The complex interaction between multidimensional health indicators and depression among Chinese older adults: A cross-sectional network analysis of health indicators and depressive symptoms.

Abstract

Objective: This cross-sectional study aimed to construct a network model linking multidimensional health indicators to depressive symptoms in Chinese older adults, using data from the 2020 wave of the China Health and Retirement Longitudinal Study (CHARLS; N = 9636, aged ≥60 years).

Methods: A 20-node network comprising ten CES-D-10 depressive symptom items, six IADL sub-items, and four single-item health indicators (self-rated health, self-rated memory, physical pain, and life satisfaction) was estimated using the EBICglasso algorithm (γ = 0.5). Five centrality metrics-expected influence, strength, closeness, betweenness, and bridge expected influence-were computed, and the stability of centrality estimates was assessed via correlation-stability (CS) coefficients. Sensitivity analyses included varying γ and residualizing each node for sociodemographic covariates and chronic-disease count; network invariance across sex, age group, and urban-rural residence was tested with the Network Comparison Test (NCT).

Results: Of 190 potential edges, 133 non-zero edges were retained, forming two internally cohesive sub-networks. Among health-indicator nodes, IADL-2 (expected influence = 1.033) and IADL-1 (1.032) were most central; among depressive-symptom nodes, CES-D3 (1.116) and CES-D4 (1.079) ranked highest. Life satisfaction (bridge expected influence = 0.429) and physical pain (0.333) were the predominant bridge nodes, followed by CES-D4 (0.290) and CES-D8 (0.289). The strongest associations were IADL-1-IADL-2 (0.559), CES-D5-CES-D8 (0.335), and self-rated health-pain (0.299). CS coefficients for expected influence and bridge expected influence both reached 0.75. Centrality and bridge-node rankings remained stable across γ values and after covariate residualization (Spearman ρ ≥ 0.99), and network structure was largely invariant across subgroups.

Conclusions: Life satisfaction and physical pain bridge the association between multidimensional health status and depressive symptoms, whereas CES-D3 and CES-D4 represent the most central depressive symptoms. These cross-sectional, hypothesis-generating findings require confirmation in longitudinal cohorts and intervention trials.