A total of N = 886 university students in Germany participated in the cross-sectional online study between April and July 2023. The survey was conducted using the survey tool of the University of Potsdam. Participants were recruited nationally via email through universities’ student councils, with requests to spread the survey link among their students, as well as locally at the University of Potsdam via the participant pool of the Cognitive Sciences and by distribution of flyers at key locations on campus during mental health awareness days.

Inclusion criteria were student status, studying in Germany, age 18 and above, and giving informed consent. The latter was obtained from all participants, ensuring thorough understanding of their rights, the study's objectives, risks, and benefits. Abiding by institutional guidelines of the University of Potsdam, data collection maintained complete anonymity, strictly limiting sociodemographic information asked, and adhering rigorously to national and EU data protection laws. Participant characteristics can be found in Table 1.

The questionnaire commenced with a brief assessment of sociodemographic variables, namely gender (female, male, non-binary), age (in categories to further ensure anonymity, e.g., 21-25), relationship status, and study term.

To compare post-pandemic levels of psychological distress to students’ mental health in the beginning of the COVID-19 pandemic, we used data from a survey conducted at our university in April 2020 (Löscher, 2020). A total of N = 1,062 (n = 706, 66.5% female; n = 345, 32.5% male; n = 11, 1% non-binary) university students in Germany participated in this cross-sectional online survey. Participants were aged 18 to 51 years (M = 24.01; SD = 4.68). The procedure was analog to the assessment detailed for 2023, except for the distribution of flyers on campus.

All analyses were conducted in R (R Core Team, 2023). There were no missing data. However, for the study term question, some exclusions were made due to non-digit entries (e.g., “bachelor”, 23 cases). The level of significance for all analyses was α = .05. Analyses were conducted as previously registered, with minor deviations detailed below due to unmet assumptions (Hunsmann & Kühne, 2023S, March 31). First, we computed descriptive statistics. In exploratory manner, we examined the correlations between perceived burdens and psychological distress, applying the Bonferroni correction for multiple comparisons (Armstrong, 2014). We also explored differences in loneliness by gender.

Proceeding with our hypotheses, we examined the relationships between psychological distress, loneliness, and emotion regulation strategies, using Spearman’s rank correlations due to violations of the assumption of normality. To better understand how distressed students were in the aftermath of the COVID-19 pandemic, we used a Mann-Whitney U-test to compare students’ levels of psychological distress in both samples. The Mann-Whitney U-test was conducted due to violations in the assumptions of equal variances and normality. Effect sizes are reported, with r = .1 considered a small, r = .3 a moderate, and r = .5 a large effect (Cohen, 1988).

Next, we conducted hierarchical multiple regression analyses to predict psychological distress from loneliness and emotion regulation strategies, while controlling for health behaviors and person-related factors. First, we considered a baseline model with health behaviors and person-related factors. In a second step, we added the emotion regulation strategies suppression and reappraisal as predictors. Similarly, we added the predictor loneliness to the baseline model. Finally, we computed a multiple regression model predicting psychological distress from all variables. As the assumption of homoskedasticity was not confirmed for all regression models, we employed heteroskedasticity-consistent, robust estimations using the HC3 method in all regression analyses (Hayes & Cai, 2007; Long & Ervin, 2000; White, 1980). Where the assumption of normality of residuals was violated, we conducted bootstrapped regression analyses with 5000 bootstrap samples to assess the stability of the original regression models’ results. If not specified otherwise, the bootstrapped regression analyses supported the initial models.

Finally, we conducted analyses of variance (ANOVAs) to investigate differences in loneliness, emotion regulation strategies, and health behaviors between participants depending on their history of mental health conditions. Post hoc contrasts were computed using the Tukey Honestly Significant Difference (HSD) test. Effect sizes were calculated using omega squared (ω²) as recommended by Kroes and Finley (2023). Field’s (2013) recommendation for interpreting ω² was used (ω² ≥ 0.01 small, ω² ≥ 0.06 medium, ω² ≥ 0.14 large).

An overview over sociodemographic information is provided in Table 1, experiences related to the COVID-19 pandemic in Table 2.

Of those students who had a history of mental health conditions (n = 252; 28.4%), either currently or previously received psychotherapy, 58.3% (n = 147) reported worsening symptoms since the COVID-19 pandemic. Furthermore, 44.5% (n = 394) of participants were classifiable as psychologically distressed.

As can be seen in Figure 1, 10.1% of students reported feeling strongly or very strongly burdened by the COVID-19 pandemic. In comparison, the percentage of students reporting to feel strongly or very strongly burdened by the other stressors were 17.6% for the war in Ukraine, 26.9% for the energy crisis, and 45.8% for the climate crisis. Yet, there was a small but significant correlation between students’ distress and how burdened they felt by the COVID-19 pandemic (r = .26, p < .001). Likewise, there were small correlations between psychological distress and perceived burdens due to the energy crisis (r = .28, p < .001), the war in Ukraine (r = .25, p < .001), and the climate crisis (r = .22, p < .001, Bonferroni correction p < .0125).

Descriptive statistics for psychological distress, loneliness and emotion regulation are provided in Table 3.

A one-way ANOVA revealed a small yet significant difference in loneliness by gender, F(2, 883) = 8.37, p < .001, ω² = .016, 95% CI [.004, .039]. Post hoc contrasts showed that participants identifying as non-binary were significantly lonelier (M = 3.5, SD = 0.98) than participants identifying as female (M = 2.92, SD = 0.88) or male (M = 2.79, SD = 0.96). As for differences in loneliness by COVID-19 related group status, a t-test revealed that participants at risk for a severe course of infection reported significantly higher loneliness (M = 3.21, SD = 0.92) than participants not belonging to a risk-group (M = 2.86, SD = 0.91; t(884) = -3.19, p = .001, d = 0.39), and that participants who reported lasting consequences of a past infection reported significantly higher loneliness (M = 3.09, SD = 0.95) than participants without post-COVID ramifications (M = 2.87, SD = 0.91), t(884) = -2.10, p = .036, d = 0.24.

The Mann-Whitney U-test revealed a significant difference in psychological distress between students in 2020 and 2023, U = 332516, z = -8.211, p < .001, r = .19, with students in 2023 reporting significantly higher distress (Mdn = 0.74) than students in 2020 (Mdn = 0.45).

There was a large positive correlation between psychological distress and loneliness, r_s(884) = .56, p < .001. A small negative correlation was found between psychological distress and reappraisal, r_s(884) = -.18, p < .001, whereas distress and suppression were positively correlated, r_s(884) = .23, p < .001.

The baseline multiple linear regression model predicting psychological distress from health behaviors and person-related factors was statistically significant, explaining 17% of variance, see Table D1 (Appendix D, Supplementary Materials). Adding the emotion regulation strategies to the baseline model resulted in a statistically significant model explaining 24% of variance in psychological distress, see Table D2 (Appendix D). The model was better than the baseline model, ΔR² = .066, F(2, 866) = 38.5, p < .001. The strategy of suppression predicted higher psychological distress, and the strategy of reappraisal lower distress. The bootstrapped regression analysis closely aligned with initial model, however, belonging to the age group 36-40 was corrected as a nonsignificant predictor.

Similarly, the predictor loneliness was added to the baseline model, resulting in a statistically significant model accounting for 39% of the variance in psychological distress, see Table D3 (Appendix D). Here, an increase in loneliness was associated with an increase in psychological distress. The model with loneliness was significantly better than the baseline model, ΔR² = .209, F(1, 867) = 301.87, p < .001.

Lastly, an exploratory regression model with all predictors accounted for R² = 41% of variance in psychological distress, see Table 4 (see also Table D4, Appendix D).

While being the strongest of the models predicting distress, it only explained 2% more variance in distress than the model with loneliness (Table D3, Appendix D), ΔR² = .024, F(2, 865) = 17.99, p < .001. This underscores the importance of loneliness as a predictor for distress.

The ANOVA revealed a small effect of prior mental health in loneliness, F(2, 883) = 20.00, p < .001, ω² = .041, 95% CI [.02, .071]. Post hoc comparisons revealed that participants with current (M = 3.28, SD = 0.95) and past mental health conditions (M = 3.09, SD = 0.93) were significantly lonelier than those without (M = 2.77, SD = 0.88).

As for differences in emotion regulation by mental health status, no significant difference was found in suppression, F(2, 883) = 1.78, p = .169. There was also no significant difference in reappraisal, F(2, 883) = 0.59, p = .556.

As for the overall score regarding health behaviors, there was no significant difference between participants with past, current, or no mental health conditions, F(2, 883) = 0.57, p = .564. Investigating each behavior individually, we found a small difference in exercising regularly between participants depending on their mental health history, F(2, 883) = 6.73, p = .001, ω² = .013, 95% CI [.002, .033]. Post hoc comparisons revealed that participants with no mental health conditions (M = 2.9, SD = 0.99) reported to exercise more regularly than participants with current (M = 2.65, SD = 1.05) or past (M = 2.6, SD = 1.03) mental health conditions. There was also a small difference in mindfulness practice by mental health group, F(2, 883) = 11.05, p < .001, ω² = .022, 95% CI [.007, .047]. However, participants with no mental health conditions (M = 1.88, SD = 0.9) reported significantly less mindfulness practice than participants with current (M = 2.2, SD = 0.91) or past (M = 2.19, SD = 1) mental health conditions.

Loneliness was the strongest predictor for psychological distress. One explanation is to understand loneliness as a persisting consequence of the pandemic. Werner and colleagues (2021) reported loneliness being highly predictive of mental health issues in a longitudinal study, with loneliness during the pandemic only marginally predicted by pre-pandemic loneliness. Greater loneliness was reported in regions with more pandemic-related restrictions in a Norwegian study among students, however, the study also found that loneliness was linked to time spent on campus and declined again from 2021 to 2022 (Hysing et al., 2023). Conversely, loneliness could stem from existing mental health conditions. Students with diagnosed mental health conditions might struggle to seek social support or engage in fulfilling interactions (Perese & Wolf, 2005). As mental health conditions remain a sensitive topic, they may hesitate to confide in others (Schnyder et al., 2017; Schomerus et al., 2019).

Regarding gender differences in loneliness, our results align with existing findings (Hysing et al., 2023; Werner et al., 2021). A rise in female loneliness during COVID-19 has been linked to declining wellbeing (Lepinteur et al., 2022), likely due to gender gaps in other areas, e.g., finances or caretaking (Flor et al., 2022). However, the 3% non-binary students in our sample reported the highest loneliness. This percentage is in line with other German and international surveys (e.g., Ipsos, 2021). Transgender and non-binary individuals are more likely to experience discrimination and violence (Aparicio-García et al., 2018), with social support and LGBTQIA+ communities playing a vital role in mitigating negative mental health outcomes (Weinhardt et al., 2019). Thus, they may have been disproportionately affected by social restrictions. Alternatively, loneliness may have been high in this group before the pandemic (Aparicio-García et al., 2018).

Despite the importance of emotion regulation in psychopathology, we found no differences by mental health status. This may reflect the limitations of our narrow assessment, with adaptive emotion regulation increasingly seen as flexible (e.g., Aldao et al., 2015) and influenced by factors like strategy access and emotional awareness (Gratz & Roemer, 2004). In line, cognitive control and flexibility moderated the association between uncertainty intolerance and emotion regulation difficulties in a multi-wave pandemic study, affecting mental health (Godara et al., 2023). Our finding that health behaviors served as adaptive coping strategies generally aligned with previous research (Chen et al., 2020; Mata et al., 2021; Oftedal et al., 2019).

Psychological distress was high in this university student sample even after the pandemic, with loneliness notably prevalent in the most distressed individuals. This underscores the importance of addressing loneliness in young adults. We also found that adaptive emotion regulation and specific health behaviors, such as adequate sleep, exercise, and maintaining daily routines, were associated with better mental health cross-sectionally. Future research should monitor loneliness among university students longitudinally. Specific interventions could address loneliness (Ma et al., 2020; Masi et al., 2011). Targeted programs in universities would benefit from particularly focusing on non-binary and female students. Counseling services could facilitate support groups and therapy referrals, and universities could expand practical assistance for challenges commonly faced by female students, such as caregiving responsibilities. By addressing mental health openly and fostering support systems, significant strides may be made towards reducing loneliness among university students.