Study on depressive symptom trajectories based on heterogeneous statistical learning

Likang Li

Abstract:

Depression, a critical and widespread mood disorder, poses a major
challenge to international health. It significantly disrupts personal,
professional, and academic pursuits beyond ordinary sadness, affecting vast numbers globally and emerging as a primary driver of
disability, per the World Health Organization. Its association with
serious physical conditions such as heart disease, diabetes, and
heightened suicide risk further exacerbates its impact on personal
well-being, familial stability, and economic health, underscoring
its extensive burden on society. This study aims to explore the relationship between educational levels and depressive symptoms
among the elderly in China, particularly through the application
of heterogeneous mediation effect analysis to identify potential
mediating variables between the two. Utilizing the longitudinal
dataset from the China Health and Retirement Longitudinal Study
(CHARLS), the study employs structural equation modeling (SEM)
to screen potential mediating variables and explores the heterogeneity of mediation effects among different groups through grouping
methods. The results indicate that subjects can be divided into four
groups, each with significantly different levels and trends of depressive symptoms. Further analysis reveals variations in the paths of
mediation effects between different groups, highlighting the complexity of the relationship between education level and depressive
symptoms. The significance of this study lies in its first application of heterogeneous mediation effect analysis to the CHARLS
dataset, offering a new perspective on understanding how education impacts the mental health of the elderly through various
mediating paths. We recommend that future research further investigate the confounding factors in the mediating relationships
between education, economic status, and depressive symptoms to
more comprehensively understand the interactions among these
variables.