The Research Of Stroke Risk Prediction And Economic Burden Simulation Based On National Stroke Screening Data

In recent years,China has witnessed an increase in the incidence of chronic diseases led by the economic changes,lifestyle shift and aging population.With the characterist ics of high incidence,high prevalence,high recurrence rate,high disability rate and high mortality,stroke has become the first cause of death for Chinese residents.However,experience in developed countries shows that stroke is preventable and controllab le.Reasonable management of the risk factors and early prevention of stroke can effective ly reduce its economic burden to families and the society.In order to controll the burden of stroke in China,the Ministry of Health of the Peoples' Republic of China initiated a national stroke screening and intervention program in May 2009.The program covers Chinese residents aged 40 years and older nationw ide.The program includes the preliminary screening and the rescreening.In the preliminar y screening,there are eight high risk factors,including hypertension,diabetes,atrial fibrillation,dyslipidemia,smoking,apparently overweight or obese,lack of exercise and positive family history of stroke.A person is considered “high-risk” if suffering from three or more than three of the risk factors or having history of stroke or transient ischemic attack(TIA)in preliminary screening.For those who have been classified to “high-r isk” group,further examination(such as imaging)is needed people with a precursor to stroke are treated accordingly.People suffering from two risk factors are considered as “intermediate risk”.People with two or more risk factors are given suggestions for stroke prevention,and their risk factors are intervened and guided through follow-up visits.Compared with the huge economic burden brought by stroke,the screening achieves remarkable results with much lower expenditure.More than 7 million stroke screening data have been accumulated nationwide so far.However,the risk level classifica t io n criteria currently used in preliminary screening and the stroke prevention and control strategy are determined by experts based on their experience,and the study on the stroke economic burden simulation is limited.Based on the stroke cross-sectional and cohort screening data,we improve the stroke risk level classification model used in the stroke screening program and constuct models to predict the stroke risk of people with different chronic diseases.We also build a model to simulate the stroke economic burden,and implement a simulation system based on it.The studies and utilization of the stroke screening data can further improve the intervention efficiency of stroke screening program and provide a strong scientific support for developing a timelier and more effective stroke prevention and control strategy.The main contents of this paper are as follows:1.The rationality of the stroke risk classification method proposed by stroke screening and intervention program is explored.Based on the cross-sectional screening data of 2012 and the cohort screening data between 2013 and 2017,the relationship of stroke and its risk factors is studied.On the whole,the probability and incidence of stroke with three risk factors are significantly higher than those with two risk factors,and the probability and incidence of stroke with two risk factors are significantly higher than those with only one risk factor.However,some probability and incidence of stroke with certain combination of two risk factors can be higher than those with some combinat io ns of three risk factors.The probability and incidence of stroke with a single risk factor can be higher than those with some combinations of two of even three risk factors in some cases.The program can take the ratio of stroke incidence with combinations of risk factors and the incidence without any of the risk factors as a variable threshold.The threshold can be adjusted to determine the stroke screening criteria and finally achieve the balance between economy and efficiency.2.In order to solve the problem that the stroke risk levels cannot be determined with risk factors including unknown values in preliminary screening,the stroke risk classification models are constructed.Based on the cross-sectional screening data of 2017,we develop logistic regression model,Na?ve Bayesian model,Bayesian network model,decision tree model,neural network model,random forest model,bagged decision tree model,voting model and boosting model with decision tree to classify stroke risk levels.We randomly select(with replacement)and construct 2,000 test sets from the experimental dataset,and calculate the averages and 95% confidence intervals of precision,recal,F1-score and AUC using them.Besides the whole test set,the data that cannot be classified using the classification method in screening program is also used to evaluate the model.The results show that the recall of the boosting model with decision tree is the highest(0.9992 and 0.9582,respectively),and the precision of the random forest model is highest(0.9733 and 0.5134,respectively).The national stroke screening program can choose the classification model according to the practical need in order to ensure the economics of program while avoiding unnecessary expenditures.3.The stroke risk of middle-aged patients with chronic diseases within 10 years is studied.Based on the cohort screening data between 2013 and 2017,the risk of stroke for people in 40-59 with the initial state of hypertension,diabete,and dyslipidemia within 10 years is studied using Markov model.The results show that the risk of stroke for people with hypertension is higher than that for people with diabete or dyslipidemia.Considering that the prevelance of hypertension in China is high,controlling hypertension is important for the prevention and control of stroke.For people with diabete or dyslipide m ia,interventions should be taken according to gender.The simulation results of stroke risk can provide decision-making basis for stroke prevention and control at the national level,and guide the key population of stroke prevention and control.4.The equilibrium model is used to simulate the stroke burden in China.The economic burden studied in this paper includes incidence,prevalence,mortalit y,hospitalization expenses,the disability adjusted life year and the GDP loss.Combined with the population prediction results by Leslie model,the equilibrium model is constructed to simulate economic burden with different degree of stroke incidence,prevalence and mortality.Based on this model,the stroke economic burden simula t io n system is developed,which can display the changes of economic burden in the form of graphs and charts.The simulation system provides a basis for establishing early warning map of stroke in China and developing early warning mechanism at the national level.In this paper,we study the stroke risk factors and stroke risk classification method,construct a model of the stroke risk of middle-aged patients with chronic diseases and an equilibrium model to simulate the stroke burden,and develop the stroke economic burden simulation system.The results can be used as strong scientific guides for the promotion of national stroke screening and intervention program in China,as well as for the development of timelier and more effective stroke prevention and control strategy.