Study On The Prediction Model For Influence Factors Of Mild Cognitive Impairment

Objective To collect information about demographic information,history of past diseases,life factors and other information,according to whether there is cognitive dysfunction,it is divided into mild cognitive impairment(MCI)patients and subjects with normal cognitive function,analyze the related factors affecting MCI,establish the predictive model of the influencing factors of MCI,and predict the influencing factors of the progress of MCI.Methods From October 2015 to October 2017,512 patients who were hospitalized in the health care department of the Affiliated Hospital of Qiingdao University and the neurology department were selected for the information questionnaire survey and data collection of demographic information,history of past diseases,life factors and so on.According to the diagnostic criteria of MCI,512 subjects were divided into MCI group(282 cases)and normal cognitive function group(230 cases).The decision tree method and Logistic regression analysis were used to analyze the influence factors of MCI,and the decision tree prediction model and Logistic regression model were constructed,and the comparison was made by comparison of the subjects' working characteristic curve(ROC curve).The area(AUC)under the curve of the two groups was compared,and the predictive ability of the two models was compared.The patients in group MCI were followed up to collect the information of the previous disease control and the change of life factors,and the patients were followed up on the discharge day of the group MCI,and the deterioration of the function was recognized as the end point of the observation.The follow-up was completed in December 31,2017.Cox proportional hazards regression analysis was used to analyze the influencing factors of MCI development,and Cox regression prediction model was established.Results 1.The accuracy of the cross validation model of the decision tree model is 76.80%,and the fitting degree of the model is better.The factors affecting the decision tree model include social interaction,years of education,diabetes,interest,smoking,tea drinking and so on.It is a classified node variable of the decision tree model,which indicates that these factors are the factors affecting the MCI.The social interaction is located in the first layer of the decision tree,which is the influence factor of the root node.It shows that the factor has the greatest impact on MCI;the years of education and tea drinking habits,smoking and interest,diabetes and interest are respectively located in the second,third,fourth layers of the tree model,which are the secondary factors of MCI.The results of Logistic regression analysis showed that the factors involved in the Logistic regression model were 4 factors,such as residence,long-term residence,diabetes,social communication,and so on.2.The decision tree model predicted that the AUC of the elderly MCI was 0.765(95%CI 0.723 ~ 0.807),and the Logistic regression model predicted that AUC of MCI was 0.722(95%CI 0.677 ~ 0.767).3.During the follow-up period,102 subjects had cognitive deterioration,and the Cox proportional risk regression analysis found that the independent risk factors of MCI progression were reduced,the protective factors were lower(60-69 years old),the blood pressure was well controlled,and the blood lipid control was good.The Cox regression model of MCI progression was h(t,X)=h0(t)exp(-1.114 X1-1.985 X2-0.594 X3+0.924 X4).(X1: age;X2: hypertension control;X3: Hyperlipidemia control;X4: interest).Conclusion 1.Diabetes and smoking are the risk factors for MCI in the elderly;frequent participation in social interaction,higher education,tea drinking and interest are protective factors for the elderly MCI.2.In the prediction of MCI,the decision tree model is superior to the Logistic regression model.3.Decreased interest is a risk factor for MCI progression.Low age(60-69 years old),good blood pressure control and good lipid control are protective factors for MCI progression.Early intervention of risk factors that can be controlled is an important way to prevent MCI progression.