The COVID-19 pandemic and subsequent lockdowns profoundly and diversely impacted mental health and well-being. However, understanding individual experiences (both positive and negative) and their relationship to mental health, alongside other contextual variables, remained limited. Existing research often focused on narrow aspects of mental health, neglecting diverse psycho-socio-economic factors and self-perceived pandemic impact. Studies also suffered from recruitment bias due to targeted sampling methods. This research aimed to address these limitations by employing a large-scale, data-driven approach to investigate the complex interplay between the pandemic's impact, mental health, and a wide range of population factors. The study aimed to identify population segments most affected by the pandemic and to compare these findings with expert opinions. This was crucial for informing policies, healthcare decisions, and guiding researchers and clinicians. The multivariate nature of the relationships between pandemic impact and psycho-socio-economic profiles posed a major methodological challenge, necessitating the analysis of large-scale population data using multivariate and machine learning methods.

Prior research on the impact of pandemics on mental health was limited, with studies often focusing on narrow aspects of mental health and failing to consider a wide range of psycho-socio-economic variables that could potentially modulate the impact of the pandemic. Existing studies had also predominantly used promotional materials explicitly mentioning COVID-19 and targeted established cohorts, potentially leading to recruitment bias from non-representative subpopulations. This study aimed to address this gap by considering a broader range of factors, such as age, ethnicity, social networks, financial and occupational circumstances, caregiving responsibilities, pre-existing mental health symptoms, maladaptive online technology use, personality traits, and compulsive behaviors. The study used a comprehensive approach involving a large-scale population survey, including a bespoke Pandemic General Impact Scale (PD-GIS), to capture self-perceived pandemic impact, providing a broader understanding of the pandemic's mental health implications.

The study employed a large-scale online survey, accessible from December 26, 2019, through a website promoted via prominent media outlets. Data collection occurred in three epochs: pre-pandemic (December 2019-January 2020), early-pandemic (February-May 2020), and mid-pandemic (May-June 2020). Participants under 16 were excluded, as were those completing the questionnaire in under 4 minutes. The pre- and mid-pandemic epochs yielded the largest datasets. Sociodemographic data showed generally close correspondences across epochs, with the exception of a younger skew in the smaller early-pandemic sample, which was excluded from further analysis. The final analysis included 233,268 pre-pandemic and 112,046 mid-pandemic participants. Mental health was assessed using standard self-report measures (anxiety, depression, concentration, sleep, tiredness). In May 2020, the questionnaire was expanded to include the PD-GIS, a bespoke scale measuring self-perceived pandemic impact (positive and negative). Data analysis involved several steps: 1. Comparing pre- and mid-pandemic mean population mood scores using General Linear Models (GLMs) while controlling for various demographic and socioeconomic factors. 2. Quantifying the dimensionality of self-perceived pandemic impact using Principal Component Analysis (PCA). 3. Testing the explanatory power of positive and negative aspects of self-perceived impact on mental health measures using multivariate analyses. 4. Using GLMs to quantify the relative importance of population factors in predicting self-perceived pandemic impact. Canonical correlation analysis (CCA) assessed the relationships between PD-GIS subscales and mental health assessments. GLMs further investigated relationships between PD-GIS scores, sociodemographic variables, home context, and work arrangements. Finally, linear modeling explored the relationship between online technology use, personality traits, compulsivity, and self-perceived pandemic impact.

Analysis revealed subtle overall population differences in standard mental health measures between the pre- and mid-pandemic periods, with anxiety showing the most significant increase. However, significant differences were observed within specific subpopulations. Older adults experienced a greater increase in anxiety, while younger adults showed increased sleep duration and decreased depression, even though pre-pandemic levels were already elevated. Retired people, workers, homemakers, those with lower incomes, and individuals identifying as 'other' sex experienced the greatest increase in anxiety. The PD-GIS revealed surprising levels of agreement with positive statements regarding the pandemic's impact, such as increased time for loved ones and appreciating simple things in life. Strong agreement with negative statements included health concerns for loved ones and loss of leisure activities. PCA identified seven components underlying self-perceived pandemic impact: 1. More time, less stress/tiredness. 2. Disrupted lifestyle. 3. Increased health concerns. 4. Positive outlook. 5. More conflict at home. 6. Improved environment. 7. More time for people at home. CCA confirmed substantial relationships between these components and standard mental health measures, with the strongest correlations between health concerns and disruption of normal life. Age was the most significant predictor of self-perceived impact, with non-linear relationships observed across different age groups. Occupational status and cohabiting situations also showed strong associations. Healthcare workers reported less positive impact, while furloughed individuals experienced the most positive changes. People with pre-existing conditions, especially those predisposing to COVID-19 vulnerability, reported higher health concerns. Personality traits (insecurity, cognitive rigidity, reward drive) and technology addiction significantly predicted negative components of self-perceived pandemic impact (health concerns and disrupted lifestyle). Access to outdoor space was associated with more positive self-perceived impact.

The study's findings highlight the subtle yet significant impact of the COVID-19 pandemic on mental health, with variations across age groups and other demographic factors. The multidimensional nature of self-perceived pandemic impact underscores the need for tailored interventions. Older adults and healthcare workers require specific attention due to their disproportionate challenges. The study identifies access to outdoor space as a crucial factor linked to improved mental health, suggesting that environmental factors could play a significant role in mitigating negative pandemic impacts. The relationship between personality traits, technology use, and compulsivity with mental health provides valuable insights for designing interventions, emphasizing the need to consider both individual characteristics and the context of technology use. The study confirms the importance of considering a wide range of factors and the complexity of human behavior. The complex relationships highlighted necessitate multifaceted and personalized interventions.

This large-scale study provided valuable insights into the multifaceted effects of the COVID-19 pandemic on mental health in the UK. The study highlights the importance of considering individual differences and the complex interplay of various factors when designing and implementing interventions. Future research could explore longitudinal effects, test the effectiveness of specific interventions, and examine the longer-term impact on mental health. The PD-GIS may prove a valuable tool for future pandemic-related research and clinical practice.

The study's cross-sectional design limits causal inferences, although the large sample size and control for confounding variables mitigate this to some extent. The reliance on self-report measures also introduces potential biases, as does the inherent limitations of any non-probabilistic sampling approach. Future longitudinal studies are needed to confirm these findings and explore causal relationships.