Study On Identification Risk Factors Of Mild Cognitive Impairment And Construction Of BP Network Prediction Model

Research Background:Mild cognitive impairment(MCI)is the stage as the "optimal intervention window stage" for dementia prevention and treatment,and its early identification is helpful for early implementation of cognitive function intervention.Therefore,in order to better prevent the occurrence and development of MCI,it has become a new topic and direction in the field of dementia prevention and treatment to prevent the occurrence and development of MCI by identifying MCI risk factors and establishing MCI occurrence risk prediction model.Research Purpose:1.Identify the influencing factors of cognitive dysfunction in the elderly MCI population;2.Construct a neural network prediction model of risk prediction of elderly with MCI;3.Based on the risk prediction model of the elderly with MCI,some cognitive management suggestions will be put forward to the predictors.4.According to the results of predictive model analysis,it provides a theoretical basis for the formulation of health management related administrative department policy and community to carry out cognitive management for the elderly.Research Methods:1.Through literature review and cross-sectional study,the related literature is analyzed,and the attribution indexes of cognitive dysfunction in MCI patients are summarized.2.From December 2017 to October 2019,1409 elderly people in Huzhou and Chifeng communities were selected by the method of multi-stage sampling combined with convenient sampling.The cognitive function and related factors of elderly people in the community were investigated by general data and risk factors questionnaire,Montreal Cognitive Assessment,Activity of Daily Living Scale and Geriatric Depression Scale.T test,chi-square test,univariate analysis of variance and other statistical methods were used to determine the influencing factors of cognitive dysfunction in the elderly MCI population in the community and the final inclusion variables of the neural network model.According to the modeling principle of the neural network model,the neural network prediction model of the risk of MCI in the elderly in the community was constructed.Research Results:1.A total of 1409 elderly people were included in this study,including 1186 MCI screened.The subjects were 449 males(31.90%)and 960 females(68.10%);the subjects were 60~89 years of age,with a total of 1153(82%)and 1348(96%)in lower secondary and lower education;and 1170(83%)in occupations dominated by manual labour;2.A single-factor analysis showed that factors that were statistically significant and correlated with the incidence of MCI in the elderly were: location,sex,age,education,occupational nature,residence status,monthly income,smoking status,alcohol consumption,family history of dementia,social activities,cognitive activities,comorbidities,metabolic disorders,number of cognitive activities,number of physical exercises,number of social activities.3.Construction of BP neural network model: extraction training group 1127 people,prediction group 282 people,each risk factor distribution in the two groups is more uniform,the two groups are basically homogeneous,comparability is good.The training data are divided into training data and verification data according to the ratio of 7:3.BP neural network is constructed with neuralnet packets in the R.The number of input neurons in the input layer is the same as the number of input 17 variables,the number of hidden layer units is 7,and the output layer variables are two,MCI and non-MCI,respectively.The results of constructing BP neural network showed that the top five ranking of the importance of risk factors for MCI occurrence were age 76~80(100.00%),rural residence(100.00%),no social activities(100.00%),physical labor(100.00%),and no drinking(94.71%).He prediction accuracy of the model was 93.44%,and the sensitivity and specificity were 0.96 and 0.93,respectively.The predicted baseline data were brought into the BP neural network model to predict the risk of disease in the population.The results showed that the overall accuracy of the model prediction was 0.89 and the F1 value was 0.60.Research Conclusion:1.By analyzing the factors influencing the risk of MCI occurrence of elderly people in the community,the variables included in this study were: place of residence,sex,age,education,occupational nature,residence,monthly income,smoking,drinking,family history of dementia,social activities,cognitive activities,comorbidities,metabolic disorders,number of cognitive activities,number of physical exercises,number of social activities.the BP neural network model shows that the top five variables in importance are: age 76~80(100.00%),rural residence(100.00%),no social activities(100.00%),physical labor(100.00%),and no alcohol(94.71%).The above five variables have important reference value for predicting the risk of MCI occurrence.2.This study will use BP neural network to predict the risk of MCI occurrence,which has good predictive ability and discriminant ability.Compared with the traditional prediction model,the BP neural network considers the nonlinear and collinearity problems.The prediction accuracy of the training model is 93.44,and the accuracy of the prediction model is 89.00.It has good fitting ability and prediction ability.