Health Risk Study In The Ageing Society

Population ageing arising from longer life expectancy and declining fertility rate is a global phenomenon,especially in China,where the population of the baby boomer steps into ageing.The population of middle old and oldest elderly in China is expected to grow rapidly over the next few decades.Population ageing makes the non-communicable diseases,the main type of which are cardiovascular dis-eases,cancers,and diabetes,the leading cause of mortality and the major burden of diseases,since chronic diseases is highly related to the ageing group.Disability is the usual consequence of the functional decline sug-gested by the chronic disease and the complication for the elderlies.The disabled population in China reached40.6 million at the end of 2015,and it is expected to reach 98 million in year 2050.The chronic disease and disability show to be the predominant health risk in ageing China.As the growing incidence of chronic disease and disability is an universal thesis today,attention on the prevention guided amounts of research in recent decades,and some factors are verified correlated to the diabetes risk and thus can work as predictive factors.For example,interventions on lifestyle have been approved effective in reducing the diabetes risk.Finds the predictors of diabetes is important as it is the basis of risk prevention.The traditional findings focus more on the clinical data,which is a collective sample of those who have already been suffering diabetes.But some areas such as insurance product pricing need evidences from the normal population,which include both those who are with the disease and those who are not,to give a more general view of the incidence of diabetes risk and the key predictors.Besides,the policy design of the public health system also relies on evidences from larger sample with greater disparity.Due to the availability of the individual-level longitudinal data of the elderlies'health status,study can be done to deeply assess strong factors contributing to healthy-diabetes transition and consequently predict the diabetes risk and its changes.This paper follows the risk analysis framework to investigate the chronic disease and disability characteris-tics to illustrate and evaluate the health risk in the ageing society in China.The main contents of the study are as follows,1.The identification of the determinants of chronic disease and disability.The paper pre-selects more than 100 indexes from 10 categories including genetic factors,lifestyle factors,social-economic status factors,communication or isolation factors,stressful events factors,availability of med-ical resources factors,nutrition factors and disease factors based on literatures as the candidate predictors,and incorporate them into the modelling of the incidence rates of chronic disease and disability using the machine-learning algorithm Extreme Gradient Boosting(XGBoost).We use the L₁norm based regularization technique and cross-validation to avoid the data-driving and over-fitting problem that are commonly seen in the machine-learning models.2.The calculation of the incidence rates of chronic disease and functional disability.For the purpose of facilitating the health insurance product pricing and effectively locating public health resources,finding health risk predictors using a more generalized population sample set rather than clinical data is in imperative need.This paper designs two kinds of models to calculate the transition probability from disease/disability-free to the incident of two major chronic disease,diabetes and heart disease,and the functional disability.One is to use a tree-based machine learning technique Extreme Gradient Boosting(XGBoost),which is both efficiency and compatible for incorporating amounts of variables into the modelling.The other is to incorporate the latent factor into the modelling of the health status transition of the elderly in China.It include the age,gender and time as the basic variables,and use the latent factor to represent the stochastic dynamics in the transition rates.The introduction of the latent factor balances the model's accuracy with the availability of data,drawing important features out while mimicking the real process of health status transitions.3.We further study the cross impacts of incidence of disability,recovery and mortality on a representative's life path to provide insights on some important questions.For example,how does the disability,death risk and the corresponding uncertainty evolves in one's life?Are people living longer active lives as well as longer lives?Does the ratio of active(disability-free)life expectancy change as time passes?How long will the disabled be expected to live?4.We also provide a cross-country comparison of the health risk characteristics between China and the US.5.Considering the historical segregation of urban and rural in China,it interests us whether the residence factor makes a difference in the incidence of disability,or more generally,in health status transitions.We extend our model to include an additional residence factor to provide a discussion of its impact.We use of U.S.Health and Retirement Study(HRS)and Chinese Longitudinal Healthy Longevity Survey(CLHLS)over the period of 1998 through 2014 for the model fit.The empirical results provide evidence of a disparity of the chronic disease risk in populations with different social-economic conditions.The economic conditions,life-style,social isolation,stressful life events and the access to the medical service all attribute to the prediction of the chronic disease and disability risk,but relative importance of social-economic conditions such as economic conditions,life-style and the access to the medical service rises in the investigated period.The dominant predictor in social-economic conditions,the years of schooling,is heavily associated with more severe disability.There is large urban-rural disparity concerning in incidence of disability.Urban residents have higher dis-ability risk and lower mortality risk than their rural peers.Among the factors that are usually applied for the transition rates modelling,it is interesting to find that age and time are still of great power in predicting the incidence of disability,while gender's impact largely shrinks when other predictor categories enter the model.The empirical results also show an improvement of the disability rates in China,but not in the US.The improvement of mortality applied to both the healthy and disabled group in China,while it limited to the healthy group in the US.But recovery rates deteriorates in both countries.Although both countries experience an ex-panding life expectancy for the healthy group,we find the contributors are different.It is contributed by both longer active lives and more years spent in disabled in China,while by longer active life expectancy which can compensate the shrinking years with disability in the US.It suggests Chinese people experienced an expansion of years with disability,while the US enjoy a compression of the disabled living years in the investigated period.In terms of the uncertainty,it plays a vital role in the incidence of disability in China.Consequently,there are wider confidence intervals in the life expectancy and the ratio of disability-free life expectancy in China,suggesting a larger disparity of the disability risk among the Chinese elderlies.The present study fills the literature gap in three aspects.First,the paper extends previous studies on China to a more comprehensive framework which takes uncer-tainty into accounts.It provides a universal model to capture the dynamics of the health status transitions in the Chinese elderly,which makes it simple to apply in practice.Second,the paper applies the big-data analysis into the identification of the health risk determinants for the first time,providing the statistically selected significant predictors through a thorough examination of all acknowledgeable factors.It draws out the contributory predictors of functional disability from a great variety of indexes that can portray our observations.It expands the knowledge of the functional disability precipitators and consequently facilitates the modelling and prediction of the disability risk.It improves the previous study by overcoming the lagging of the calculation and the partial consideration in the predictor identification.We adopt the machine learning algorithm Extreme Gradient Boosting(XGBoost)to handle the high-dimensional data.A Bayesian-based algorithm is applied to search for the optimal hyper-parameters for the XGBoost.Cross-validation is used both in the hyper-parameter training and the out-of-sample prediction.Reg-ularization is used to avoid over-fitting.The predictive model reaches a classification accuracy of above 85%,which is around 30%higher than the regularly used Lasso GLM model.Third,the recovery process is considered in the health states transitions of the elderlies in China,which calibrates the underestimation of life expectancy in previous study on China.Last but not the least,we provide a cross comparison of the health risks between China and US,including the factor impact in the health status transition rates,the total and active life expectancies and the distribution of the disability along age.It can provide the insight about the current development stage of the health risk in the ageing society in China.