Factors contributing to depression in the elderly: a longitudinal data analysis based on linear mixed effect

Abstract

Objective To investigate the prevalence of depression in the elderly and identify the contributing factors by fitting the Geriatric Depression Scale (GDS) scores using a linear mixed effect (LME) model. Methods The data were obtained from a longitudinal study by the United States National Alzheimer's Coordinating Center (NACC). Starting from 2011, this study investigated the prevalence of depression using Short GDS initially among 1 345 elderly subjects, from whom 230 were followed up continuously for 4 years. According to the minimization principle of the values of Akaike information criterion (AIC) and Bayesian Information Criteria (BIC), a variance-covariance structure was chosen to fit the data and estimate the model, and the MIXED module in SAS software was used to model and analyze the longitudinal data. Results Analysis of the LME model showed that a longer education time was associated with a lower GDS score (β=-0.103, P=0.016). The elderly who divorced (β=0.742, P=0.025) and those who lived alone (β=1.495, P=0.024) were more likely to have a higher GDS score than those in marriage. The elderly who required assistance for complex activities had higher GDS scores than those who were completely independent (β=0.420, P=0.036). The elderly reporting frequent depressive feelings in the past 2 years had higher GDS scores (β=1.176, P < 0.000 1). The elderly patients with dementia (β=1.068, P=0.003) and those with mild cognitive impairment (β=1.020, P=0.001) had higher GDS scores than the elderly with normal cognitive function. Conclusion The LME model allows efficient analysis of the longitudinal data from this elderly cohort. A lower level of education, a lowered self-care ability, cognitive impairment and divorce are all important factors contributing to depression in the elderly

Keywords

• linear mixed effect model
• longitudinal data
• elderly depression