Research On The Growth Of Medical Expenditure And Its Influencing Factors Based On Empirical Mode Decomposition

BackgroundThe increasing of medical expenses is one of the hot issues of health research at home and abroad,Both developed and developing countries are very concerned about it.The increasing of medical expenses is special,it has many factors,including policy factors,economic development,price level,disease spectrum and so on.Because of the different intensity and time of different factors,the time series of medical expenses are often nonlinear and unstable.The traditional time series analysis method is not ideal in this kind of data analysis,and it can not systematically analyze the influence factors on the medical cost and time.the Empirical Mode Decomposition(EMD)is believed to be the best approach for the analysis of nonlinear and unsteady time series at present,The time series can be decomposed into different IMFs and RES,IMF is a stable time series with different intensity and different period.it represents the random influence factor,and RES represents its long-term trend.So,EMD is suitable for the reseach on the trend and influencing factors of medical expenses in theory.Now it was widely used in the field of engineering,marine meteorology and other fields,but there is no literature report on the research of the time series analysis in health field.Objectives(1)To explore the applicability and application conditions of EMD and its improvement in the study of medical expenses growth,to create the conditions for the popularization and application of EMD method in the time series analysis in health field.(2)To understand the model and the influencing factors of the growth of medical expenses,to provide reference for the control of medical expenses.Materials and methods1.Sources:The determination of indicators in this study was based on the previous studies and the requirements of sequence data of EMD,the main indexes which can reflect the growth of medical expenses are selected as the analysis indexes of this study.In the indicators of annual data(from 1978 to 2014),‘Out-of-pocket Health Payments Per Capita' was considered as the indicator of the growth of medical expenses,‘Real GDP Per Capita',‘Number of medical institutions Per million',‘Retail Price Index',‘Consumer Price Index',‘Social Health Expenditure Per Capita' and ‘Government Health Expenditure Per Capita' were acted as the influencing factors.While the indicators of monthly data(from 2006 to 2015)focused on the consumer price index,‘The consumer price index of health care and personal products 'was acted as the medical expense growth index,seven indicators including ‘Food Consumer Price Index ',‘Consumer Price Index for Alcoholic Drinks and Tobacco',‘Consumer Price Index for Clothing Category',‘Household Equipment and Maintenance Services Consumer Price Index',‘Consumer price index for transportation and communication',‘Consumer Education and Cultural Goods and Services Consumer Price Index',‘Residential Consumer Price Index' were considered as the influencing factors.All of these data are from the official website of the National Bureau of Statistics.2.Methods:(1)The descriptive statistical analysis was used to analyze the current situation of the growth of medical expenses in China.Main outcome measures included: development speed with linkrelative method,development speed with fixed basis,increment speed with linkrelative method,increment speed with fixed basis,average speed of increase,average speed of development.(2)The application of EMD was explored in the growth of medical expenses,and the key indicators of EMD was analysed: the decomposition of the original sequence using different iterations and white noise,using R software.(3)On the basis of RES and the original series trend,choose different models for the trend fitting,calculate the sum of squares of distances between the estimated values of RES and the original sequence,the smaller the square distance,the closer the RES to the original sequence.The optimal number of iterations and the white noise are determined by the sum of the square of the distance between the RES and the original sequence,the boundary effect and the mode mixing phenomenon(4)EMD analysis method was used to decompose all the research indexes,and the contribution rate of each component was calculated.The pseudo-component was deduced by time series correlation.(5)The Vector Auto Regression Model,Impulse Response Function and Variance Decomposition Method were used to analyze the same frequency components,to explore the main impact and contribution rate of the medical expenses growth.Results1.The current situation of medical expenditure in China(1)Out-of-pocket Health Payments Per Capita growthThe growth rate of year-on-year in 1979 was 12.99%,and the growth rate reached the peak in 1983(67.53%).During the period from 1978 to 1986,the change of growth rate was relatively larger.Medical expenses grew rapidly during 1987 to 1996,fluctuate between 23%-35%.During the 1997-2007 year,growth began to slow down,fell from 21.86% to 4.51%,it rose to 14.66% in 2008,a new peak of 19.48% in 2011,3 years as a cycle during the 2008-2013.2014-2015 period,the change of growth rate is small,for 4.73% and 5.65%.During 1979-2015,the average speed of development was 118.04%,with an average speed of increase 18.04%.(2)The proportion of Out-of-pocket Health Payments to the total health costsThe change of the proportion of Out-of-pocket Health Payments Per Capita to the total health costs in the past 37 years old was decomposed into five stages.In the first stage,during the period from 1979 to 1984,the proportion of Government Health Expenditure increased,that of Social Health Expenditure continued to decline,and that of personal cash expenditure was steadily increasing.Until 1984,the proportion of Government Health Expenditure,that of Out-of-pocket Health Payments Per Capita and that of Social Health Expenditure was 37%,33%,and 30%,which the proportion of social health spending reached the lowest point in the past 37 years.Then during the next period,between 1985 to 1992,the proportion of Government Health Expenditure kept decreasing while that of the Out-of-pocket Health Payments Per Capita and Social Health Expenditure remained increasing.In the third stage,the exploration of medical reform,from 1993 to 2000,the proportion of Government Health Expenditure continued to decline,and the proportion of Social Health Expenditure started to decrease gradually,which led to the sharp increase of the proportion of Out-of-pocket Health Payments Per Capita.In the fourth stage,from 2001 to 2011,the government increased investment in health spending and the social health spending kept a stable increase,thus the proportion of Out-of-pocket Health Payments Per Capita presented an obvious declining trend.In the final stage,from 2012 to 2015,the proportion of government health spending remained stable(nearly by 30%),the proportion of Social Health Expenditure increased constantly which has increased to 40% in 2015,thus that of personal health cash expenditure kept declining and decreased to 29% in 2015.2.The applicability and application of EMD in the research of the change trend of medical expenditure.EMD was used to analyze the growth rate of Out-of-pocket Health Payments Per Capita,and four IMFs and one RES could be decomposed by different iterations,and the decomposition results were no longer changed after more than seven iterations.The sum of the square of the distance between RES and the original sequence trend line is 9.08,when the iteration number is four.There is no obvious boundary effect and mode mixing phenomenon,so it is considered that the EMD decomposition result is ideal when the iteration number is four.EEMD method was used to analyze the growth rate of Out-of-pocket Health Payments Per Capita.When the number of iterations and the intensity of white noise signal were different,all of them can be decomposed into 4 IMF4 IMFs and 1 RES.It is found that when the number of iterations is 4,and the white noise signal intensity is 0.001,the sum of the squared distance is the smallest,which is 9.39.It is considered that when the number of iterations is 4,and the white noise signal intensity of the input is 0.001,the EEMD decomposition is the bestThe decomposition results of EEMD show that the lower the intensity of the white noise signal was better,which can be concluded that the result is consistent with the EMD when the white noise signal is 0.This result was according with the basic theory of EEMD,EEMD is equivalent to EMD when the intensity of the white noise signal was 0 and the number of repetitions was 1.3.The main influencing factors of medical expenses growth(1)Annual IndicatorsIn 10 years of the change of ‘Out-of-pocket Health Payments Per Capita',the contribution of itself maintained a high degree,and the first phase of the contribution rate was as high as 28.01%,while the contribution rate of ‘the per capita Social Health Expenditure' was 0.03%,the rest of contributions can be negligible.With the increase of time,the contribution rate of self-contribution was declining,and the contribution rate of the rest factors was increasing.In 1-10 years,The contribution rate of it self,CPI,PGDP,RPI,the number of hospitals per 10000 people,the per capita Social Health Expenditure and the per capita Government Health Expenditure was63.63-62.08%,0.02-0.473%,2.10-2.30%,1.57-1.71%,0.05-0.20%,5.14-5.67% and 0.01-0.30%.In the analysis of the annual index of this study,the fluctuation cycle influence index analysis results show the high frequency components for 3 years,CPI,national health expenditure,hospital number of million people per capita personal health expenditure has positive impact.The PGDP,per capita personal cash health expenditure was first during the period of their own the positive impact,then the rapid decay of RPI,SW have negative impacts,and then quickly return to around 0.The fluctuation cycle influence index analysis results of 5 years of the show,there are cyclical fluctuations.The fluctuation cycle for the long-term impact of 8 years,only the number of hospital million people has a negative impact effect.(2)Monthly IndicatorsIn the forecast values of 24-month,The Consumer Price Index of Health Care and Personal Products had a high self-contribution rate,and kept stable nearly by 52.92% in a long-term.Food and tobacco and alcohol consumption index of the medical consumer price index is increasing.12 the predicted results: Residential Category contribution was 0.33%-3.47%,Daily Commodity contribution was 6.55%-11.47%,Food category contribution was 0.91%-2.59%,Tobacco and Alcohol category contribute up to 5.16%-5.09%,the Traffic catefory contribution was 1.08%-3.22%,the Clothing catefory contribution was 0.96%-1.89%,the Educational catefory contribution was 0.73%-3.93%.For the high frequency fluctuation cycle of 3 months,the Alcohol and Tobacco catefory has positive impact,the rest is fluctuate around 0,no obvious rule.Effect of high frequency component results factors show that the impact strength of the Residential consumer price index for the medical consumer price index is increased.The low frequency component in research the price index of medical consumption,we found that the Residence category,Daily Commodity category,Food category,Transportation category,Education category have negative impact,but the Tobacco category and Clothing category have positive impact.Conclusions1.The importance of healthcare was recognized by the state and society,their investment in health maintained constantly increasing in recent years,which led to the decline of the proportion of Out-of-pocket Health Payments Per Capita to the total health cost.Out-of-pocket Health Payments Per Capita was still increasing greatly,but the growth rate has declined in recent years2.The EMD and EEMD were suitable for application in the study of the increasing trend of medical expenses,the non-stationary time series can be effectively decomposed into a number of stationary time series,each IMF hold a unique information that can be used for further study in medical costs,and can grasp the change trend,cycle fluctuations,changes in intensity as well as the influence of different wave on its own more accurately.and the square sum of the trend line / RES can be used as the criterion of the decomposition effect.3.The change of medical expenditure in our country is greatly influenced by national policy.For the changes of medical expenses,whether long or short term,the main impacts were from their own,30-40% of which could be caused by its own fluctuations The economic development got the greatest impact on medical expenses if removing their own impact.The Government Health Expenditure,inflation,the development of national economic,people 's living standards can increase personal medical expenses at some extent,while the growth of the number of hospitals can reduce the Out-of-pocket Health Payments Per Capita.