Multilevel Model Of Repeated Measurement Data For Cognitive Impairment And Depression Co-existing Among The Elderly In Taiyuan

Objective:The aims of this study were:(1) to describe the prevalence of cognitive impairment and depression for the elderly in Taiyuan based on the epidemic survey data;(2) to apply the multilevel model of repeated measurement data to the progress of cognitive impairment and depression to find out related factors, to make an exploratory study in the relationship between cognitive impairment and depression and to provide scientific bases for formulating policies to improve their quality of life (3) to discuss the strength of multilevel model for longitudinal data analysis and to provide methodological reference for exploring influencing factors and variation patterns for the progress of other chronic diseases co-existing.Methods:Our data came from three waves of repeated measurement data of517older adults(≥60years old) in Taiyuan. Multiple stepwise regression analysis was used to identify significant factors related to MoCA and GDS scores. Multilevel model of repeated measurement data was applied to find out the trends and the trend differences of cognitive impairment and depression among the elderly.Results:Multiple linear regression model and multilevel model both fitted well. For the cognition study, multiple linear regression analysis indicated that baseline depression, education, age, reading books or newspapers, smoking and marriage were the significant factors. The older adults who had depression at the first survey, lower education level, more aged, less reading, smokers, singles had lower cognition level; Multilevel model analysis showed that individual background covariates such as age, marriage, education, living, hearing, reading books or newspapers, smoking were the predictive factors for the cognition. After controlling these individual background covariates, we found that there was no statistical significance for the main effect of depression, but interaction of depression and time was still statistically significant (t=-1.96, P=0.049), and decline rate of MoCA scores for depression group was greater than non-depression group. For the depression study, multiple linear regression analysis indicated that baseline cognition, reading books or newspapers, living, diabetes, use of aluminum cooking utensils, heart disease, occupation were the predictive factors, the older adults who had cognitive impairment at the first survey, less reading, living alone, diabetes, high-frequency use of aluminum cooking utensils, heart disease, manual workers had higher level of depression. The results of multilevel model analysis showed that individual background covariates such as occupation, source of income, living, reading books or newspapers, use of pots and pans made of aluminum, diabetes, heart disease were the predictive factors for the depression. After controlling these individual background covariates, we found that both main effect of cognition(t=2.09, P=0.038) and interaction of cognition and time(t=2.56, P=0.011) were statistically significant, the average baseline GDS scores of cognitive impairment group were higher than non-cognitive impairment group, and accumulate rate of GDS scores for cognitive impairment group was greater than non-cognitive impairment group.Conclusion:Various factors are related to cognition and depression in older adults. The depression could promote the decline of cognition, and the cognitive impairment could worsen the depression in older adults. More attention should be payed to the prevention of depression in cognitive impairment and cognitive impairment in depression in older adults to avoid cognitive impairment and depression co-existing. Compared with the traditional method, multilevel model can provide more overall analysis about the influencing factors and the trends in chronic diseases co-existing.