ABSTRACT
Research has suggested an increase in loneliness during the COVID-19 pandemic, but much of this work has been cross-sectional, making causal inferences difficult. In the present research, we employed a longitudinal design to identify loneliness trajectories within a period of twelve months during the COVID-19 pandemic in Belgium (N = 2106). We were particularly interested in the potential protective role of self-compassion in these temporal dynamics. Using a group-based trajectory modelling approach, we identified trajectory groups of individuals following low (11.0%), moderate-low (22.4%), moderate (25.7%), moderate-high (31.3%), and high (9.6%) levels of loneliness. Findings indicated that younger people, women, and individuals with poor quality relationships, high levels of health anxiety, and stress related to COVID-19, all had a higher probability of belonging to the highest loneliness trajectory groups. Importantly, we also found that people high in two of the three facets of self-compassion (self-kindness and common humanity) had a lower probability of belonging to the highest loneliness trajectory groups. Ultimately, we demonstrated that trajectory groups reflecting higher levels of loneliness were associated with lower life satisfaction and greater depressive symptoms. We discuss the possibility that increasing self-compassion may be used to promote better mental health in similarly challenging situations.
Disclosure statement
No potential conflict of interest was reported by the authors.
Ethical standards
The studies comply with the American Psychological Association (APA) ethical regulations for research on human subjects and all participants gave online informed consent, as approved by the institutional review boards of the principal investigator (Project 2021-13).
Notes
1 This study is part of a larger international and longitudinal research project on COVID-19, health behaviours and mental health (see Bigot et al., Citation2021; Dekeyser et al., Citation2023; Wollast et al., Citation2021).
2 The education data was not available for 4.5% of participants.
3 Table S1, available in the online supplementary material, presents the sociodemographic characteristics for the full sample, indicating that the subsample is representative of the full sample in terms of age, gender, and education.
4 Given the wording of the respective items between loneliness and relationship quality, we performed a confirmatory factor analysis (CFA) to test whether these two constructs were distinct. Model fit was assessed using multiple fit indices (Tanaka, Citation1993). The Root Mean Square Error of Approximation (RMSEA) should be ≤ .08, and the Comparative Fit Index (CFI) should be ≥.90 (see Byrne, Citation2004). The models were based on maximum likelihood estimation. CFA evidenced that the model distinguishing loneliness and relationship quality demonstrated good and greater model fit statistics (χ2 = 93.286, df = 8, CFI = .963, RMSEA = .071) as compared to the model combining both constructs (χ2 = 384.572, df = 9, CFI = .838, RMSEA = .141). All standardized factor loadings were significant and above the conventional threshold (> .40), and the correlation between both latent constructs was significant (r = −.26, p < .001), confirming that loneliness and relationship quality are related but distinct constructs.
5 The smallest cluster of the five-group model contains 9.6% that translates into 202 participants which generates sufficient power to detect the significance of slopes within trajectory classes (Frankfurt et al., Citation2016).