Analysis Method On Intervention Index And Effect Of High Risk For Type2Diabetes

With continuous improvement of living standard, the prevalence of type2diabetes isincreasing in the world. And it becomes one of the most common chronicnon-communicable diseases. Type2diabetes is difficult to cure and has long duration. Itcauses kinds of diabetes complications, which might bring heavy economic burden to thestate and individals seriously affecting the quality of life of patients. Prevention is theprimary strategy for disease control. Once the prevention and control of type2diabetes isachieved early, the disease will be avoided effevtively, the diabetes prevalence will bereduced greatly, the life quality of risk population will be improved, and the huge economicexpenses will be decreased. In this context, the aim of this research is to build theintervention guidance system of type2diabetes high risk, to provide the quantifyintervention guidance of high risk person, and to prevent or delay disease. The technicalguidance of quantify intervention at high risk is deeply conducted. The main line of thedissertation is around the intervention guidance system building of high risk person, withdiscussing the conditions, contents, key problems and applicative values.In the dissertation, the data mining is adopted to research the high risk intervention. Atthe beginning, the quantify analysis of the influence of multiple factors on blood isconducted. And the intervention points are chosen. The intervention effect of high riskfactors is simulated and predicted. Then, the intervention guidance system of type2diabetes was built. These researches could provide basis for type2diabetes prevention. Atthe same time, the effectiveness of intervention points is verified.The main contributions of this dissertation are listed below:(1) A quantitative analysis method of the influence of risk factors on blood glucose level isproposed in this dissertation.To provide theory basis for understanding the characteristics of blood glucose changeand confirming the intervention points of type2diabetes, the quantitative analysis methodof the influence of risk factors on blood glucose level is proposed. The quantitative methodis proposed to analyze the influence of risk factors on blood glucose using back propagation (BP) neural network. Ten risk factors are screened first. Then the cohort is divided into ninegroups by gender and age. According to the minimum error principle, nine BP neuralnetwork models are trained respectively. The quantitative values of the influence ofdifferent risk factors on the blood glucose change can be obtained by sensitivity calculation.The experiment results indicate that weight is the leading cause of blood glucose change(0.2449). The second factors are cholesterol, age and triglyceride. The total ratio of thesefour factors reaches to77%of the nine screened risk factors. And the sensitivity sequencescan provide judgment method for individual intervention.This method can be applied to risk factors quantitative analysis of other diseases andpotentially used for clinical practitioners to identify high risk populations for type2diabetes as well as other diseases.(2) A method of the intervention point selection and validation is proposed in thisdissertation.In order to make the intervention point targeted, a method of the intervention pointselection and validation is proposed. The model of intervention point selection andvalidation is trained using qualitative analysis and quantitative computation combinedmethod. When the model is constructed, machine learning and statistical method areadopted. The subgroups attribution model is built using H-SVMs algorithm. In differentsubgroups, intervention points are adjusted quantitatively. And in order to vertify theeffectiveness of intervention points, the proportion of state transition which transferredfrom high risk state to moderate risk state, low risk state and normal state is calculated. Thesimulative computation results indicate that intervention points are confirmed in16subgroups at high risk. When we decrease BMI5percent (weight loss:3-5kg) in persons athigh risk class A state, the proportion of state transition from high risk to a lower state(moderate risk state, low risk state or normal state) could reach15-25percent. In high riskclass B1state, changing BMI5percent (weight loss:3-5kg) is advised. In high risk classB2state, reviewing GLU and2-hour postprandial glucose is advised. In high risk class Cstate, CHOL is adjusted2.5percent (0.13-0.34mmol/L), the proportion of state transition is10-25percent.Selecting and vertifying the intervention point at high risk could provide theoretical basis and guidance suggestion for targeting intervention at high risk persons, and couldimprove intervention effect.(3) A simulation and prediction analysis method of intervention effect on high risk factorsof type2diabetes is proposed in this dissertation.In order to predict the intervention effect at the beginning of the intervention period,the simulation research work is conducted. According to the current physical condition oftype2diabetes high risk person, the intervention effect is simulated. And the simulationand prediction analysis method of intervention effect on high risk factors of type2diabetesis proposed. The simulation and prediction model of high risk is trained using machinelearning and statistical methods. The model of state transition probability is constructedusing C4.5algorithm, to compute the state transition probability with the top two or threerisk factors adjusted. Then the glucose value is predicted using BP model. At last, the statetransition probability is computed, and the intervention effect is given. The experimentresults indict that the high risk is divided into four groups according to the RSD riskdetermination, sensitivity analysis of intervention points and general characteristics of thepopulation. And the intervention effect is computed affer adjusting the intervention points.Comparing the state transition probability between the four groups and class A in the wholegroup can verify the different feature in different subgroups which is divided by gender andage. Different intervention point has different effect degree on the state transition. Nomatter gender or age, loss weight can be chosen as the preferred intervention point.20%reduction in weight could bring35-40%state transition probability. The more interventionpoints are adjusted, the better intervention effect is caused. The best intervention effect canreach95%. The sensitive degree is different when intervention points are adjusted in thefour groups. The plateau of state trantion in men>50is around10%. The plateau inwomen>50is around20%. The plateau in men≤50is around20%. The plateau in women≤50is around14%. When reduce the same probablitity of intervention points, theintervention efficiency before the plateau is higher than after the plateau. In the samegender and age group, the different sensitive population is distinguished into class A, classB1, class B2and class C, which is benefit to the prediction and intervention work.The prediction analysis method can provide technology support for understand risk degree and intervention efficiency. The research can facilitate the resources usageeffectively and improve intervention efficiency.(4) The intervention and guidance sysem of type2diabetes is designed and implemented inthe dissertation.In order to improve the efficiency of intervention implementation, the results oftheoretical research and requirements of actual application are combined. The interventionand guidance system of type2diabetes is designed and implemented. The system canscreen the high risk, attribute the subgroups, confirm the intervention points and computethe intervention degree and state transitioin probability. Then by combining the personalexercise and dietary habits, individual intervention guidance program is formulated. At thesame time, the stability, the reliability and the fault tolerance of system are verified throughthe experiment.