and surgery or not (0.130) were the main influences. Multicollinearity and some strong influential points were found in multiple linear regression, so multiple linear regression model was not appropriate to anlalyze factors affecting hospitalization expense of primary liver cancer inpatient. Then, the model fit test found that the two models were different, the R2and R2adadjof BP neural network model were better than multiple linear regression model’s. The result showed that the expense of hospitalization is analyzed by BP neural network model appropriately.Conclusion:The hospitalization expense of the insured primary liver cancer inpatients in Tianjin from2008to2011were on the rise. The internal structure of hospitalization expense was unreasonable, and the proportion of drug expense was the highest which was the emphasis and difficulty of controlling hospitalization expense. The length of stay, the hospital level and surgery or not were the main influences on hospitalization expense of insured inpatients with primary liver cancer. Compared with the multiple linear regression method, BP neural network model was better. The BP neural metwork model was more suitable for the analysis of the factors affecting hospitalization expense data, and it is worth popularizing in the analysis of the factors affecting hospitalization expense.