The longitudinal relationships among loneliness, source-specific perceived social support, and depressive symptoms in Chinese children and adolescents: a three-wave cross-lagged panel network analysis

Loneliness, perceived social support, and depressive symptoms are closely linked in childhood and adolescence; however, less is known about how depressive symptoms, loneliness, and source-specific deficits in perceived support are prospectively associated at different levels of measurement. Prior studies using aggregate scores have limited insights into how specific symptoms interact with distinct support sources over time. This three-wave cross-lagged panel network (CLPN) study examined these relationships in a large school-based sample of ~5000 Chinese children and adolescents. Depressive symptoms were assessed using the Patient Health Questionnaire (PHQ-9), loneliness was measured with a single item, and perceived social support was evaluated across four sources: parents, teachers, classmates, and close friends. We estimated both cross-sectional and cross-lagged panel networks to examine centrality indices, bridge expected influence, and longitudinal predictive pathways. Network structures remained stable across waves. Fatigue consistently emerged as a central depressive symptom, and low classmate support was central within the support domain. Loneliness showed the strongest bridge effect, linking depressive symptoms with low perceived social support. Longitudinally, fatigue and loneliness showed relatively high outward expected influence and prospectively predicted later increases in low perceived support, although these temporal pathways should not be interpreted as causal effects. The associations between suicidal ideation and support were similar for parent and teacher sources but differed for peer support during a specific time period, demonstrating source-specific interpersonal correlates. These findings suggest that fatigue, loneliness, and source-specific support deficits may be important targets for hypothesis generation and future intervention research, while causal mechanisms require further confirmation using designs specifically suited for causal inference.