During trying times, a number of factors can help individuals survive, manage and adapt. Bassi et al. (2021) found that Italian health workers classified as thriving individuals were less likely to have post-traumatic stress (PTSD) during the pandemic. The term thriving denotes the state of fully functioning in mental, physical, and social terms (Su et al., 2014). Thriving includes a range of dynamic factors including, but not limited to, gratitude (Emmons et al., 2003; Cheng et al., 2015; Otto et al., 2016), belonging (Baumeister and Leary, 1995), Social- contribution, integration, and actualization (Provencher and Keyes, 2011), meaning (Schueller and Seligman, 2010), pride (Tracy and Robins, 2007; Williams and DeSteno, 2008; Fredrickson, 2013), compassion (Seppala et al., 2013), and learning from the situation (Jenkins and Mostafa, 2015). Broader literature about mental health has highlighted the protective role thriving has in buffering against mental illness (Provencher and Keyes, 2011).

Additional lines of investigations point to the contribution of regular routines to mental health. A recent longitudinal study conducted during the three first months of the pandemic in Germany (Bendau et al., 2020) found that the following factors protected against anxiety and depression: self-efficacy, normalization of routines, maintaining social contacts, and knowledge about where to get medical support. Moreira et al. (2021), investigating the same topic during the COVID-19 outbreak in Portugal, found that working (online or in-person) and physically exercising on a regular basis, not having previous psychological/physical diseases, not consuming COVID-19 information and doing remote psychotherapy served as protective factors for mental health. These results align with a body of literature that points out that regularity has a beneficial effect on mental health (Sano et al., 2017; Murray et al., 2020) in the same way unpredictability in the environment is seen as a potential risk for later mental illness (Glynn and Baram, 2019). Other works showed that regularity of sleep, exercise and social rhythm correlated with improved mental health and well-being (Margraf et al., 2016; Boland et al., 2019; Logan and McClung, 2019).

In addition to person-focused factors, society-directed attitudes can also play a protective role in the context of the pandemic. Although some authors claim that societal trust increases after natural disasters due to the shared need to overcome the event (e.g., Toya and Skidmore, 2014), others maintain that disasters can foster suspicion conspiracy theories about the event (Wilson and Rose, 2014). These society-level outcomes, in turn, impact mental health. O'Hara et al. (2020), for instance, found that in countries that distrusted the government, an increase in policy stringency was associated with men reporting more depression—but no more worries—and women reporting both worries and depression. Thus, trust in government can impact well-being (Helliwell and Huang, 2008), and low perceived efficacy of governmental actions can reduce mental health (Mækelæ et al., 2020), especially during pandemic times when governments have to impose behavioral restrictions.

Finally, decades of research demonstrate that social connections are vital to the well-being and coping with difficult situations (Sibley et al., 2020). Thus, social support and close and caring relationships may both help individuals cope with life's adversities, as well as foster growth and development (Feeney and Collins, 2014). Social support is well-known to be a protective factor for mental disorders (Puschner, 2018), including paranoia (Freeman et al., 2011; Crush et al., 2018) and depression (Høifødt et al., 2020).

Numerous risk factors are also present during trying times, having an adverse effect on the physical and mental health of individuals. Bendau et al. (2020) found that anxiety and depression are exacerbated by routine suppression, unhealthy diet, reduced physical activity, increased substance abuse, and a longer daily time thinking about the pandemic Xiong et al. (2020), in turn, found that women, younger people (≤40 years), individuals with chronic/psychiatric illnesses, unemployed, students, and people frequently exposed to COVID-19 news experience more negative impact on their mental health in eight countries (China, Spain, Italy, Iran, USA, Turkey, Nepal, and Denmark). An additional study conducted in China (Guo et al., 2020) has identified, in addition to the factors mentioned above, the following risk factors: reduced income, having family members with chronic diseases, concerns related to COVID-19 infection for themselves/family members, living alone, having family conflicts, having sedentary time per day, and worsened sleep quality. Furthermore, a study conducted in the USA found that fear, worry, and threat were significant predictors of both depressive and anxiety symptomatology, even after controlling social vulnerability measures (Fitzpatrick et al., 2020).

The studies above suggest that excessive worry, catastrophizing thoughts, feeling scared about COVID-19 infection together with other fears, and experiencing a lack of control are all components of a negative mindset related to poor mental health. These studies also highlight vulnerability conditions, for example low socio-economic status (Link and Phelan, 1995; Reiss, 2013), low levels of education (Araya et al., 2003), presence of financial worries (Bareket-Bojmel et al., 2020), unemployment (Xiong et al., 2020), or reduced income (Guo et al., 2020), are risk factors for the mental health of specific importance during unpredictable times, such as a pandemic. Finally, the frequent exposure to COVID-19 news (Bendau et al., 2020; Xiong et al., 2020) accompanied by low perceived efficacy of governmental actions (Mækelæ et al., 2020) can make people react with suspicion and develop conspiracy theories about it (Wilson and Rose, 2014). According to Uscinski et al. (2020), during the COVID-19 outbreak there has been an increase in irrational beliefs or conspiracy theories, possibly due to decreased social interactions (Graeupner and Coman, 2017), potentially leading to detrimental outcomes for individuals (Bierwiaczonek et al., 2020) and societies alike (Jolley and Paterson, 2020; Romer and Jamieson, 2020).

Although some of the aforementioned factors are fairly universal, other factors depend to varying degrees on local spatial and temporal contexts. For example, countries vary in their age distribution, levels of trust in the local government, the prevalence of higher education among citizens, and degree of social equality to name a few parameters. Opinion papers have highlighted the potential contribution of psychological science to coping with the pandemic (Arnot et al., 2020; Garfin, 2020; van Bavel et al., 2020) and a couple of reviews (Serafini et al., 2020; Talevi et al., 2020) and empirical papers have investigated the impact of COVID-19 on the mental health of individuals (e.g., Bendau et al., 2020; Martínez et al., 2020). However, only some of these studies about COVID-19's detrimental impact on mental health have been conducted across countries (see e.g., Gobbi et al., 2020; Xiong et al., 2020; Alzueta et al., 2021; Gato et al., 2021).

Several international organizations published first recommendations (e.g., WHO, 2020) highlighting potential risk and protective factors that might assist in the prevention of mental health issues arising from the COVID-19 pandemic. Here we aimed to further characterize the robustness of factors that help maintain mental health during the COVID-19 outbreak through a two-wave observational study conducted across five countries. The goal of this work was two-fold. First, we wanted to determine whether the prolonged disruption of normal life and imposed social restrictions increased general distress and paranoia. Second, we aimed to identify which factors contribute to maintaining mental health during the pandemic. We hypothesized that the negative effects of social restrictions will be attenuated by the presence of protective factors (having high trust in the government/authorities, thriving, exercising, engaging in actions for own and others' well-being, maintaining a regular schedule, and having no financial worries). We further hypothesized that these negative effects will be amplified by the presence of risk factors (having high perceived risk, financial worries, and lower education). Identifying the factors that improve mental health (by reducing general distress and paranoia) can assist governments worldwide in handling the long-term social and economic costs associated with coping with the societal and psychological aspects of a pandemic (Nicola et al., 2020).

The present study used a longitudinal observational design with two waves of data collection (April and July 2020) in a convenience sample composed of participants from five countries: Brazil, Colombia, Germany, Israel and Norway. In mid-April 2020 we launched a survey in seven languages targeting five countries: Norway (Norwegian), Germany (German), Israel (Hebrew and Arabic), Colombia (South American Spanish), Brazil (Brazil-Portuguese). We included these specific countries to allow rapid data collection in an early stage of the pandemic (see Mækelæ et al., 2020). The survey was also available in English in all countries but Israel.

The first wave (W1) occurred in mid-April 2020 and was composed of over 2,200 participants. The second wave (W2) happened in July and was composed of over 700 participants who took the follow-up survey. Table 1 reports the sample characteristics for the five countries collected in wave 1.

The inclusion criteria were to have internet access and to be older than 18 years old. The exclusion criteria were to complete the survey in <3 min and to answer fewer than 70% of the items on a scale. Participants were recruited via social media (FB ads) and snowballing. All participants were encouraged to answer all items. Responding took around 15 min in April and around 10 min in July. All participants provided their informed consent and they did not receive any compensation. After answering the first wave, we asked respondents whether they would volunteer to partake in a follow-up in ~2 months later. If they consented they were transferred to a new survey collecting their email addresses.

Our survey included several distinct constructs, some developed anew, e.g., trust scale, thriving and negative mindset scale, others modified from established scales (e.g., CORE-10, CAPE-42, epistemic belief), across four categories. We describe the measures for each category below.

To test which factors predict general distress (hypothesis 1) we conducted a linear mixed model with the general distress score (CORE-9) as the outcome, country as a random effect and thriving, regularity, trust, financial worry, paranoia, negative mindset, perceived risk, perceived efficacy of actions, gender, and education as the predictors. Survey distribution time (April or July) was entered as a fixed effect.

To address hypothesis 2 we run a linear mixed model for the paranoia score (CAPE-P10) as the dependent variable and as fixed effects: thriving, regularity, trust, financial worry, negative mindset, general distress, perceived risk, perceived efficacy of actions, gender and education. Country was entered as a random effect and wave as fixed effect. We also run a linear mixed model with the same predictors complemented with the conspiracy score for the April data.

Since we expected a positive relationship between general distress and paranoia (Sun et al., 2019; Mækelæ et al., 2020), we applied mediation analysis with general distress being the predictor, paranoia score the outcome and thriving, regularity, trust, negative mindset, perceived risk and perceived efficacy as mediators.

To examine the predictors of satisfaction and dissatisfaction with the actions taken to counteract the outbreak in one's country (hypothesis 3) we conducted a generalized mixed model with satisfaction as the binary outcome variable with the following predictors: thriving, regularity, trust, financial worry, conspiracy score, paranoia, negative mindset, general distress, perceived risk, efficacy of actions, gender and education. Country was included as a random effect. For comparing April and July (fixed effect) the predictor conspiracy score was omitted.

To examine the relationship between usual day activities and general distress we performed correlation analyses and compared April and July correlations using z transformation.

We used frequentist analysis and a significance criterion of p < 0.05 for the two mixed models and for the logistic regression (pre-registered hypotheses). All predictors and outcomes were centered. For the exploratory analysis we corrected for multiple testing. We used Jamovi (The Jamovi Project, 2021) and R (R Core Team, 2017) for data analysis and visualization. Results without country as a predictor are reported in the Supplementary Material (SOM).

The analyses in this study were formally pre-registered. However, the analyses presented below deviate from the pre-registration and are only conceptually similar, as our statistical approach focused on longitudinal outcomes and required us to collect over 1,000 responses in July. In reality, we were able to collect only 731 valid responses in July.

There is considerable variation in the recommendation of required samples for multiple and hierarchical regression analysis. We follow a rule of thumb for multiple regression (Brysbaert, 2019), i.e., 100 participants plus another 100 for every predictor. Since we had up to 11 predictors, we aimed to recruit at least N = 1,200 participants to partake in each survey.

The linear mixed model testing our first hypothesis explained 44.7% of the variance. There was no significant change in distress from April (Mean= 9.73, SD = 0.291) to July (Mean = 10.11, SD = 0.395). Perceived efficacy of governmental reactions and trust in authorities were not significant predictors of general distress. Thriving and maintaining a regular schedule demonstrated protective qualities, i.e., higher scores yielded lower general distress. A negative mindset, paranoia, high perceived COVID-19 infection risk, financial worries and being female predicted more distress. The random effect of country was significant (LRT¹ = 9.27, p = 0.002). Table 2 reports the estimates of the mixed model and Figures 2, 3 shows the estimates per country and for the April and July survey, respectively.

The linear mixed model for paranoia and for April and July explained 25.5% of the variance. The fixed effect of wave was small, F_{(1, 2,498)} = 4.042, p = 0.045. We therefore report the model for April only but including the conspiracy score as predictor. The mixed model for paranoia in April explained 24.5% of the variance. The paranoia score differed by country (LRT = 75.4, p < 0.001). The less regularity and trust respondents reported, and the lower their level of education, the more paranoia they experienced. Being male was also associated with paranoia, so was a higher conspiracy score and general distress. Financial worries, negative mindset, perceived risk, perceived efficacy and trust were not associated with more paranoia. Table 3 reports the estimates and Figure 4 shows that there is more variation between the indices than the countries.

We found the same pattern of results when analyzing data from April and July separately, using all data irrespective of country. Thriving and regularity were protective whereas paranoia, negative mindset, financial worry and being female were risk factors for general distress (see Supplementary Material for details). Similarly, general distress was a significant predictor for paranoia. The overall pattern for general distress and paranoia emerged also when using the longitudinal subset (n = 525), i.e., participants who took part both in April and July (see Supplementary Material for details, page 20).

As Figures 2–4 show there is a positive relationship between general distress and paranoia. We therefore investigated (using the April data) whether any of the protective and risk factors could mediate the relationship, i.e., would there be a significant indirect effect. Mediators were thriving, regularity, trust, negative mindset, perceived risk and perceived efficacy. The direct effect between distress and paranoia explained 98% of the variance whereas the indirect effect of thriving explained only 2%, Z = 1.75, p = 0.079. Similarly a negative mindset did not mediate the relationship, 2.68%, Z = 1.65, p = 0.1. On the other hand regularity significantly mediated the relationship between distress and paranoia, 9.27%, Z = 5.03, p < 0.001, so did trust: 17.1%, Z = 9.27, p < 0.001, perceived risk: 2.51%, Z = 2.62, p = 0.009 and perceived efficacy: 7.46%, Z = 5.86, p < 0.001. Thus, a more regular schedule, higher trust in authorities, lower perceived risk and higher perceived efficacy of actions reduced the association between distress and paranoia.

Finally, our third hypothesis that predicted more distress and paranoia among those who think that their country overreacted was not confirmed. Firstly, the relative number of participants stating that their country overreacted was lower in July than in April [estimate = −4.42, exp(B) = 0.012, z = −4.84, p < 0.001]. Secondly, applying generalized mixed model (logit link function, country as cluster variable/random intercept) for the April data (Table 4), dissatisfied participants had a significant lower score on thriving (χ² = 21.05, p < 0.001), trust in authorities (χ² = 17.11, p < 0.001) and perceived efficacy (χ² = 10.09, p < 0.001), but also demonstrated less characteristics of a negative mindset (χ² = 27.35, p < 0.001) and endorsed conspiracy theories more (χ² = 36.73, p < 0.001). They showed no differences in general distress, financial worries, regularity, education, gender or paranoia. Their perceived risk was somewhat higher but not significantly so. Overall the model explained 68.5% of the variance (R² conditional). Applying a similar model without the conspiracy score but with wave as fixed effect yielded very similar results, i.e., thriving, trust in authorities, perceived efficacy and negative mindset were significant predictors. In addition, lower perceived risk significantly predicted dissatisfaction with the governmental response [estimate = −0.02, exp(B) = 0.978, z = −3.102, p = 0.002].

In addition to the data described so far, we have also collected data in March 2020, investigating the perceived efficacy of COVID-19 restrictions and how those affect mental health (Sun et al., 2019; Mækelæ et al., 2020). The following will capitalize on similarities in the collected measures to cross-sectionally compare general distress, paranoia, conspiracy score and perceived risk, four measures that were identical (distress and paranoia) or had overlapping items (conspiracy score, perceived risk) in all three surveys, see Figure 5.

Distress increased from March to April, F_{(2, 4,671)} = 11.851, p < 0.001, ηp2 = 0.004. Paranoia, on the other hand, decreased in all countries from March to April, F_{(2, 4,677)} = 16.189, p < 0.001, ηp2 = 0.006. There was a large reduction in perceived risk from March to April and an even greater reduction in July, F_{(2, 4,656)} = 359.994, p < 0.001, ηp2 = 0.129. Finally, we compared the conspiracy score in March with April, and this score increased, F_{(1, 3,891)} = 11.981, p < 0.001, ηp2 = 0.003, indicating that in April participants were more likely to endorse Covid-19 related conspiracy theories. Details by country and the interactions can be found in the Supplementary Material, page 21ff.

Participants were asked to indicate how much time on a usual day they engage in each of 17 activities. Apart from procrastination we assumed that all other activities are either neutral (e.g., child care) or beneficial for mental health (being outdoors). Briefly, we found that procrastinators had a higher score on general distress than participants who spent no or little time procrastinating (ρ = 0.249, p < 0.001). General distress was also lower among those who exercised (ρ = −0.131, p < 0.001) and could and did go outdoors (ρ = −0.107, p < 0.001). Details are reported in the Supplementary Material, page 30.

Our study is a convenience sample, biased toward educated and female respondents and only very few have had a SARS-CoV-2 infection or a close family member who have had it. Particularly in Brazil, a country with high social inequality and unequal access to the internet (Silva et al., 2020), our sample is mostly from the middle class.

To ensure a high completion rate of the survey we limited the number of protective and risk factors. We did not ask about sleep quality, though sleep quality is beneficial for mental health, e.g., Lüdtke et al. (2021). We used an abbreviated version of the CAPE positive scale and the short form of the clinical outcome (CORE) measure. We also did not ask about protective behavior, or used scales to measure anxiety.

The different scales implemented in this study generally had good internal consistency, both across and within countries (see Supplementary Material for details). One notable exception pertains to the perceived efficacy of actions scale, with internal consistency around 0.6. Accordingly, the lack of effect of this scale on distress and paranoia, as well as its role in mediating the relationship between distress and paranoia, should be interpreted with caution.

The selection of the five countries is due to the researcher's location and access to those data. Our focus was not in comparing those five countries, quite the opposite. We set out to find similarities despite known economic and cultural differences.

Our study exploited the unprecedented opportunity to measure general distress and paranoia in the general population across five countries varying in their socio-economy during the early phase of the COVID-19 outbreak. Despite large differences in the countries' welfare system and handling of the pandemic, we found the same psychological factors being associated with distress and paranoia. Maintaining a regular schedule was greatly beneficial for both general distress and retaining low paranoia. Across all countries, thriving was highly beneficial for reducing general distress, whereas financial worries and a negative mindset were associated with increased general distress. In summary, our data show the remarkably adaptive and resilient nature of human beings as they grow accustomed to new situations when they have a supportive community and a sense of meaning in life. We hope that by shedding light on the factors contributing to adversarial growth our society will be better prepared for the upcoming events in this pandemic and in future prolonged negative society-level experiences.