Disentangling Loneliness, Depression, and Social Media Use: A Gaussian, Mixed, and Bayesian Network Approach in the EU Loneliness Study

Background: Loneliness, depression, and social media use (SMU) are increasingly interconnected phenomena in modern mental health research. While prior studies have demonstrated associations between these domains, their mutual dynamics and directional relationships remain unclear. Traditional analytical approaches often treat these constructs as unitary, failing to capture the complexity of symptom-level interactions. Aims: This study aimed to disentangle the relationships between loneliness, depression, and SMU by applying advanced network modeling approaches to a large and representative European sample. Method: Using a large sample from the EU Loneliness Survey (N = 25,646), we applied Gaussian Graphical Models (GGMs), Moderated Mixed Graphical Models (MGMs), and Bayesian Network Analysis to explore the structure and potential causal pathways among loneliness, depression, and SMU symptoms. Moderation analyses tested whether gender, religiosity, income, and education influenced the network structure. Results: Time spent on social media emerged as the most central symptom connecting loneliness and depression to SMU. Moderation analyses revealed that gender and religiosity significantly influenced specific network connections. In our sample, women exhibited stronger links between social rejection and emotional isolation, while religious individuals showed heightened associations between online engagement and problematic use. Bayesian network analysis identified a directional path from perceived lack of support to emotional disconnection, leading to depression and culminating in compensatory SMU. Conclusions: Our findings emphasize the central role of loneliness in triggering depressive symptoms and maladaptive SMU, with meaningful differences across sociodemographic groups. These insights support the development of targeted, symptom-level interventions in digital mental health.