Longitudinal relationships between sleep-related symptoms and depressive symptoms in a rural-to-urban Chinese middle-aged and older population: a cross-lagged panel network analysis

There is evidence for a two-way relationship between sleep and depression; depression makes sleep disorders worse, which raises the chance of developing depression again. This study explored the directional relationship between sleep-related symptoms and depressive symptoms from novel perspective, and identified key symptoms using a cross-lagged panel network (CLPN) model. We followed middle-aged and older residents of three cities in Shandong, China from April 2021 to July 2024. Participants completed the social demography questionnaire, sleep-related symptoms questionnaire (selected from the Composite International Diagnostic Interview; CIDI3.0-screening section) and Patient Health Questionnaire-9 (PHQ-9) at two waves separated by 1-1.5 years. In total, 7054 participants were included. We used a CLPN model to estimate the auto-regression and cross-lagged effects of sleep-related symptoms and depressive symptoms after controlling for baseline age. The primary Gaussian LASSO retained 97 of 156 cross-lagged paths. In 1,000 participant-level bootstrap resamples, 18 paths had descriptive 95% percentile intervals excluding zero, including 10 cross-domain paths: seven from sleep-related symptoms to depressive symptoms and three from depressive symptoms to sleep-related symptoms. The largest cross-domain paths were difficulty falling asleep (Fas) at W1 to PHQ-9 sleep problems (Sle item 3: “Trouble falling or staying asleep, or sleeping too much”) at W2 ( B = 0.219, 95% interval 0.142–0.299), difficulty maintaining sleep (Slm) at W1 to PHQ-9 sleep problems at W2 ( B = 0.186, 0.090–0.299), and daytime lethargy (Dal) at W1 to low energy (Eng item 4: “Feeling tired or having little energy”) at W2 ( B = 0.142, 0.081–0.225). Exploratory node-level summaries identified difficulty falling asleep and difficulty maintaining sleep as having the largest outgoing predictive-effect values. Shared paths retained by the Gaussian and outcome-distribution-matched models had consistent directions, although network density was sensitive to the regularization rule. The findings suggest cross-lagged symptom-level associations in both directions. The three largest cross-domain paths were all from W1 to W2: difficulty falling asleep and difficulty maintaining sleep were associated with PHQ-9 sleep problems, while daytime lethargy was associated with low energy. These exploratory regularized associations should be interpreted cautiously and not as causal or within-person temporal effects. Not applicable.