Methodology Study on Prediction and Determinants of Total Health Expenditures in China 

Author  ZhuFengMei 
Tutor  YangTuBao 
School  Central South University 
Course  Epidemiology and Biostatistics, 
Keywords  Total health expenditure Cointegration Theory Error correction model Granger causality test Influencing factors Autoregressive moving average model Residuals from the regression model Statespace model Prediction accuracy 
CLC  R197.1 
Type  Master's thesis 
Year  2011 
Downloads  129 
Quotes  0 
Objective: To evaluate the scientific and reasonable method to identify the main factors affecting the growth of total health expenditure in China, to explore the longrun equilibrium and shortterm fluctuations in the relationship between total health expenditure and its influencing factors. Three forecasting methods to predict the future development trend of China's total health costs, and fitting a comprehensive evaluation of its predicted effect. Provide a reference for the healtheconomic policymaking, effective use of health resources, as well as the forecast of the total health costs. Methods: Based on the theory of cointegration, error correction model and Granger causality test, select the demandside factors (income, health expenditure, the aging of the population, the level of urbanization) and supplyside factors (number of physicians, the number of hospital beds) independent variables, the total health expenditure as the dependent variable to establish the influencing factors of total health expenditure in dynamic regression model to explore the longrun equilibrium with shortterm fluctuations in the relationship between the total health costs in China and its influencing factors, validation and causality between variables. Digital conversion and differential research sequence variance homogeneity of smoothing, the use of autoregressive moving average model, residuals from the regression model, and the statespace model of total health expenditure in 19782005 data model fitting, and the introduction of the mean square error, mean absolute percentage error and root mean square error three statistics 20062008, total health expenditure data for the model to predict the effect of test, finally, on this basis, further to China 20092020 total health expenditure trends to predict. Statistics and data to establish a database using Excel software to use SAS9.13 statistical analysis and forecast of statistical hypothesis testing using a twosided test inspection level α = 0.05. Results: (1) total health cost factors regression model analysis found that total health expenditures and income (GDP), health expenditure, population aging, longrun equilibrium relationship between the level of urbanization as well as the number of hospital beds; aging population is health The main factors of the total cost of shortterm fluctuations, followed by income and the number of hospital beds. Total health expenditure relative to the longterm income elasticity coefficient is 2.110, and the shortterm elasticity coefficient of 0.581. This shows 19782008, China's GDP per growth by one percentage point, the total health expenditure growth by 2.110 percentage points, the total health expenditure growing faster than the rate of economic growth; year GDP for every increase of one percentage point will lead this year's Health The growth in the total cost of 0.581 percent. Granger causality test results show: the total health expenditure is the reason for the increase of economic growth and the number of hospital beds, and the aging of the population and economic growth are the reasons for the growth of total health expenditure. Total health expenditure in the trend of the future development of the study found that the predictive value to 2020, real per capita health costs three models were 691.536 yuan, 720.130 yuan and 944.466 yuan, which the predictive value of the statespace model of real per capita health costs significantly higher than the other two model forecast results; residuals from the regression model of real per capita health costs predicted value is slightly higher than the autoregressive moving average model. ② three prediction methods comparison analysis revealed that the fitting precision of the statespace model is optimal, since the regression residuals model followed the ARIMA model fitting average absolute percentage error of the smallest in the three evaluation; in the model prediction accuracy, the state space model, and far superior to the ARIMA model the autoregressive error of prediction accuracy, which, since the sample regression error model to predict the mean square error is small, and the state space model sample forecasting mean absolute percentage error than small. Conclusion: ① income (GDP), health expenditure, the aging of the population, the level of urbanization as well as the number of hospital beds is a major factor in longterm changes in the total health expenditure, and the impact of population aging on the total cost of our health has gradually dominated. However, health expenditure, the level of urbanization has little effect on the shortterm fluctuations of the total health expenditure. ② total health expenditure is caused by economic growth as well as the reason for the change of the number of hospital beds. The entrance of the aging and economic growth is the growth in total health expenditure. (3) statespace model can be applied to time series prediction of the total health expenditure in China, capable of influencing factors included in the study, and to eliminate the impact of external shocks, policy changes and other unpredictable factors. Total health expenditure forecast study has important practical significance.