Mental Health, Resilience and Sports Activity in the Initial Phase of The First COVID-19 Lockdown in Germany

Quarantine and social-distancing measures during the COVID-19 pandemic situation resulted in a radical change in lifestyle behaviors. While the reduction of total physical activity is assumed to negatively impact psychological health (higher stress and anxiety levels), regular sports activity during lockdown conditions has benecial effects on health. Mechanisms are however unclear. The present analysis therefore examined the associations of sports activity with mental health, and assessed whether this is due to a direct effect on experiencing positive emotions and mental health, due to a stress-buffering mechanism, and/or through protecting/enhancing resilience. Moderation analyses conrmed this direct effect of sports activity on affective symptoms and psychological wellbeing, but not life satisfaction. Furthermore, we found a moderation between sports activity and resilience. In those with lower resilience, higher sports activity was linked to better wellbeing and reduced affective psychopathology. Stress-buffering effects were not conrmed for either outcome and a resilience-protecting effect was seen for psychological wellbeing and affective symptoms, but not life satisfaction. Importantly, mental health levels of those participants with higher resilience were never reached. The gender-separated analyses showed that, in women, lower resilience accompanied by higher stress experience was associated with highest depressive symptoms and symptoms of anxiety. The resilience-protecting effect of sports activity was no longer evident in the gender-separated analyses.

On March 11 th , 2020, the World Health Organization announced the coronavirus (COVID-19) disease outbreak a pandemic 1 . In response, countries worldwide have had introduced protective measures like (self-)quarantine and physical distancing to decrease cases and prevent the virus' spread. The Federal and State Governments of Germany adopted comprehensive contact restrictions starting on March 10 th and a signi cant tightening of measures around March 27 th2 . At the beginning of May, the greatest restrictions were gradually lifted. During the summer months, restrictions were rather local and temporary, and of small to moderate degree 3,4 . In the fourth quarter of 2020, measures intensi ed again. During the rst lockdown, attention was soon focused on the psychological impact of the pandemic situation and the con nement measures with most studies reporting on negative psychological effects. Across various populations, anxiety, depression, and stress considerably increased 5 . It is therefore of utmost importance to understand what can be done to mitigate the negative consequences of pandemic situations and quarantine measures.
Lifestyle behaviors like exercise, diet and sleep play major roles in the prevention and treatment of mental illness 6 . The role of regular physical activity and sports activity in mental and physical health is particularly well-researched. A number of meta-analysis show sports activity to aid the prevention and management of mental illness 7,8 . Hence, an active lifestyle and sport are assumed effective means to reduce negative impact of con nement measures 9,10 . According to the model of possible effects of sports activity in the transactional process of disease development, ve possible mechanisms were proposed described as the direct, preventive, protective (stress-buffering), resource-creating and resourceprotective effect of sport 11 . One such resource in the context of sport is resilience, i.e. adapting, managing and negotiating adversity. Theory suggests resilient individuals to bounce back from negative experiences quicker and more effectively 12 . While in the past, resilience was primarily understood as a trait, today it is generally accepted that resilience exists on a continuum, resilience may differ across different domains of life and may change over time 13 . The present investigation adopted this theoretical assumption. Regular sports activity, more speci cally the related psycho-physiological adaptations, are important contributors and mechanisms of resilience 14 . Resilience, on the other hand, is known to be related to less anxiety/depressive symptoms and more positive emotions 15 . Finally, a body of research con rms the interplay between sports activity, resilience, and mental health 16,17 . Although, the exact directions of action are not yet understood. For the present study, we adopted the above mentioned model of sports' direct and stress-buffering effects on health and additionally considered resilience to be a resource for mental health.
In the context of lifestyle behaviors, current COVID-19-related literature shows that adopted preventive measures resulted in a radical change in lifestyle behaviors 18,19 . For physical and sports activity in particular, physical activities decreased signi cantly during the lockdown across all age groups 20,2122,23 . This reduction of total physical activity is assumed to have a profoundly negative impact on psychological health (higher stress and anxiety levels) and well-being 24,25 . Data on physical and sport activity during the COVID19 con nement measures from the German general population are however rare Page 4/20 26-28 and there is only initial evidence for the effects sport may have on reducing distress and fostering mental health during the current pandemic situation. In a large-scale, global cross-sectional study, Brand and colleagues showed that those who exercised almost daily during lockdown conditions had best subjective well-being 29 . Albeit with only small effects, another cross-country study revealed a correlation between change in physical activity from "before" to "during" home con nement and change in wellbeing, life satisfaction and depressive symptoms. Those participants who showed a decrease in their physical activity levels showed poorer psychological health 30 . Another cross-country study provided initial evidence for sport activity (single item measure) to mediate the relationship between depressive symptoms and burden caused by the COVID-19 pandemic 31 . While thus associations between sports behavior and mental health during the COVID-19 pandemic can be assumed, mediating and moderating variables are barely understood. From above summarized research we propose stress and resilience to be such variables.
Hence, the main question for this study was whether sports activity provides an advantage for overcoming the COVID19 pandemic situation and quarantine measures and if so, how. The present analysis therefore examined the associations between sports activity and mental health, and assessed whether this is due to a direct effect on experiencing positive emotions and mental health, due to a stressbuffering mechanism 32 , and/or through protecting/enhancing resilience 16 . Due to differences in stress perception and coping behaviors between men and women 33 , gender-speci c associations were assumed and tested.

Participants and Design
Data for this study stems from a larger longitudinal project aiming to characterize the impact of COVID-19-related restrictions on physical and sports activity, and subjective well-being. Data herein reports on the initial baseline assessment that started on April 7 th 2020, about 2 weeks after a variety of stricter lockdown measures were implemented throughout Germany 2, 4 . Though the baseline survey was accessible until July 6 th , only questionnaires completed until April 30 th were considered to cover the period with most serious restrictions in mobility. The longitudinal project involved ve bi-weekly repeated invitation to take part in a follow-up assessment. This data will be presented elsewhere (Strahler et al., in preparation). This study recruited a convenience sample from the general population and the survey was presented online using the platform SosciSurvey. The study was advertised using university mailing lists and social media. In agreement with the most recent version of the Helsinki declaration, participation was voluntary and informed consent was obtained. Before starting the survey, a study agreement was presented and participants were required to click a checkbox indicating that they agree to the terms. The only further inclusion criteria was being 16 years of age or older.
Overall, the baseline survey was started 1141 times and completed 901 times (79.0%). From these 901 data sets, 159 had to be excluded: 37 were questionnaires lled out with "no consent", six questionnaires were lled extremely fast (<6 min and relative speed index >2), one data set contained responses making a misunderstanding of the questionnaire likely, ve data sets provided unreasonable high sports (>40h/week) or physical activity (>98h/week, see also below), and 110 records date from the period after May 1 st . The nal data set thus comprised 742 subjects including 534 (72.0%) women and ranging in age from 16 to 83 years (mean: 28.13±11.46 years). Two subjects indicated "diverse" as their gender. Both data sets were excluded from gender-speci c analyses.

Measures
The complete survey lasted about 10 to 15 minutes and gathered data on anxious and depressive symptoms, psychological wellbeing, life satisfaction, physical and sports activity, stress, and resilience. This was complemented by sociodemographic and physical health data.
Mental health: Anxious and depressive symptoms were measured by the Patient Health Questionnaire-4 (PHQ-4) 34 . This ultra-brief measure consists of a 2-item depression scale and 2-item anxiety scale with response options ranging from "not at all" (=0) to "nearly every day" (=3). The composite PHQ-4 total score (ranging from 0 to 12) was used in the present report. The total score showed acceptable internal consistency (α Cronbach = .78). Wellbeing was assessed by the WHO Wellbeing Index (WHO-5) 35 and the Life Satisfaction Scale (L1) 36 . The 5 items of the WHO-5 (e.g., "In the past two weeks, I have felt cheerful and in good spirits") are to be scored from 5 ("all of the time") to 0 ("none of the time"). A sum score is created and multiplied by 4 to obtain a percentage scale from 0 (absence of well-being) to 100 (maximal well-being). Internal consistency was good herein (α Cronbach = .82). The L1 is a single-item measure recording the cognitive evaluation of one's own quality of life ("All things considered, how satis ed are you with your life these days? ") with an 11-point scale (0 "not at all satis ed" to 10 "completely satis ed"), with moderate re-test stability and adequate construct validity.
Physical and sports activity: Physical and sports activity were examined by means of the Physical Activity, Exercise, and Sport Questionnaire (BSA) 37 which is based on the Frequency -Intensity -Timeand -Type principle. The BSA records different leisure time and sport activities of the last four weeks. Participants record the frequency and duration (min) of each activity or sport episode. Finally, a "daily activity" index (including job-related and household activities), a "sport activity" index (e.g., swimming, athletic sports, jogging, or yoga), and a "total activity" index (min/week) are determined. According to the model proposed above and according to research showing moderate-and vigorous-intensity activity to be more bene cial for mental health as compared to low intensity activity 38 , present analyses focused on the sports activity index capturing planned, structured, and repetitive activities performed to improve or maintain physical tness. Daily and total activity were reported for descriptive purposes but did not enter multivariate analyses.
Stress and resilience: To examine general current stress levels (not speci c to the current lockdown situation) participants rated a single item "I fell stressed out" on a visual analogue scale with the poles 0 (not at all) and 100 (completely). Single-item measures of stress are often used when stress has to be monitored repeatedly and in large-scale survey research. Their content, criterion, and construct validity were shown to be satisfactory 39 . Resilience was assessed via four recently proposed items to capture momentary feelings of resilience, i.e. felling determined, cheerful, content, and being interested in the things one is doing 40 . While the authors proposed 7-point Likert scales, we decided to adapt responding to the stress item and also used visual analogue scales with the poles 0 (not at all) and 100 (completely).
The mean score was created for analyses and this measure showed good reliability (α Cronbach = .84).

Statistical analysis
First, descriptive statistics were performed for men and women separately. In a next step, we tested all observed variables regarding their distribution features. As not all of the variables were normally distributed, Mann-Whitney-U-Test was used to identify differences between men and women regarding mental health scores (WHO-5, PHQ-4, L-1), physical activity (total physical activity, activity of daily living, sports activity) resilience and stress experience.
We further used Spearman's rank correlation coe cients to determine the relationship between mental health (as indicated by WHO-5, L-1, and PHQ-4), physical activity (activity of daily living, sports activity, total activity), resilience and feelings of stress for women and men separately.
Lastly, we analyzed the impact of resilience, sports activity and feelings of stress as potential predictors of mental health and wellbeing using multiple regression analyses for the whole group as well as for men and women separately. We further added four product terms to the model in order to test for interactions between sports activity, resilience and stress. Prior to the analysis, we z-standardized all variables of interest. We further checked for multicollinearity. VIF was always < 10. All residuals were normally distributed as indicated by normal P-P-plots. There was no clear distribution pattern in scatterplot of residuals versus predicted values indicating homoscedasticity. Durbin Watson test revealed no autocorrelation between the included regressors.
JASP, version 14.0, was used for statistical analysis and partly for graphical illustration. P values <.05 will be considered signi cant. The Benjamini-Hochberg procedure (with δ = .05) was used in order to decrease false discovery rate (FDR) in multiple hypothesis testing.

Results
Descriptives: Mental health represented by WHO-5, PHQ-4 and L-1 Bivariate associations between physical activity scores, mental health (re ected by L-1, WHO-5, PHQ-4), resilience and stress experience in men and women The three activity indices were correlated with the mental health scores, resilience and stress experience in the entire group, as well as in men and women separately (for a detailed depiction, see Fig. 1). In the total sample, the subcategory total activity signi cantly correlated with WHO-5 (r = .134, p < .001) and PHQ-4 (r = -.125, p < .001). Sports activity though showed a signi cant correlation with resilience (r = .131, p < .001), WHO-5 (r =.174, p < .001), and PHQ-4 (r = -.163, p < .001). There were no correlations with activity of daily living (all p > .05).
For women, we found signi cant correlations between sports activity with WHO-5 (r = .205, p < .001), PHQ-4 (r = -.182, p < .001), and resilience (r = .156, p < .001). The activity of daily living score signi cantly correlated with PHQ-4 only (r = -.149, p < .001). For men, we found no signi cant associations between the activity parameters, mental health, wellbeing, resilience and stress. All correlations for the whole group and by gender can be found in Fig. 1A.
Multivariate analysis: Prediction of the mental health by resilience, stress experience and sports activity during lockdown Multiple regressions analyzed whether mental health (represented by L1, WHO-5, PHQ-4) was predicted by resilience, stress experience, sports activity, as well as their interactions. Regarding life satisfaction, increases in L1-scores correlated signi cantly with increases in resilience (β = .648, p < .001), as well as decreases in stress experience (β =-.109, p < .001) The inclusion of the product terms did not explain signi cant additional variance in the L1 score (R 2 Change = .003, F Change(4, 734) = 1.032, p > .05), revealing no signi cant interactions. The total variance explained by the model as a whole was R 2 = .460, F(3, 738) = 211.291, p < .001. Results further showed that the WHO-5 score was signi cantly moderated by resilience (β = .677, p < .001), stress (β = -.212, p < .001) and sports activity (β = .060, p = .014). The inclusion of product terms further revealed an interaction between resilience and sports activity (β = -.064, p = .009) re ecting that wellbeing of those participants with a low resilience and low sports activity is impaired the most (see Fig. 2A). However, this model explained only slightly more variance in the WHO-5 score (R 2 Change = .006, F Change (4, 734) = 2.436, p = .046). The total variance explained by the model as a whole was R 2 = .569, F(3, 738) = 326.882, p < .001. Moreover, increases in PHQ-4 scores correlated signi cantly with decreases in resilience (β = -.593, p < .001), decreases in sports activity (β = -.063, p = .017) and increases in stress (β = .263, p < .001). The inclusion of the product terms did explain only little additional variance in the PHQ-4 score (R 2 Change = .014, F Change (4, 734) = 5.201, p < .001) revealing signi cant interactions between stress experience and resilience (β = -.093, p < .001) re ecting a lower resilience as well as a higher stress experience during lockdown being associated with more anxious and depressive symptoms. Furthermore, we found a signi cant interaction between sports activity and resilience (β = .073, p = .006) (Fig. 2B). Thus, low sports activity and low resilience was associated with highest symptoms of depression and anxiety. The total variance explained by the model as a whole was

Summary of main ndings
The main question for this study was whether sports activity provides an advantage for overcoming a pandemic situation and quarantine measures. Therefore, sports activities' direct effects on mental health, stress-buffering, and resilience-protecting effects were examined in a cross-sectional study during the rst COVID19-related lockdown in April/May 2020. In addition, gender differences were tested. In this convenience sample, women showed higher anxiety and depressive symptoms but similar stress, resilience and activity scores. Across all participants, sports activity was related to higher wellbeing and resilience but lower affective psychopathology. Moderation analyses con rmed this direct effect of sports activity on affective symptoms and psychological wellbeing, but not life satisfaction. Furthermore, we found a moderation between sports activity and resilience. In those with lower resilience, higher sports activity was linked to better wellbeing and reduced affective psychopathology. Stress-buffering effects were not con rmed for either outcome and a resilience-protecting effect was seen for psychological wellbeing and affective symptoms, but not life satisfaction. Importantly, mental health levels of those participants with higher resilience were never reached. The gender-separated analyses showed that, in women, lower resilience accompanied by higher stress experience was associated with highest depressive symptoms and symptoms of anxiety. The resilience-protecting effect of sports activity was no longer evident in the gender-separated analyses.
Direct effects of sports activity on mental health, stress and resilience during lockdown The direct effect of sports activity on mental health, feelings of stress and resilience is well documented in the general literature 7,8,15 . In contrast to previous reports, however, effects can only be regarded smallsized. This might be traced back to the heterogeneity of sports activity levels in the current sample without specifying intensity of such activities. Here, moderate to vigorous intensity seems more bene cial for mental health than light intense activities 42 . However, studies conducted during COVID19-related mitigation measures con rm these rather low correlations 24, 25, 31 . Unexpectedly, there was no correlation between sports activity level and stress in the present sample. This contrast previous reports of the stress-reducing effect of physical activity and exercise 43 . Relatively low stress levels (about 38 on a scale from 0 to 100) in our sample could have resulted in a oor effect. Studies in individuals at risk for heightened stress during a pandemic situation are still pending and need to con rm this assumption of critical stress levels beyond which sports activity shows its bene cial effects. In chronically stressed populations, such as caregivers, stress is linked to generally low physical activity levels 44 . While this may be understood as activity's "open potential", that is highly stressed individuals might bene t more from high activity, such populations are also less likely to initiate physical and sports activity 45 . Understanding possible barriers for mental health-enhancing activity levels remains an open task for future research.
Overall, present results con rm sports activity as an effective mean to increase mental health and wellbeing. To contribute to our understanding of moderating variables, the present study consideredbesides stress -links between sport and resilience.

Prediction of mental health by resilience and stress experience moderated by sports activity
From the model of possible effects of sports activity in the transactional process of disease development, we also tested sports' stress-buffering capabilities on health and additionally considered resilienceprotective mechanisms. In contrast to our assumptions, there was no indication of a stress-buffering mechanism, neither for measures of wellbeing nor affective psychopathology. This also contradicts larger epidemiological studies 46 as well as recent COVID19-related research where overall physical activity level was predictive of general stress with a medium effect size 32 . What present data, however, showed was a resilience-protecting mechanism for sports activity on psychological wellbeing and affective symptoms. In participants with higher resilience scores, mental health was unrelated to sport activity. High-resilient individuals always scored above the group mean. In those with lower resilience scores, sports activity appears to have a resilience-protecting effect in addition to the direct effect on mental health. Different psychobiological mechanisms are proposed to underlie such an effect. There is evidence for physical tness to optimize the neuroendocrine and physiological stress response, to promote an anti-in ammatory state, and to improve certain neuronal processes (e.g., neuronal plasticity or the expression of growth factors) [47][48][49][50][51] . In addition, regular sports activity results in gains in selfregulation 52 and provides another coping resource via social support, self-e cacy or better sleep quality 53 . This line of research highlights the potential of sports activity for mental health, especially during times of crisis.
As also noted above, when considering the positive effects of physical and sports activity on mental health, it is also important to keep in mind that the initiation or maintenance of activity depends on the individual's mental state and condition 45 . Psychological stress levels (and thus lower mental well-being) predicted lower physical and sports activity as well as higher sedentary time. Besides evidence for this behavioral inhibition effect under stress, there is also evidence for stress-related behavioral activation.
Some studies suggest increased activity levels under stress. Here, it can be concluded that activity was used as a coping strategy to better manage the stress load. Studies that examined activity levels before as compared to during the pandemic did so only retrospectively 20,2122,2329,30  Gender-speci c associations Due to gender-related differences in perception and coping behaviors 33 , this study also examined genderspeci c associations. Gender-separate analyses no longer hold the main effect of sports activity on wellbeing and affective symptoms. Again, our rather healthy sample and the limited range in mental health impairment may explain the missing effect. Furthermore, the moderation between stress and resilience on depression and anxiety levels was only found among women. Due to the unequal sample size and the relative small sample of men, a sample-size bias must be taken into account in the interpretation.
However, the fact that stress had the most impairing effect in women with low resilience corroborates recent research showing meaning and purpose in life to be bene cial for various dimensions of mental health across cultures and life span 59 . Neither among men nor among women, a resilience-enhancing effect of sport could be shown. This contrasts to earlier COVID19-related literature that suggests physical activity to be a more important mental health factor in both genders with slightly higher associations among women 24 . Important to mention is that in the present study all associations were only smallsized.
Limitations Some limitations of our study must be considered. First, with our survey procedure we had to rely on selfreported physical and sports activity levels. Several aspects of bias may have accompanied this selfreport, that is social desirability, selective recall or introspective ability. While this study anonymously collected data, thereby reducing social desirability bias, current stress and mental health levels may have limited the individual's ability to assess their behaviors and health status accurately. Physical activity assessment also did not allow to explore different levels of intensity or mode of activity. From previous research, moderate to vigorous intensity is expected to be more bene cial for mental health than light intensities 42 . Another limitation was the cross-sectional design of our study. Having not assessed pre-COVID19 mental health and activity levels precluded us to explore the effects of pre-pandemic levels or whether behavioral changes in response to restrictions rather than the current behaviors are more decisive. This could be interesting because Lesser and Niehuis 60 showed that the positive associations between sports and wellbeing were primarily found in people who had previously been inactive and started sporting activities with the beginning of the COVID19 restrictions. Furthermore, our recruitment procedure could have resulted in some selection bias. Relatively high activity and low stress levels in our sample may underlie small effect sizes and limit generalizability.

Conclusion
Taken together, this study provides evidence that sports activity was related to higher wellbeing and resilience but lower affective psychopathology during the rst COVID19-related lockdown in April/May 2020. This was clearly demonstrated by a moderation effect between sports activity and resilience. People with low resilience in particular bene t from sport activities in terms of their well-being. At the same time it became clear that in women, lower resilience accompanied by higher stress experience was associated with highest depressive symptoms and symptoms of anxiety, and the resilience-protecting effect of sports activity was no longer evident in the gender-separated analyses. Future studies should be longitudinal and include previous activity levels in order to investigate the relationship between sport and well-being in more detail. The gender-speci c results should also be examined with the inclusion of other psychological and biological characteristics.

Figure 2
A Associations between low/high sports activity, resilience, and WHO5 B and PHQ4. B Associations between low/high sports activity, resilience, and PHQ4.