Multi-level Analysis Of Influencing Factors Of Type 2 Diabetes In Residents Of Bengbu City Based On Bayesian Estimation

Objective To explore the influencing factors of type 2 diabetes in residents of bengbu City by multi-level model based on Bayesian estimation.To provide reference for relevant departments to prevent and control type 2 diabetes,in order to improve the quality of life of patients.Methods The stratified random sampling method was used to extract the adult residents of Longzihu District of Bengbu City,and then the questionnaire survey was conducted on the selected objects.the demographic characteristics,objective health status,and health behavior factors of the research subjects were collected.SPSS 22.0 and MLwiN were used for statistical analysis of the collected data.Chi-square test was used for the univariate analysis of diabetes.Two-level zero model was used to analyze the clustering of the research subjects at the level of community health service stations.On the basis of univariate analysis,Multi-factor analysis was performed on the factors affecting residents' prevalence of type 2 diabetes using Bayesian multilevel logistic regression model.Compared with the Bayesian single-level logistic regression model including the same variables,the superiority of the Bayesian multi-level logistic regression model in analyzing the influencing factors of type 2 diabetes was verified.Results(1)Among the 3354 residents surveyed,1468 were male,1886 were female,and 357(10.6%)had diabetes.(2)single factor analysis showed that the factors associated with type 2 diabetes have gender,age,marital status,education level,hypertension,monthly income,family history of diabetes,mental health status,BMI,HbA1 c,smoking,drinking,vegetable intake,fruit intake frequency,exercise Above all of the prevalence of type 2 diabetes difference was statistically significant(P < 0.05).(3)The single-level logistic regression model under the Bayesian method has a good overall convergence when the MCMC chain length is equal to 30,000.The results suggest that gender,age,hypertension,marital status,family history of diabetes,mental health status,BMI,HbA1 c,alcohol consumption,vegetable intake,and fruit intake frequency may be the influencing factors for type 2 diabetes.(4)The two-level zero model based on Bayesian estimation indicates clustering at a high level,that is,clustering at the community service station level.Therefore,a multilevel logistic regression model based on the Bayesian method was established to analyze the factors affecting type 2 diabetes.And when the MCMC chain length is 30,000,the overall convergence of the model is acceptable.The results suggest that gender,age,marital status,hypertension,family history of diabetes,mental health status,BMI,HbA1 c,and frequency of fruit intake may be Influencing factors of type 2 diabetes.(5)From the perspective of deviation information criterion(DIC),the DIC value of the single-level logistic regression model based on Bayesian estimation is 1654.45,and the DIC value of the two-level logistic regression model based on Bayesian estimation is 1628.94,The difference between the two models is 21.51.The fitting effect of the two-level logistic regression model based on Bayesian estimation is better than the single-level logistic regression model based on Bayesian estimation.Conclusion The prevalence of type 2 diabetes among adult residents in Longzihu District of Bengbu City is clustered at the level of community service stations.The data has a hierarchical structure.Multilevel logistic based on Bayesian estimation can better handle multilevel structural data.To more accurately analyze the influencing factors of type 2 diabetes.At the same time,relevant departments should pay more attention to the people who are the male,who are over 60 years old,who have hypertension,who have a family history of diabetes,who have poor mental health,and who have a low frequency of fruit intake,especially those with abnormal HbA1 c.