The COVID-19 pandemic significantly impacted mental health globally, with studies showing increased psychological distress during lockdown periods. While previous research identified risk factors and mental health trajectories, the dynamic interplay of symptoms over time remains less understood. This study aimed to address this gap by employing a network analysis approach, which examines the interconnectedness of symptoms and identifies central symptoms driving symptom cascades. Unlike traditional approaches focusing on composite scores, network analysis allows for a granular understanding of individual symptom dynamics and their mutual influence. Prior research suggested loneliness and feelings of worthlessness as potentially central symptoms in mental health networks during the pandemic, but longitudinal evidence was lacking. This study utilized data from the Understanding Society study, a nationally representative longitudinal study in the UK, to investigate symptom networks across three time points: pre-pandemic (2019), pandemic onset (April 2020), and during the pandemic's peak (November 2020 and January 2021). The study hypothesized that loneliness would be a central symptom driving other symptoms, but also explored other symptoms' roles in the evolving network.

Existing research on mental health during the COVID-19 pandemic largely employed composite scores, obscuring the complex interactions between individual symptoms. Studies demonstrated a surge in psychological distress during lockdowns, with peaks coinciding with periods of stricter restrictions. While identifying demographic and psychological risk factors (e.g., being female, young, lonely), these studies lacked a nuanced understanding of the temporal dynamics of symptom interactions. Previous cross-sectional network analyses provided some evidence for the centrality of loneliness and worthlessness in mental health networks during the pandemic, but these studies couldn't capture the unfolding of symptom changes over time. The current study aimed to fill this gap by conducting a longitudinal network analysis, enabling a more comprehensive understanding of symptom dynamics under conditions of maximum and repeated stress.

This study utilized secondary data from the Understanding Society UK Household Longitudinal Study, a nationally representative prospective cohort study. The analysis focused on data from three time points: a pre-pandemic survey (2019/2020), the first wave of the pandemic (April 2020), and two subsequent waves representing pandemic peaks (November 2020 and January 2021). Mental health symptoms were assessed using the General Health Questionnaire-12 (GHQ-12), a widely used measure of psychological distress, and a single-item measure of loneliness. The sample consisted of 17,761 participants in the first COVID-19 wave. Missing data, handled as missing at random (MAR), were imputed using predictive mean matching. Three temporal cross-lagged panel network models (CLPNs) were constructed to analyze the symptom networks across the three time points: pre-COVID to first peak, first peak to second peak, and second peak to third peak. LASSO regression was used to estimate the network connections, controlling for autoregressive effects. Centrality indices (in- and out-expected influence) were calculated to determine the influence of each symptom on others and its susceptibility to influence from other symptoms. Bootstrapping was used to assess the accuracy and stability of the network estimates.

Across the three CLPN networks, the connections between symptoms displayed high consistency. The strongest connections were consistently found between "thinking of self as worthless" and "losing confidence," and between "loneliness" and "feeling unhappy and depressed." Centrality analysis revealed that "loneliness" and "thinking of self as worthless" had the strongest out-expected influence (predicting other symptoms), while "feeling unhappy and depressed" and "could not overcome difficulties" exhibited the strongest in-expected influence (being predicted by other symptoms). This suggests that loneliness and low self-worth were key drivers of symptom escalation, while depression and difficulty overcoming problems were downstream consequences of symptom cascades. The study also observed a consistent cluster of strongly connected symptoms over time: loneliness, feeling unhappy and depressed, thinking of self as worthless, losing confidence, could not overcome difficulties, under stress, and lost much sleep. Other symptoms (e.g., ability to concentrate, feeling reasonably happy) showed less consistent connections.

The findings strongly support the hypothesis that loneliness and self-worth are central drivers of mental health symptom escalation during the pandemic. The consistent patterns across the three time points highlight the enduring nature of these influences under both acute and prolonged stress. The high in-expected influence of feeling depressed and the inability to overcome difficulties suggests these symptoms represent endpoints in symptom cascades, further reinforcing the importance of addressing upstream factors like loneliness and low self-worth. The resilience of the network structure despite fluctuating COVID-19 incidence and lockdown measures underscores the robustness of these symptom relationships. Surprisingly, the symptom "enjoying day-to-day activities" did not emerge as a key driver of the network, suggesting that cognitive appraisal and coping mechanisms may play a more crucial role in mitigating the impact of restrictions. The findings highlight a vulnerability-stress model, with pre-existing loneliness potentially initiating symptom cascades upon exposure to pandemic stressors.

This study provides novel longitudinal evidence for the centrality of loneliness and self-worth in driving mental health symptom escalation during the COVID-19 pandemic. These findings underscore the need for interventions targeting these specific symptoms to prevent wider network activation and mitigate mental health deterioration. Future research should focus on testing the causal nature of these relationships and developing targeted interventions that specifically address loneliness and self-worth, potentially through single-session interventions or broader societal programs. The findings also provide valuable insights for policymakers in designing strategies that mitigate the mental health consequences of future crises.

This study relied on self-reported data, potentially subject to biases. While multiple imputation addressed missing data, the network analysis approach is not fully compatible with multiple imputations, and this should be kept in mind. Furthermore, the single-item measure of loneliness may not capture the full complexity of this construct. The cross-sectional nature of the data within each time point limits the ability to definitively establish causal relationships between symptoms. Future studies could benefit from employing more granular data collection methods (e.g., ecological momentary assessment) and experimental designs to further investigate causal pathways.