|
article-title
|
Who is the Most Stressed During the COVID‐19 Pandemic? Data From 26 Countries and Areas
|
|
alt-title
|
Perceived Stress During the Pandemic
|
|
alt-title
|
KOWAL ET AL.
|
|
abstract
|
Background
To limit the rapid spread of COVID‐19, countries have asked their citizens to stay at home. As a result, demographic and cultural factors related to home life have become especially relevant to predict population well‐being during isolation. This pre‐registered worldwide study analyses the relationship between the number of adults and children in a household, marital status, age, gender, education level, COVID‐19 severity, individualism–collectivism, and perceived stress.
Methods
We used the COVIDiSTRESS Global Survey data of 53,524 online participants from 26 countries and areas. The data were collected between 30 March and 6 April 2020.
Results
Higher levels of stress were associated with younger age, being a woman, lower level of education, being single, staying with more children, and living in a country or area with a more severe COVID‐19 situation.
Conclusions
The COVID‐19 pandemic revealed that certain people may be more susceptible to experience elevated levels of stress. Our findings highlight the need for public health to be attentive to both the physical and the psychological well‐being of these groups.
|
|
sec
|
Background
To limit the rapid spread of COVID‐19, countries have asked their citizens to stay at home. As a result, demographic and cultural factors related to home life have become especially relevant to predict population well‐being during isolation. This pre‐registered worldwide study analyses the relationship between the number of adults and children in a household, marital status, age, gender, education level, COVID‐19 severity, individualism–collectivism, and perceived stress.
|
|
title
|
Background
|
|
p
|
To limit the rapid spread of COVID‐19, countries have asked their citizens to stay at home. As a result, demographic and cultural factors related to home life have become especially relevant to predict population well‐being during isolation. This pre‐registered worldwide study analyses the relationship between the number of adults and children in a household, marital status, age, gender, education level, COVID‐19 severity, individualism–collectivism, and perceived stress.
|
|
sec
|
Methods
We used the COVIDiSTRESS Global Survey data of 53,524 online participants from 26 countries and areas. The data were collected between 30 March and 6 April 2020.
|
|
title
|
Methods
|
|
p
|
We used the COVIDiSTRESS Global Survey data of 53,524 online participants from 26 countries and areas. The data were collected between 30 March and 6 April 2020.
|
|
sec
|
Results
Higher levels of stress were associated with younger age, being a woman, lower level of education, being single, staying with more children, and living in a country or area with a more severe COVID‐19 situation.
|
|
title
|
Results
|
|
p
|
Higher levels of stress were associated with younger age, being a woman, lower level of education, being single, staying with more children, and living in a country or area with a more severe COVID‐19 situation.
|
|
sec
|
Conclusions
The COVID‐19 pandemic revealed that certain people may be more susceptible to experience elevated levels of stress. Our findings highlight the need for public health to be attentive to both the physical and the psychological well‐being of these groups.
|
|
title
|
Conclusions
|
|
p
|
The COVID‐19 pandemic revealed that certain people may be more susceptible to experience elevated levels of stress. Our findings highlight the need for public health to be attentive to both the physical and the psychological well‐being of these groups.
|
|
body
|
INTRODUCTION
The COVID‐19 pandemic brought the world to a halt. Since early 2020, social life has changed for many people around the world. Government restrictions and new social norms led to a reduction in mobility (Google, 2020), avoidance of public transport, cancellation of the majority of large events such as concerts, festivals, religious and sports events, and the temporary closing down of meeting places such as cafes, restaurants, museums or theatres. The COVID‐19 outbreak and measures undertaken by almost all the countries around the world pose numerous threats to people's psychological well‐being; thus, we believe that focusing on our closest social environment is greatly needed. Both relative lack of social relations (Tay et al., 2013) and negative emotions (Huppert, 2009) strongly predict overall mortality and disease outcomes, and the threat of SARS‐CoV‐2 may increase anxiety levels (Karwowski et al., 2020) and prejudice toward other nationalities (Sorokowski et al., 2020), rendering any effort to understand stress in isolated or quarantined individuals during the COVID‐19 pandemic even more important. Moreover, to understand possible negative effects, it is important to unpick the pre‐existing factors that can predict the stress levels of people in quarantine or isolation.
Other People
One such factor is the presence (or absence) of other people. To date, most of the existing studies on human isolation and confinement are naturalistic observational studies, such as the SEALAB project (Radloff & Helmreich, 1968), or the South African National Research station (SANAE) in Antarctica (Vermeulen, 1977). Under these difficult circumstances of a physically and socially restricted environment, participants reported high indices of severe hostility, depression, insomnia, and anxiety. Studies on prisoner populations suggest that social factors can significantly boost prisoner’s well‐being, which can otherwise be disrupted by jail isolation (Kyprianides & Easterbrook, 2020). The conditions in which participants of the aforementioned studies were put are far from what people are experiencing now. Although being relatively isolated from the outside world, many of us remain within the comfort of our own homes, usually surrounded by friends or relatives. However, there are also people who are getting through this time alone. Living alone has been previously linked to higher indices of depression and anxiety, and other common mental disorders (Jacob et al., 2019). Therefore, those who live alone might experience more stress due to the COVID‐19 situation than those living with people who may be a source of potential support (Cohen & Wills, 1985). On the other hand, being around others might be stress‐enhancing. Specifically, the company of others can be detrimental to mental health when one is subjected to overcrowding—or, in the current situation—residential or household crowding, highly prevalent in developing countries (Epstein, 1981). Studies suggest that this type of chronic stress is often accompanied by a lack of privacy, a higher number of unwanted social interactions, possible deterioration of relationships with the family or flatmates, and so forth (Fuller et al., 1996). Considering the above, we hypothesise that the relationship between the number of people one is stuck with during isolation is somewhat U‐shaped—those who live alone and those who are subjected to overcrowding experience the highest levels of stress.
Intimate Relationships
In general, married individuals are happier (Lee & Ono, 2012), live longer and healthier lives (Kiecolt‐Glaser & Newton, 2001), and are at a lower risk of committing suicide (Waite & Gallagher, 2001). A recent study by Chin et al. (2017) supported previous conclusions at a physiological level. Interestingly, the authors investigated the cortisol levels (a hormone that is often associated with stress) of married and single persons, and found that married individuals had lower cortisol levels than unmarried and previously married people, which would suggest that they experience lower levels of stress. Similar patterns have been observed among persons in intimate relationships (not necessarily married), who exhibited lower levels of stress compared with single persons when stress was assessed by blood pressure (Sisca, 1985), self‐reported measures (Hudson & O'Regan, 1994), or a broad range of well‐being indices (Sorokowski et al., 2019).
Despite numerous advantages, being in a romantic relationship also carries a number of risks, for instance, contagion of negative emotions (Roberts & Levenson, 2001). Moreover, when couples encounter difficulties, such as financial hardship, or a lack of support from the spouse, they may experience a pronounced decrease in marital satisfaction, which can lead to an increase in stress levels (Archuleta et al., 2011).
On the other hand, intimate relationships allow for dyadic coping (for a review, see Bodenmann, 2005), and often serve as a buffer against difficult situations (Gottlieb & Wagner, 1991), which may in turn translate into lower levels of stress (Chin et al., 2017). Here, we intend to investigate whether marital status is a significant predictor of stress levels during quarantine or isolation on a large dataset from numerous countries around the world. We hypothesise that during the difficult times of relative isolation, being in an intimate relationship (i.e. marriage or cohabitation) is more beneficial in terms of experienced stress levels than being single.
Number of Children
Also of interest is how having and living with children relates to the experiences of adults during the period of relative isolation. Having children is quite a positive experience, as it has been generally related to greater life satisfaction, especially amongst married couples in contrast to individuals who are separated, cohabiting or never‐married singles (Angeles, 2010). Nevertheless, having children may also foster elevated stress levels. Parental stress is the result of the gap between the challenges of being a parent and the parent’s perceived ability to cope with those challenges (Abidin, 1995). This kind of stress may be even more pronounced during the COVID‐19 outbreak, as most parents now must homeschool their children. Parental stress has been associated with numerous negative feelings and thoughts, such as feeling overwhelmed and dissatisfied, and also perceiving the child as difficult (Haskett et al., 2006). For instance, a study on parents from 94 countries showed that parenthood is negatively linked to well‐being (Stanca, 2012).
As outbreaks make individuals vulnerable to depression and diminished social support (Stein et al., 2005), the current COVID‐19 pandemic may compromise parenting and childcare practices. Concerns over a child younger than 16 years getting COVID‐19 have been reported as very common among parents in China (Wang et al., 2020). During the equine influenza outbreak in Australia, Taylor et al. (2008) found that individuals having one child had a 1.2 times higher risk of experiencing distress than those with no children. Therefore, we hypothesise that having children at home may be associated with higher levels of stress in adults.
Gender
Gender seems to be related to well‐being (Mroczek & Kolarz, 1998) and stress levels (Taylor et al., 2008). Previous studies have found that women report greater sadness, anxiety, and stress than men (Bergdahl & Bergdahl, 2002; Gao et al, 2019).
Concerning quarantine, evidence from the first outbreak of equine influenza in Australia showed no gender differences on the level of psychological distress (Taylor et al., 2008). However, a recent study by Limcaoco et al. (2020) in 25 countries on susceptibility to stress during the COVID‐19 situation indicated that women report greater levels of stress. Similar gender differences for stress, anxiety, and depression symptoms were found by Wang et al. (2020) in a Chinese sample during the initial stage of the COVID‐19 outbreak, although only a minority of the participants reported being confined. In sum, while the evidence suggests that women in normal circumstances experience more stress, support for the link between gender and stress under quarantine is inconclusive.
Age
Generally, stress levels tend to decrease with age and although older adults report poorer perceived health, they have lower stress levels and higher well‐being than young adults (Archer et al., 2015). Bergdahl and Bergdahl (2002) found that self‐reported stress levels increase from the age of 20 to a peak in the 40s, and then decrease to the lowest level in the 60s. Stone et al. (2010) argued that well‐being increases after 50 years of age, proposing an inverted U‐shaped relationship between age and well‐being (but see Frackowiak et al., 2020). Studies generally support the notion that older people are less affected by stressors than younger people (Feizi et al., 2012).
Few studies have investigated the relationship between age and stress levels during quarantine. The objective consequences of being infected by SARS‐CoV‐2 are higher for the older population (World Health Organization, 2020a, 2020b). However, while the perceptions of a highly transmitted illness threat among the elderly may prompt stress, the empirical evidence is mixed. For instance, Taylor et al. (2008) found that the negative psychological impact of quarantine is more pronounced among younger people. Investigating psychosocial responses towards a national outbreak of SARS in Singapore, Sim et al. (2010) found that psychiatric morbidities are associated with younger age and higher posttraumatic stress symptoms. Most recently, an online survey in the early stages of the COVID‐19 outbreak in China found no association between age and stress (Wang et al., 2020).
As such, although older adults face higher risks of severe disease and death due to COVID‐19 (WHO, 2020b), the evidence suggests that older people are less stressed and less affected by psychological consequences of quarantine and social isolation, while younger ones exhibit the highest levels of stress.
Individualism–Collectivism Dimension
The role of culture has been widely studied for decades, with researchers debating how cultural factors may act as a buffer to the environmental stressors or, on the contrary, exacerbate stress levels (Dar, 2017). One of the main focuses of research in this area, the individualism–collectivism dimension (Hofstede, 2001), has been linked to stress processing (Chun et al., 2006). During the current quarantine, people have been forced to renounce their personal enjoyment (e.g. sports, concerts, shopping, travel, social gatherings) for the sake of group needs. Since individualistic, rather than collectivistic, cultures put a higher value on pleasure and hedonism (Schwartz, 2009), it might be reasonable to think that the emotional cost of this quarantine period would be greater in individualistic cultures. In fact, collectivistic (vs. individualistic) cultures put more emphasis on group harmony over personal interests and enjoyment (Triandis et al., 1990). Moreover, Oarga et al. (2015) found that helping behaviours had a stronger association with life satisfaction in countries where helping others constituted a social norm. In this sense, people from collective cultures would be likely to focus on caring for others, thereby alleviating the negative psychological consequences of quarantine. Overall, under the prevailing quarantine, the level of stress is expected to be higher for individualistic cultures compared to the collectivistic ones.
Aims and Hypotheses
The present study is one of the first to test how the number of persons an individual is staying with in isolation, along with age, gender, marital status, the number of children, and culture (i.e. individualism vs. collectivism) is associated with experienced stress levels in a large sample from 27 countries and areas. Furthermore, we want to investigate the link between stress levels and educational background and the severity of the COVID‐19 situation in each country or area. Based on the literature review, our pre‐registered hypotheses (see https://osf.io/xf4mj) were as follows:
H1: Individuals living alone and those who are subjected to overcrowding while in isolation experience the highest levels of perceived stress (compared with individuals living with others).
H2: Married and cohabiting persons experience lower levels of stress compared with persons that are single.
H3: Individuals with children would report increased stress levels during the COVID‐19 pandemic compared to people living alone or with adults. Levels of perceived stress would increase as the number of children at home increases.
H4: Younger people experience more stress than older people.
H5: People in collectivistic cultures experience less stress than those in individualistic cultures.
H6: Women experience higher levels of stress compared with men (not pre‐registered hypothesis).
METHODS
Participants
Participants were volunteers, recruited online between 30 March and 6 April 2020, using a snowballing technique. The initial sample included 54,245 Internet users from 27 countries and areas who participated in the COVIDiSTRESS study. However, based on the reliability analysis, we decided to exclude one country (Kosovo). The final sample consisted of 53,524 participants from 26 countries and areas (see the Results section for more details). Participation was not remunerated. The numbers for each country with their respective individualism scores (based on the Hofstede index; Hofstede, 2001; Hofstede et al., 2010), along with detailed descriptions of all participants from each country are presented in the Supplementary Material, Table S1 (https://osf.io/cznr8/). As stated in the pre‐registration, we only included countries and areas with N > 300 participants.
This sample consisted of 34,475 women (64.6%), 18,288 men (34.2%) and 645 people who responded with “other/would rather not say” (1.2%). The age in the sample ranged between 18 and 110 years (M = 38.17, SD = 13.79). The sample included 26,429 married or cohabiting participants (49.4%), 20,465 singles (38.2%), 3,626 divorced or widowed (6.8%), and 3,004 people who would rather not answer (5.6%); 629 individuals did not attain any education (1.2%), 709 completed 6 years of school (1.3%), 741 completed 9 years of school (1.4%), 5,133 completed 12 years of school (9.6%), 11,639 completed some college, short continuing education or equivalent (21.8%), 28,983 attained a college degree, bachelor, or masters (54.3%), and 5,497 obtained a PhD/doctorate (10.3%); 24,367 participants indicated that they are isolated (45.8%), 62—isolated in medical facility or similar location (0.1%), 25,867—life carries on with minor changes (48.7%), and 2,862—life carries on as usual (5.4%) (see Supplementary Material for participants’ detailed descriptives across countries and areas).
Measures and Procedure
Both the collaborating researchers of the COVIDiSTRESS project and the participants shared the invitation to participate in the study through various online channels (e.g. social media and private communication). The questionnaire was administered via Qualtrics. Information on all the components of the survey can be found in the COVIDiSTRESS Global Survey Research Network (2020), while in this paper, we report only the measures relevant to the analyses conducted.
The levels of perceived stress were measured with the widely used Perceived Stress Scale (PSS; Cohen et al., 1983; Cohen & Williamson, 1988). It includes 10 items (four reverse‐scored) concerning various subjective feelings related to problems, behaviours, and ways of coping, on a 5‐point scale from 0 (never) to 4 (very often). Cronbach’s alphas for each of the language versions of the scale are presented in the Supplementary Material (https://osf.io/cznr8/).
Statistical Analyses
A detailed description of the cleaning process of the database can be found in the Supplementary Material. Furthermore, we transformed the continuous variables of interest that violated the normal distribution (i.e. the number of adults, dependants, children, and a measure of COVID‐19 severity), by square rooting them.1 As a robustness check, in a separate model, we have also included these variables in a raw form: the results remained almost the same.
We began our analyses by testing the cross‐cultural equivalence of the PSS‐10. To this end, in a multi‐group factor analysis in R (R Core Team, 2014), using the following packages: lavaan (Rosseel, 2012), and semTools (Jorgensen et al., 2019), we compared the models assuming a two‐factor structure (positive and negative, with the latter consisting of reversed items; Roberti et al., 2006) across 26 countries and areas (configural invariance), with a model with factor loadings and latent correlations constrained to be equal (metric invariance), and items' intercepts to be the same in all groups (scalar invariance). When evaluating the model fit, we relied on the usually applied criteria (Hu & Bentler, 1999), in which a comparative fit index (CFI) and Tucker Lewis Index (TLI) above .90 indicate adequate fit, whereas a root‐mean‐square error of approximation (RMSEA) below .08 and a standardised root‐mean‐square residual (SRMR) below .06 indicates no misfit. When evaluating measurement equivalence, we compared the configural invariance model with the metric invariance model, and then the metric invariance model with the scalar invariance model. As these models were characterised by a growing complexity (each subsequent model was nested within the previous one), while assessing models’ superiority we relied on the cut‐off criteria recommended for testing measurement invariance: a change of CFI (∆CFI) less than .01 (∆CFI < .01), a change of RMSEA of less than .015 (∆RMSEA < .015), and a change of SRMR less than .01 (∆SRMR < .01) would indicate that the two models compared do not differ in terms of model fit (Chen et al., 2008; Cheung & Rensvold, 2002).
After computing Cronbach’s alpha for the PSS‐10 for all the countries and areas separately and jointly, we tested multilevel regression models, using Jamovi (Version 1.2.16.). In these models, participants were nested within countries, and the number of adults, children, marital status, education level, and gender were group‐mean‐standardised. In the first model, we regressed a composite score of perceived stress levels on the number of adults and the number of children, as well as participants’ gender, age, marital status, education level, COVID‐19 severity (the number of confirmed cases at the time of data collection (World Health Organization, 2020a, 2020b), divided by the number of the country’s citizens (Worldometer, 2020; Level‐2 variable), and culture (i.e. collectivistic vs. individualistic: Level‐2 variable). The first model we created included a random intercept of stress on a country level. The second model differed, as we additionally estimated random slopes of the number of children. The third model tested for potential curvilinear links by adding two variables in a squared form (i.e. the number of adults and children). The final, fourth model included random intercept for the stress level as well as all random slopes for age, gender, and marital status.
RESULTS
The PSS‐10 was found to be highly reliable (overall Cronbach’s alpha: .869), with the exception of Kosovo (Cronbach’s alpha: .688). Confirmatory factor analysis indicated that the overall model fit the data well in the case of configural invariance (CFI = .996, TLI = .995, SRMR = .028, RMSEA = .028), but not metric invariance (CFI = .986, TLI = .985, SRMR = .044, RMSEA = .046). Based on the poor reliability of Kosovo, we decided to exclude data from this country and re‐estimated the multi‐group CFA model. After excluding data from Kosovo, the model slightly improved: configural (CFI = .996, TLI = .995, SRMR = .028, RMSEA = .028), metric (CFI = .987, TLI = .996, SRMR = .043, RMSEA = .044), and scalar invariance (CFI = .950, TLI = .955, SRMR = .073, RMSEA = .081). However, as the scalar invariance was not reached (a change in fit model indices was above recommended criteria, i.e. ΔCFI and ΔSRMR> .01), in the next step, we tested for a partial scalar invariance (Byrne et al., 1989; Steenkamp & Baumgartner, 1998). The final model fit the data well, and partial scalar invariance was reached (CFI = .983, TLI = .983, SRMR = .047, RMSEA = .050). Table 1 depicts mean stress levels across countries, and Table 2 presents correlations across variables of interest. The mean stress level in the present study in, for instance, the USA (M = 17.50) was significantly higher than the mean stress level (M = 13.19) indicated by the PSS‐10 norms (Cohen et al., 1994; 13.19), t(3896) = 19.45, p < .001, d = 0.63.
TABLE 1 Mean Levels of Perceived Stress Across Countries and Areas
Country Number of participants Mean stress level (SD) Country Number of participants Mean stress level (SD)
Argentina 2286 17.53 (7.38) Indonesia 938 18.00 (5.75)
Belgium 339 16.38 (7.27) Italy 525 16.04 (6.89)
Bosnia and Herzegovina 660 18.95 (6.73) Japan 3190 20.08 (5.74)
Bulgaria 2413 18.78 (7.10) Lithuania 4450 15.32 (6.88)
Canada 307 17.81 (6.66) Mexico 5781 17.33 (7.32)
Croatia 900 19.55 (6.75) Poland 2094 20.19 (7.28)
Czech Republic 461 17.54 (6.95) Portugal 425 18.81 (7.25)
Denmark 5200 14.22 (7.21) Spain 363 16.80 (7.19)
Finland 7464 14.85 (7.50) Switzerland 719 13.93 (6.66)
France 9395 15.71 (7.38) Taiwan 869 15.97 (6.85)
Germany 616 16.52 (6.87) Turkey 683 21.71 (6.67)
Greece 304 17.43 (6.41) United Kingdom 886 17.53 (7.50)
Hungary 745 17.60 (6.06) United States 1511 17.50 (7.36)
Note
Mean stress scores on 10 items on a 5‐point Likert scale, ranging from 0 – almost never, to 4 – always.
John Wiley & Sons, Ltd This article is being made freely available through PubMed Central as part of the COVID-19 public health emergency response. It can be used for unrestricted research re-use and analysis in any form or by any means with acknowledgement of the original source, for the duration of the public health emergency.
TABLE 2 Correlation Matrix on Main Variables of Interest
1 2 3 4 5 6 7
1. Perceived stress —
2. Gender .09*** —
3. Age −.24*** .04*** —
4. Adults a .06*** ‐.04*** ‐.14*** —
5. Children b .04*** .06*** ‐.03*** .14*** —
6. People c .07*** .00 ‐.13*** .90*** .56*** —
7. Dependents d −.02*** .06*** .20*** .08*** .53*** .31*** —
8. Education e −.04*** −.03*** ‐.01 ‐.01** .03*** .00 .03***
Note:
aAdults–the number of adults staying with the participants during the relative isolation. bChildren–the number of children staying with the participants during the relative isolation. cPeople–the number of people staying with the participants during the relative isolation. dDependants–the number of dependants staying with the participants during the relative isolation. eEducation–education status, with a higher number representing a higher completed education level. *p < .05; **p < .01; ***p < .001.
John Wiley & Sons, Ltd This article is being made freely available through PubMed Central as part of the COVID-19 public health emergency response. It can be used for unrestricted research re-use and analysis in any form or by any means with acknowledgement of the original source, for the duration of the public health emergency. Having established configural, metric, and partly scalar invariance, we proceeded with multilevel models to examine whether the number of people (adults and children) our participants were staying with during the pandemic was associated with their perceived stress level, controlling for gender, age, education level, marital status, scores on the individualism measure, and COVID‐19 severity. Taking into consideration the four models we described in the Statistical Analyses section, based on the Bayesian Information Criterion, the fourth model (with random effects for both slopes and intercepts) showed a superior fit when compared to the others (i.e. ΔBIC > 10; Raftery, 1999; Δ between the first and the fourth model BIC = 432, the second and fourth ΔBIC = 431, the third and fourth ΔBIC = 450). The model that included the squared number of adults and children surrounding the participants in their isolation had substantially worse fit than the second model (ΔBIC = −19); thus, we infer that there is a linear rather than a curvilinear relationship between the number of adults and children, and stress.
As illustrated in Table 3, younger people perceived higher levels of stress compared with older people (β = −0.076, SE = 0.016, p < .001); the number of children an individual was staying with was positively related to perceived stress levels (β = 0.021, SE = 0.003, p < .001); women were more stressed than men (β = 0.065, SE = 0.012, p < .001); married or cohabiting people were less stressed than singles (β = −0.132, SE = 0.048, p < .01); education was negatively related to stress levels (β = −0.022, SE = 0.003, p < .001), meaning that the higher the completed education, the less stressed a participant was, and participants from countries and areas with higher COVID‐19 severity were more stressed (β = 0.090, SE = 0.044, p < .05). The countries’ and areas’ score on the individualism–collectivism continuum was not related to perceived stress (β = −0.087, SE = 0.045, p = .063), and similarly the number of adults the participant was staying with (β = 0.001, SE = 0.003, p = .806).
TABLE 3 Results of a Multilevel Model, with Participants Nested within Countries and Areas, and Random Effects for Both Intercepts and Slopes for Age, Gender, and Marital Status
Fixed effects β SE p
Culturea −0.087 0.045 .063
Age −0.076 0.016 <.001
Adultsb 0.000 0.003 .806
Childrenc 0.021 0.003 <.001
Gender (Men vs. Women) 0.065 0.012 <.001
Educationd −0.022 0.003 <.001
COVID‐19 severitye 0.090 0.044 .050
Marital status (Single vs. Married)f −0.132 0.048 .006
Random effects g SD LRT p
Age 0.077 816.83 <.001
Gender (Men vs. Women) 0.054 366.59 <.001
Marital status (Married vs. Single)f 0.238 368.13 <.001
Note
aCulture–Hofstede’s score of a country's individualism (i.e. the higher the number, the more individualistic the country). bAdults–the number of adults staying with the participants during the relative isolation. cChildren–the number of children staying with the participants during the relative isolation. dEducation–education status, with a higher number representing a higher completed education level: 6–PhD/Doctorate; 5–College degree, bachelor, master; 4–Some College, short continuing education or equivalent; 3–Up to 12 years of school; 2–Up to 9 years of school; 1–Up to 6 years of school; 0–None. eCOVID‐19 severity–the number of confirmed cases divided by the number of countries’ citizens. fMarital status–single individuals are coded as 0, and married or cohabiting individuals as 1. gRandom effects–the likelihood ratio test for each random effect, representing variability of stress levels and the association between stress levels and other variables of interest between countries. For the intercept: SD = 0.261, ICC = 0.130. Significant results are bolded.
John Wiley & Sons, Ltd This article is being made freely available through PubMed Central as part of the COVID-19 public health emergency response. It can be used for unrestricted research re-use and analysis in any form or by any means with acknowledgement of the original source, for the duration of the public health emergency.
DISCUSSION
The results of the present study, based on analyses of data from 53,524 respondents from 26 countries and areas, provide evidence that higher levels of stress are reported by women, single people, people of younger age, people staying with more children, the less educated, and from countries and areas with a more severe COVID‐19 situation. The level of the country's individualism and the number of adults the individual was staying with were not related to stress levels. What is worth highlighting is that the variable which was the most strongly associated with stress levels was marital status, while the standardised coefficients of other variables showed very small effect sizes. Moreover, the analysis of random effects revealed that countries and areas differed in both stress levels and the association between stress levels and marital status, gender, and age. Such differences and their underlying mechanisms warrant further investigations.
H1. The Number of People Together—Do People Living Alone Experience the Highest Level of Stress? No
Even though almost half of our sample was stuck at home (46.1%), while 20.2 per cent lived alone, and 77.5 per cent stayed with one or more adults, we found very little evidence for the association between the number of adults and perceived stress levels (beyond marital status). The results of our study contradict previous findings, which suggested that either living alone or having too many people around can negatively affect one’s well‐being (Fuller et al., 1996). It is possible that living alone may be a preference for many people in our sample, and that they are able to provide for themselves and to pay the rent for a single apartment. Many of those who live alone still maintain high levels of connections with their non‐household relatives or friends (De Vaus & Qu, 2015). Previous studies have shown that people use social media to connect with others (Kowal et al., 2020), and these valuable (and, most importantly, intentional) connections and interactions are not hindered by the current situation, as they can be kept online.
H2. Marital Status—Do Single People Experience More Stress Than Married People? Yes
The results of our study provided support to our hypothesis that married (or cohabiting) individuals experience lower levels of stress than single individuals. These findings fall in line with previous studies highlighting the benefits of being in an intimate relationship (e.g. Braithwaite et al., 2010). Dyadic coping can make up for the potential shortcomings of being in a relationship, and overall, provide more benefits than harm (Merz et al., 2014). Such a protective role may be even more pronounced during difficult times, for instance during financial hardship (Helms et al., 2014). Within the COVID‐19 context, our findings are in line with those of Odriozola‐González et al. (2020), who provided evidence that single (as compared with married) individuals score higher on measures of anxiety, depression, and stress.
On the other hand, our results differ from Wang et al. (2020) and Tian et al. (2020), who found no differences related to marital status on perceived stress, and higher levels of distress among single (vs. married) Chinese citizens, respectively. However, the sample analysed by Wang et al. (2020) may have been underpowered to detect the effect of being in a relationship, as it had 1,210 participants, and tested numerous relationships between various variables. Moreover, Tian et al. (2020) did not control for other relevant factors that could have accounted for the differences (e.g. age, gender). Our study used a high‐powered sample from different countries and areas as well as multivariate analysis techniques to control for several confundings.
H3. Children—Do People with Children Experience More Stress? Yes
Our data confirm the hypothesis that individuals with children report increased stress levels during the COVID‐19 pandemic compared with people living alone or with adults. Moreover, the levels of perceived stress increased with the growing number of children at home. The presence of children during relative isolation or quarantine, probably due to increased home strain and need for childcare, might put demands on adults living with them. In addition, concerns over a child being sick with COVID‐19 have been previously shown to be associated with greater severity of stress and anxiety symptoms in parents (Wang et al., 2020). Having children has also been associated with decreased financial satisfaction (Stanca, 2012). Future studies should examine whether the ages or gender of children living at home moderate the relationship between the number of children and adults’ stress levels.
H4. Age—Do Younger People Experience More Stress than Older People? Yes
The results of our study provided evidence for a negative linear relationship between age and stress levels—younger people experience more stress than older people—contrasting with the existing literature suggesting an inverted U‐shaped relationship between age and stress (Stone et al., 2010). The findings of the present study are in contrast with a recent study of the COVID‐19 outbreak, which showed no association between age and stress levels in a Chinese sample (Wang et al., 2020). Rather, our results support the findings of Limcaoco (2020), who, in a sample of people in quarantine across 41 countries, found age to be significantly related to stress levels. Our evidence warns that we cannot neglect the mental health condition of the younger generations as they seem to be the most stressed during the COVID‐19 outbreak.
H5. Culture—Do People from Individualistic Cultures Experience More Stress? No
In contrast to our hypothesis that people from individualistic cultures might experience more stress than those from collectivistic cultures, we observed no differences in perceived stress levels between countries with varying levels of individualism. Collectivism demands that people feel the obligation and responsibility towards in‐group members. Therefore, due to living in traditional families (Persike & Seiffge‐Krenke, 2012) and feeling more guilt and remorse for seeking help (Kim et al., 2008), people from collectivistic cultures may feel more stressed over their financial burdens than people from individualistic cultures, whereas people from individualistic cultures may treat the current situation as a threat to their need for self‐expression and freedom (Kim & Sherman, 2007) which translates into equal levels of stress.
H6. Gender—Do Women Experience More Stress than Men? Yes
Our data provide evidence that women experience significantly higher stress levels than men during quarantine. These findings support the existing literature on the relationship between gender and stress levels (e.g. Bergdahl & Bergdahl, 2002; Stone et al., 2010), but are contradictory to recent studies on the COVID‐19 outbreak, which found gender to be unrelated to stress levels (Li et al., 2020; Limcaoco et al., 2020; Wang et al., 2020). However, conclusions from these studies ought to be drawn with caution, as they either focused exclusively on samples within one country (e.g. Wang et al., 2020) or included a small number of participants (e.g. Limcaoco et al., 2020). Women are found to be more family‐oriented and provide a supportive role (Tigani et al., 2011); however, during this highly unprecedented time of quarantine, they are overwhelmed by the need to support their families, and such a commitment is beyond their coping capabilities. Moreover, during the relative isolation, it is women who are more likely to not only take care of their work responsibilities, but also take care of the children and home (Minello, 2020).
Limitations
Although the present study contributes to a better understanding of the relationship between the presence of others and perceived stress during the COVID‐19 pandemic, it is not without limitations. First, in order to reach out to participants that are currently under lockdown, it was necessary to use self‐reports and to make use of the individuals’ internet access, although it might limit our findings to a population that is wealthier and has a higher educational background than the general population. Moreover, perceived stress level was tested using only a single self‐reported measure (10‐item Perceived Stress Scale), which warrants some caution. Future studies should also include other measures of stress to expand our knowledge of various aspects of well‐being during the pandemic.
Second, it would not have been possible to use an experimental design, so all our findings are correlative, and we cannot put forward causal relationships. Shortly before the lockdown, people might have chosen to move away from their families because they work in health care and did not want to infect their family members. People that lived on their own before the pandemic might have agreed to isolate with others to combat their loneliness or to help with care work in the family. Those examples are likely rare and might only mildly affect our findings. Even though we were not able to control for confounding variables, some pre‐existing factors, such as age and gender, may at least indicate a direction of effect.
Third, longitudinal effects of the pandemic were not measured. We present the data of the first week when the COVIDiSTRESS Global Survey went online, and participants were only allowed to participate once. It would be beneficial for future studies to measure the long‐term effects and the development of stress within a sample.
Fourth, the pandemic had been spreading in countries and areas at various rates during data collection. Therefore, people from different countries and different regions within a country possibly were under different levels of threat as well as different time spans of exposure. It is important for future studies to further investigate whether there are dose–response relationships between stress and the level of threat in the given countries.
CONCLUSIONS
The present study contributes to a better understanding of the relationship between the presence of others and perceived stress during the COVID‐19 lockdown. Moreover, our results add to the limited literature on people during the pandemic. These insights highlight the need to focus on neglected groups (i.e. women, younger, single people, with children, and with lower education status) both in research and in public health interventions, as they may be especially susceptible to experiencing elevated levels of stress, with their psychological well‐being being compromised.
Conflict of interest
The authors declare no conflict of interest.
Supporting information
Table S1. Descriptive statistics of gender, marital status, age, the number of adults and children under 12 individuals were staying with during the data collection, and country’s score on Hofstede individualism scale.
Table S2. Means, standard deviations and Cronbach’s alphas of the Perceived Stress Scale across countries.
Table S3. Results of a Multilevel Model, With Participants Nested Within Countries, and Random Effects for Both Intercepts and Slopes for age, gender, isolation, and marital status with variables in a raw form.
Table S4. Responses on the question: ‘What best describes your current situation?’, with respect to countries.
Table S5. The number of citizens across countries and the number of confirmed cases across days of data collection (with an early launch in Denmark – March 28 and 29).
Table S6. The number of participants with the given education level (the highest completed education) with respect to countries.
Figure S1. Three scatter plots, depicting relationships between perceived stress and the number of people (A), children (B), and adults (C) an individual was staying with during the relative isolation.
Click here for additional data file.
|
|
sec
|
INTRODUCTION
The COVID‐19 pandemic brought the world to a halt. Since early 2020, social life has changed for many people around the world. Government restrictions and new social norms led to a reduction in mobility (Google, 2020), avoidance of public transport, cancellation of the majority of large events such as concerts, festivals, religious and sports events, and the temporary closing down of meeting places such as cafes, restaurants, museums or theatres. The COVID‐19 outbreak and measures undertaken by almost all the countries around the world pose numerous threats to people's psychological well‐being; thus, we believe that focusing on our closest social environment is greatly needed. Both relative lack of social relations (Tay et al., 2013) and negative emotions (Huppert, 2009) strongly predict overall mortality and disease outcomes, and the threat of SARS‐CoV‐2 may increase anxiety levels (Karwowski et al., 2020) and prejudice toward other nationalities (Sorokowski et al., 2020), rendering any effort to understand stress in isolated or quarantined individuals during the COVID‐19 pandemic even more important. Moreover, to understand possible negative effects, it is important to unpick the pre‐existing factors that can predict the stress levels of people in quarantine or isolation.
Other People
One such factor is the presence (or absence) of other people. To date, most of the existing studies on human isolation and confinement are naturalistic observational studies, such as the SEALAB project (Radloff & Helmreich, 1968), or the South African National Research station (SANAE) in Antarctica (Vermeulen, 1977). Under these difficult circumstances of a physically and socially restricted environment, participants reported high indices of severe hostility, depression, insomnia, and anxiety. Studies on prisoner populations suggest that social factors can significantly boost prisoner’s well‐being, which can otherwise be disrupted by jail isolation (Kyprianides & Easterbrook, 2020). The conditions in which participants of the aforementioned studies were put are far from what people are experiencing now. Although being relatively isolated from the outside world, many of us remain within the comfort of our own homes, usually surrounded by friends or relatives. However, there are also people who are getting through this time alone. Living alone has been previously linked to higher indices of depression and anxiety, and other common mental disorders (Jacob et al., 2019). Therefore, those who live alone might experience more stress due to the COVID‐19 situation than those living with people who may be a source of potential support (Cohen & Wills, 1985). On the other hand, being around others might be stress‐enhancing. Specifically, the company of others can be detrimental to mental health when one is subjected to overcrowding—or, in the current situation—residential or household crowding, highly prevalent in developing countries (Epstein, 1981). Studies suggest that this type of chronic stress is often accompanied by a lack of privacy, a higher number of unwanted social interactions, possible deterioration of relationships with the family or flatmates, and so forth (Fuller et al., 1996). Considering the above, we hypothesise that the relationship between the number of people one is stuck with during isolation is somewhat U‐shaped—those who live alone and those who are subjected to overcrowding experience the highest levels of stress.
Intimate Relationships
In general, married individuals are happier (Lee & Ono, 2012), live longer and healthier lives (Kiecolt‐Glaser & Newton, 2001), and are at a lower risk of committing suicide (Waite & Gallagher, 2001). A recent study by Chin et al. (2017) supported previous conclusions at a physiological level. Interestingly, the authors investigated the cortisol levels (a hormone that is often associated with stress) of married and single persons, and found that married individuals had lower cortisol levels than unmarried and previously married people, which would suggest that they experience lower levels of stress. Similar patterns have been observed among persons in intimate relationships (not necessarily married), who exhibited lower levels of stress compared with single persons when stress was assessed by blood pressure (Sisca, 1985), self‐reported measures (Hudson & O'Regan, 1994), or a broad range of well‐being indices (Sorokowski et al., 2019).
Despite numerous advantages, being in a romantic relationship also carries a number of risks, for instance, contagion of negative emotions (Roberts & Levenson, 2001). Moreover, when couples encounter difficulties, such as financial hardship, or a lack of support from the spouse, they may experience a pronounced decrease in marital satisfaction, which can lead to an increase in stress levels (Archuleta et al., 2011).
On the other hand, intimate relationships allow for dyadic coping (for a review, see Bodenmann, 2005), and often serve as a buffer against difficult situations (Gottlieb & Wagner, 1991), which may in turn translate into lower levels of stress (Chin et al., 2017). Here, we intend to investigate whether marital status is a significant predictor of stress levels during quarantine or isolation on a large dataset from numerous countries around the world. We hypothesise that during the difficult times of relative isolation, being in an intimate relationship (i.e. marriage or cohabitation) is more beneficial in terms of experienced stress levels than being single.
Number of Children
Also of interest is how having and living with children relates to the experiences of adults during the period of relative isolation. Having children is quite a positive experience, as it has been generally related to greater life satisfaction, especially amongst married couples in contrast to individuals who are separated, cohabiting or never‐married singles (Angeles, 2010). Nevertheless, having children may also foster elevated stress levels. Parental stress is the result of the gap between the challenges of being a parent and the parent’s perceived ability to cope with those challenges (Abidin, 1995). This kind of stress may be even more pronounced during the COVID‐19 outbreak, as most parents now must homeschool their children. Parental stress has been associated with numerous negative feelings and thoughts, such as feeling overwhelmed and dissatisfied, and also perceiving the child as difficult (Haskett et al., 2006). For instance, a study on parents from 94 countries showed that parenthood is negatively linked to well‐being (Stanca, 2012).
As outbreaks make individuals vulnerable to depression and diminished social support (Stein et al., 2005), the current COVID‐19 pandemic may compromise parenting and childcare practices. Concerns over a child younger than 16 years getting COVID‐19 have been reported as very common among parents in China (Wang et al., 2020). During the equine influenza outbreak in Australia, Taylor et al. (2008) found that individuals having one child had a 1.2 times higher risk of experiencing distress than those with no children. Therefore, we hypothesise that having children at home may be associated with higher levels of stress in adults.
Gender
Gender seems to be related to well‐being (Mroczek & Kolarz, 1998) and stress levels (Taylor et al., 2008). Previous studies have found that women report greater sadness, anxiety, and stress than men (Bergdahl & Bergdahl, 2002; Gao et al, 2019).
Concerning quarantine, evidence from the first outbreak of equine influenza in Australia showed no gender differences on the level of psychological distress (Taylor et al., 2008). However, a recent study by Limcaoco et al. (2020) in 25 countries on susceptibility to stress during the COVID‐19 situation indicated that women report greater levels of stress. Similar gender differences for stress, anxiety, and depression symptoms were found by Wang et al. (2020) in a Chinese sample during the initial stage of the COVID‐19 outbreak, although only a minority of the participants reported being confined. In sum, while the evidence suggests that women in normal circumstances experience more stress, support for the link between gender and stress under quarantine is inconclusive.
Age
Generally, stress levels tend to decrease with age and although older adults report poorer perceived health, they have lower stress levels and higher well‐being than young adults (Archer et al., 2015). Bergdahl and Bergdahl (2002) found that self‐reported stress levels increase from the age of 20 to a peak in the 40s, and then decrease to the lowest level in the 60s. Stone et al. (2010) argued that well‐being increases after 50 years of age, proposing an inverted U‐shaped relationship between age and well‐being (but see Frackowiak et al., 2020). Studies generally support the notion that older people are less affected by stressors than younger people (Feizi et al., 2012).
Few studies have investigated the relationship between age and stress levels during quarantine. The objective consequences of being infected by SARS‐CoV‐2 are higher for the older population (World Health Organization, 2020a, 2020b). However, while the perceptions of a highly transmitted illness threat among the elderly may prompt stress, the empirical evidence is mixed. For instance, Taylor et al. (2008) found that the negative psychological impact of quarantine is more pronounced among younger people. Investigating psychosocial responses towards a national outbreak of SARS in Singapore, Sim et al. (2010) found that psychiatric morbidities are associated with younger age and higher posttraumatic stress symptoms. Most recently, an online survey in the early stages of the COVID‐19 outbreak in China found no association between age and stress (Wang et al., 2020).
As such, although older adults face higher risks of severe disease and death due to COVID‐19 (WHO, 2020b), the evidence suggests that older people are less stressed and less affected by psychological consequences of quarantine and social isolation, while younger ones exhibit the highest levels of stress.
Individualism–Collectivism Dimension
The role of culture has been widely studied for decades, with researchers debating how cultural factors may act as a buffer to the environmental stressors or, on the contrary, exacerbate stress levels (Dar, 2017). One of the main focuses of research in this area, the individualism–collectivism dimension (Hofstede, 2001), has been linked to stress processing (Chun et al., 2006). During the current quarantine, people have been forced to renounce their personal enjoyment (e.g. sports, concerts, shopping, travel, social gatherings) for the sake of group needs. Since individualistic, rather than collectivistic, cultures put a higher value on pleasure and hedonism (Schwartz, 2009), it might be reasonable to think that the emotional cost of this quarantine period would be greater in individualistic cultures. In fact, collectivistic (vs. individualistic) cultures put more emphasis on group harmony over personal interests and enjoyment (Triandis et al., 1990). Moreover, Oarga et al. (2015) found that helping behaviours had a stronger association with life satisfaction in countries where helping others constituted a social norm. In this sense, people from collective cultures would be likely to focus on caring for others, thereby alleviating the negative psychological consequences of quarantine. Overall, under the prevailing quarantine, the level of stress is expected to be higher for individualistic cultures compared to the collectivistic ones.
Aims and Hypotheses
The present study is one of the first to test how the number of persons an individual is staying with in isolation, along with age, gender, marital status, the number of children, and culture (i.e. individualism vs. collectivism) is associated with experienced stress levels in a large sample from 27 countries and areas. Furthermore, we want to investigate the link between stress levels and educational background and the severity of the COVID‐19 situation in each country or area. Based on the literature review, our pre‐registered hypotheses (see https://osf.io/xf4mj) were as follows:
H1: Individuals living alone and those who are subjected to overcrowding while in isolation experience the highest levels of perceived stress (compared with individuals living with others).
H2: Married and cohabiting persons experience lower levels of stress compared with persons that are single.
H3: Individuals with children would report increased stress levels during the COVID‐19 pandemic compared to people living alone or with adults. Levels of perceived stress would increase as the number of children at home increases.
H4: Younger people experience more stress than older people.
H5: People in collectivistic cultures experience less stress than those in individualistic cultures.
H6: Women experience higher levels of stress compared with men (not pre‐registered hypothesis).
|
|
title
|
INTRODUCTION
|
|
p
|
The COVID‐19 pandemic brought the world to a halt. Since early 2020, social life has changed for many people around the world. Government restrictions and new social norms led to a reduction in mobility (Google, 2020), avoidance of public transport, cancellation of the majority of large events such as concerts, festivals, religious and sports events, and the temporary closing down of meeting places such as cafes, restaurants, museums or theatres. The COVID‐19 outbreak and measures undertaken by almost all the countries around the world pose numerous threats to people's psychological well‐being; thus, we believe that focusing on our closest social environment is greatly needed. Both relative lack of social relations (Tay et al., 2013) and negative emotions (Huppert, 2009) strongly predict overall mortality and disease outcomes, and the threat of SARS‐CoV‐2 may increase anxiety levels (Karwowski et al., 2020) and prejudice toward other nationalities (Sorokowski et al., 2020), rendering any effort to understand stress in isolated or quarantined individuals during the COVID‐19 pandemic even more important. Moreover, to understand possible negative effects, it is important to unpick the pre‐existing factors that can predict the stress levels of people in quarantine or isolation.
|
|
sec
|
Other People
One such factor is the presence (or absence) of other people. To date, most of the existing studies on human isolation and confinement are naturalistic observational studies, such as the SEALAB project (Radloff & Helmreich, 1968), or the South African National Research station (SANAE) in Antarctica (Vermeulen, 1977). Under these difficult circumstances of a physically and socially restricted environment, participants reported high indices of severe hostility, depression, insomnia, and anxiety. Studies on prisoner populations suggest that social factors can significantly boost prisoner’s well‐being, which can otherwise be disrupted by jail isolation (Kyprianides & Easterbrook, 2020). The conditions in which participants of the aforementioned studies were put are far from what people are experiencing now. Although being relatively isolated from the outside world, many of us remain within the comfort of our own homes, usually surrounded by friends or relatives. However, there are also people who are getting through this time alone. Living alone has been previously linked to higher indices of depression and anxiety, and other common mental disorders (Jacob et al., 2019). Therefore, those who live alone might experience more stress due to the COVID‐19 situation than those living with people who may be a source of potential support (Cohen & Wills, 1985). On the other hand, being around others might be stress‐enhancing. Specifically, the company of others can be detrimental to mental health when one is subjected to overcrowding—or, in the current situation—residential or household crowding, highly prevalent in developing countries (Epstein, 1981). Studies suggest that this type of chronic stress is often accompanied by a lack of privacy, a higher number of unwanted social interactions, possible deterioration of relationships with the family or flatmates, and so forth (Fuller et al., 1996). Considering the above, we hypothesise that the relationship between the number of people one is stuck with during isolation is somewhat U‐shaped—those who live alone and those who are subjected to overcrowding experience the highest levels of stress.
|
|
title
|
Other People
|
|
p
|
One such factor is the presence (or absence) of other people. To date, most of the existing studies on human isolation and confinement are naturalistic observational studies, such as the SEALAB project (Radloff & Helmreich, 1968), or the South African National Research station (SANAE) in Antarctica (Vermeulen, 1977). Under these difficult circumstances of a physically and socially restricted environment, participants reported high indices of severe hostility, depression, insomnia, and anxiety. Studies on prisoner populations suggest that social factors can significantly boost prisoner’s well‐being, which can otherwise be disrupted by jail isolation (Kyprianides & Easterbrook, 2020). The conditions in which participants of the aforementioned studies were put are far from what people are experiencing now. Although being relatively isolated from the outside world, many of us remain within the comfort of our own homes, usually surrounded by friends or relatives. However, there are also people who are getting through this time alone. Living alone has been previously linked to higher indices of depression and anxiety, and other common mental disorders (Jacob et al., 2019). Therefore, those who live alone might experience more stress due to the COVID‐19 situation than those living with people who may be a source of potential support (Cohen & Wills, 1985). On the other hand, being around others might be stress‐enhancing. Specifically, the company of others can be detrimental to mental health when one is subjected to overcrowding—or, in the current situation—residential or household crowding, highly prevalent in developing countries (Epstein, 1981). Studies suggest that this type of chronic stress is often accompanied by a lack of privacy, a higher number of unwanted social interactions, possible deterioration of relationships with the family or flatmates, and so forth (Fuller et al., 1996). Considering the above, we hypothesise that the relationship between the number of people one is stuck with during isolation is somewhat U‐shaped—those who live alone and those who are subjected to overcrowding experience the highest levels of stress.
|
|
sec
|
Intimate Relationships
In general, married individuals are happier (Lee & Ono, 2012), live longer and healthier lives (Kiecolt‐Glaser & Newton, 2001), and are at a lower risk of committing suicide (Waite & Gallagher, 2001). A recent study by Chin et al. (2017) supported previous conclusions at a physiological level. Interestingly, the authors investigated the cortisol levels (a hormone that is often associated with stress) of married and single persons, and found that married individuals had lower cortisol levels than unmarried and previously married people, which would suggest that they experience lower levels of stress. Similar patterns have been observed among persons in intimate relationships (not necessarily married), who exhibited lower levels of stress compared with single persons when stress was assessed by blood pressure (Sisca, 1985), self‐reported measures (Hudson & O'Regan, 1994), or a broad range of well‐being indices (Sorokowski et al., 2019).
Despite numerous advantages, being in a romantic relationship also carries a number of risks, for instance, contagion of negative emotions (Roberts & Levenson, 2001). Moreover, when couples encounter difficulties, such as financial hardship, or a lack of support from the spouse, they may experience a pronounced decrease in marital satisfaction, which can lead to an increase in stress levels (Archuleta et al., 2011).
On the other hand, intimate relationships allow for dyadic coping (for a review, see Bodenmann, 2005), and often serve as a buffer against difficult situations (Gottlieb & Wagner, 1991), which may in turn translate into lower levels of stress (Chin et al., 2017). Here, we intend to investigate whether marital status is a significant predictor of stress levels during quarantine or isolation on a large dataset from numerous countries around the world. We hypothesise that during the difficult times of relative isolation, being in an intimate relationship (i.e. marriage or cohabitation) is more beneficial in terms of experienced stress levels than being single.
|
|
title
|
Intimate Relationships
|
|
p
|
In general, married individuals are happier (Lee & Ono, 2012), live longer and healthier lives (Kiecolt‐Glaser & Newton, 2001), and are at a lower risk of committing suicide (Waite & Gallagher, 2001). A recent study by Chin et al. (2017) supported previous conclusions at a physiological level. Interestingly, the authors investigated the cortisol levels (a hormone that is often associated with stress) of married and single persons, and found that married individuals had lower cortisol levels than unmarried and previously married people, which would suggest that they experience lower levels of stress. Similar patterns have been observed among persons in intimate relationships (not necessarily married), who exhibited lower levels of stress compared with single persons when stress was assessed by blood pressure (Sisca, 1985), self‐reported measures (Hudson & O'Regan, 1994), or a broad range of well‐being indices (Sorokowski et al., 2019).
|
|
p
|
Despite numerous advantages, being in a romantic relationship also carries a number of risks, for instance, contagion of negative emotions (Roberts & Levenson, 2001). Moreover, when couples encounter difficulties, such as financial hardship, or a lack of support from the spouse, they may experience a pronounced decrease in marital satisfaction, which can lead to an increase in stress levels (Archuleta et al., 2011).
|
|
p
|
On the other hand, intimate relationships allow for dyadic coping (for a review, see Bodenmann, 2005), and often serve as a buffer against difficult situations (Gottlieb & Wagner, 1991), which may in turn translate into lower levels of stress (Chin et al., 2017). Here, we intend to investigate whether marital status is a significant predictor of stress levels during quarantine or isolation on a large dataset from numerous countries around the world. We hypothesise that during the difficult times of relative isolation, being in an intimate relationship (i.e. marriage or cohabitation) is more beneficial in terms of experienced stress levels than being single.
|
|
sec
|
Number of Children
Also of interest is how having and living with children relates to the experiences of adults during the period of relative isolation. Having children is quite a positive experience, as it has been generally related to greater life satisfaction, especially amongst married couples in contrast to individuals who are separated, cohabiting or never‐married singles (Angeles, 2010). Nevertheless, having children may also foster elevated stress levels. Parental stress is the result of the gap between the challenges of being a parent and the parent’s perceived ability to cope with those challenges (Abidin, 1995). This kind of stress may be even more pronounced during the COVID‐19 outbreak, as most parents now must homeschool their children. Parental stress has been associated with numerous negative feelings and thoughts, such as feeling overwhelmed and dissatisfied, and also perceiving the child as difficult (Haskett et al., 2006). For instance, a study on parents from 94 countries showed that parenthood is negatively linked to well‐being (Stanca, 2012).
As outbreaks make individuals vulnerable to depression and diminished social support (Stein et al., 2005), the current COVID‐19 pandemic may compromise parenting and childcare practices. Concerns over a child younger than 16 years getting COVID‐19 have been reported as very common among parents in China (Wang et al., 2020). During the equine influenza outbreak in Australia, Taylor et al. (2008) found that individuals having one child had a 1.2 times higher risk of experiencing distress than those with no children. Therefore, we hypothesise that having children at home may be associated with higher levels of stress in adults.
|
|
title
|
Number of Children
|
|
p
|
Also of interest is how having and living with children relates to the experiences of adults during the period of relative isolation. Having children is quite a positive experience, as it has been generally related to greater life satisfaction, especially amongst married couples in contrast to individuals who are separated, cohabiting or never‐married singles (Angeles, 2010). Nevertheless, having children may also foster elevated stress levels. Parental stress is the result of the gap between the challenges of being a parent and the parent’s perceived ability to cope with those challenges (Abidin, 1995). This kind of stress may be even more pronounced during the COVID‐19 outbreak, as most parents now must homeschool their children. Parental stress has been associated with numerous negative feelings and thoughts, such as feeling overwhelmed and dissatisfied, and also perceiving the child as difficult (Haskett et al., 2006). For instance, a study on parents from 94 countries showed that parenthood is negatively linked to well‐being (Stanca, 2012).
|
|
p
|
As outbreaks make individuals vulnerable to depression and diminished social support (Stein et al., 2005), the current COVID‐19 pandemic may compromise parenting and childcare practices. Concerns over a child younger than 16 years getting COVID‐19 have been reported as very common among parents in China (Wang et al., 2020). During the equine influenza outbreak in Australia, Taylor et al. (2008) found that individuals having one child had a 1.2 times higher risk of experiencing distress than those with no children. Therefore, we hypothesise that having children at home may be associated with higher levels of stress in adults.
|
|
sec
|
Gender
Gender seems to be related to well‐being (Mroczek & Kolarz, 1998) and stress levels (Taylor et al., 2008). Previous studies have found that women report greater sadness, anxiety, and stress than men (Bergdahl & Bergdahl, 2002; Gao et al, 2019).
Concerning quarantine, evidence from the first outbreak of equine influenza in Australia showed no gender differences on the level of psychological distress (Taylor et al., 2008). However, a recent study by Limcaoco et al. (2020) in 25 countries on susceptibility to stress during the COVID‐19 situation indicated that women report greater levels of stress. Similar gender differences for stress, anxiety, and depression symptoms were found by Wang et al. (2020) in a Chinese sample during the initial stage of the COVID‐19 outbreak, although only a minority of the participants reported being confined. In sum, while the evidence suggests that women in normal circumstances experience more stress, support for the link between gender and stress under quarantine is inconclusive.
|
|
title
|
Gender
|
|
p
|
Gender seems to be related to well‐being (Mroczek & Kolarz, 1998) and stress levels (Taylor et al., 2008). Previous studies have found that women report greater sadness, anxiety, and stress than men (Bergdahl & Bergdahl, 2002; Gao et al, 2019).
|
|
p
|
Concerning quarantine, evidence from the first outbreak of equine influenza in Australia showed no gender differences on the level of psychological distress (Taylor et al., 2008). However, a recent study by Limcaoco et al. (2020) in 25 countries on susceptibility to stress during the COVID‐19 situation indicated that women report greater levels of stress. Similar gender differences for stress, anxiety, and depression symptoms were found by Wang et al. (2020) in a Chinese sample during the initial stage of the COVID‐19 outbreak, although only a minority of the participants reported being confined. In sum, while the evidence suggests that women in normal circumstances experience more stress, support for the link between gender and stress under quarantine is inconclusive.
|
|
sec
|
Age
Generally, stress levels tend to decrease with age and although older adults report poorer perceived health, they have lower stress levels and higher well‐being than young adults (Archer et al., 2015). Bergdahl and Bergdahl (2002) found that self‐reported stress levels increase from the age of 20 to a peak in the 40s, and then decrease to the lowest level in the 60s. Stone et al. (2010) argued that well‐being increases after 50 years of age, proposing an inverted U‐shaped relationship between age and well‐being (but see Frackowiak et al., 2020). Studies generally support the notion that older people are less affected by stressors than younger people (Feizi et al., 2012).
Few studies have investigated the relationship between age and stress levels during quarantine. The objective consequences of being infected by SARS‐CoV‐2 are higher for the older population (World Health Organization, 2020a, 2020b). However, while the perceptions of a highly transmitted illness threat among the elderly may prompt stress, the empirical evidence is mixed. For instance, Taylor et al. (2008) found that the negative psychological impact of quarantine is more pronounced among younger people. Investigating psychosocial responses towards a national outbreak of SARS in Singapore, Sim et al. (2010) found that psychiatric morbidities are associated with younger age and higher posttraumatic stress symptoms. Most recently, an online survey in the early stages of the COVID‐19 outbreak in China found no association between age and stress (Wang et al., 2020).
As such, although older adults face higher risks of severe disease and death due to COVID‐19 (WHO, 2020b), the evidence suggests that older people are less stressed and less affected by psychological consequences of quarantine and social isolation, while younger ones exhibit the highest levels of stress.
|
|
title
|
Age
|
|
p
|
Generally, stress levels tend to decrease with age and although older adults report poorer perceived health, they have lower stress levels and higher well‐being than young adults (Archer et al., 2015). Bergdahl and Bergdahl (2002) found that self‐reported stress levels increase from the age of 20 to a peak in the 40s, and then decrease to the lowest level in the 60s. Stone et al. (2010) argued that well‐being increases after 50 years of age, proposing an inverted U‐shaped relationship between age and well‐being (but see Frackowiak et al., 2020). Studies generally support the notion that older people are less affected by stressors than younger people (Feizi et al., 2012).
|
|
p
|
Few studies have investigated the relationship between age and stress levels during quarantine. The objective consequences of being infected by SARS‐CoV‐2 are higher for the older population (World Health Organization, 2020a, 2020b). However, while the perceptions of a highly transmitted illness threat among the elderly may prompt stress, the empirical evidence is mixed. For instance, Taylor et al. (2008) found that the negative psychological impact of quarantine is more pronounced among younger people. Investigating psychosocial responses towards a national outbreak of SARS in Singapore, Sim et al. (2010) found that psychiatric morbidities are associated with younger age and higher posttraumatic stress symptoms. Most recently, an online survey in the early stages of the COVID‐19 outbreak in China found no association between age and stress (Wang et al., 2020).
|
|
p
|
As such, although older adults face higher risks of severe disease and death due to COVID‐19 (WHO, 2020b), the evidence suggests that older people are less stressed and less affected by psychological consequences of quarantine and social isolation, while younger ones exhibit the highest levels of stress.
|
|
sec
|
Individualism–Collectivism Dimension
The role of culture has been widely studied for decades, with researchers debating how cultural factors may act as a buffer to the environmental stressors or, on the contrary, exacerbate stress levels (Dar, 2017). One of the main focuses of research in this area, the individualism–collectivism dimension (Hofstede, 2001), has been linked to stress processing (Chun et al., 2006). During the current quarantine, people have been forced to renounce their personal enjoyment (e.g. sports, concerts, shopping, travel, social gatherings) for the sake of group needs. Since individualistic, rather than collectivistic, cultures put a higher value on pleasure and hedonism (Schwartz, 2009), it might be reasonable to think that the emotional cost of this quarantine period would be greater in individualistic cultures. In fact, collectivistic (vs. individualistic) cultures put more emphasis on group harmony over personal interests and enjoyment (Triandis et al., 1990). Moreover, Oarga et al. (2015) found that helping behaviours had a stronger association with life satisfaction in countries where helping others constituted a social norm. In this sense, people from collective cultures would be likely to focus on caring for others, thereby alleviating the negative psychological consequences of quarantine. Overall, under the prevailing quarantine, the level of stress is expected to be higher for individualistic cultures compared to the collectivistic ones.
|
|
title
|
Individualism–Collectivism Dimension
|
|
p
|
The role of culture has been widely studied for decades, with researchers debating how cultural factors may act as a buffer to the environmental stressors or, on the contrary, exacerbate stress levels (Dar, 2017). One of the main focuses of research in this area, the individualism–collectivism dimension (Hofstede, 2001), has been linked to stress processing (Chun et al., 2006). During the current quarantine, people have been forced to renounce their personal enjoyment (e.g. sports, concerts, shopping, travel, social gatherings) for the sake of group needs. Since individualistic, rather than collectivistic, cultures put a higher value on pleasure and hedonism (Schwartz, 2009), it might be reasonable to think that the emotional cost of this quarantine period would be greater in individualistic cultures. In fact, collectivistic (vs. individualistic) cultures put more emphasis on group harmony over personal interests and enjoyment (Triandis et al., 1990). Moreover, Oarga et al. (2015) found that helping behaviours had a stronger association with life satisfaction in countries where helping others constituted a social norm. In this sense, people from collective cultures would be likely to focus on caring for others, thereby alleviating the negative psychological consequences of quarantine. Overall, under the prevailing quarantine, the level of stress is expected to be higher for individualistic cultures compared to the collectivistic ones.
|
|
sec
|
Aims and Hypotheses
The present study is one of the first to test how the number of persons an individual is staying with in isolation, along with age, gender, marital status, the number of children, and culture (i.e. individualism vs. collectivism) is associated with experienced stress levels in a large sample from 27 countries and areas. Furthermore, we want to investigate the link between stress levels and educational background and the severity of the COVID‐19 situation in each country or area. Based on the literature review, our pre‐registered hypotheses (see https://osf.io/xf4mj) were as follows:
H1: Individuals living alone and those who are subjected to overcrowding while in isolation experience the highest levels of perceived stress (compared with individuals living with others).
H2: Married and cohabiting persons experience lower levels of stress compared with persons that are single.
H3: Individuals with children would report increased stress levels during the COVID‐19 pandemic compared to people living alone or with adults. Levels of perceived stress would increase as the number of children at home increases.
H4: Younger people experience more stress than older people.
H5: People in collectivistic cultures experience less stress than those in individualistic cultures.
H6: Women experience higher levels of stress compared with men (not pre‐registered hypothesis).
|
|
title
|
Aims and Hypotheses
|
|
p
|
The present study is one of the first to test how the number of persons an individual is staying with in isolation, along with age, gender, marital status, the number of children, and culture (i.e. individualism vs. collectivism) is associated with experienced stress levels in a large sample from 27 countries and areas. Furthermore, we want to investigate the link between stress levels and educational background and the severity of the COVID‐19 situation in each country or area. Based on the literature review, our pre‐registered hypotheses (see https://osf.io/xf4mj) were as follows:
|
|
p
|
H1: Individuals living alone and those who are subjected to overcrowding while in isolation experience the highest levels of perceived stress (compared with individuals living with others).
|
|
p
|
H2: Married and cohabiting persons experience lower levels of stress compared with persons that are single.
|
|
p
|
H3: Individuals with children would report increased stress levels during the COVID‐19 pandemic compared to people living alone or with adults. Levels of perceived stress would increase as the number of children at home increases.
|
|
p
|
H4: Younger people experience more stress than older people.
|
|
p
|
H5: People in collectivistic cultures experience less stress than those in individualistic cultures.
|
|
p
|
H6: Women experience higher levels of stress compared with men (not pre‐registered hypothesis).
|
|
sec
|
METHODS
Participants
Participants were volunteers, recruited online between 30 March and 6 April 2020, using a snowballing technique. The initial sample included 54,245 Internet users from 27 countries and areas who participated in the COVIDiSTRESS study. However, based on the reliability analysis, we decided to exclude one country (Kosovo). The final sample consisted of 53,524 participants from 26 countries and areas (see the Results section for more details). Participation was not remunerated. The numbers for each country with their respective individualism scores (based on the Hofstede index; Hofstede, 2001; Hofstede et al., 2010), along with detailed descriptions of all participants from each country are presented in the Supplementary Material, Table S1 (https://osf.io/cznr8/). As stated in the pre‐registration, we only included countries and areas with N > 300 participants.
This sample consisted of 34,475 women (64.6%), 18,288 men (34.2%) and 645 people who responded with “other/would rather not say” (1.2%). The age in the sample ranged between 18 and 110 years (M = 38.17, SD = 13.79). The sample included 26,429 married or cohabiting participants (49.4%), 20,465 singles (38.2%), 3,626 divorced or widowed (6.8%), and 3,004 people who would rather not answer (5.6%); 629 individuals did not attain any education (1.2%), 709 completed 6 years of school (1.3%), 741 completed 9 years of school (1.4%), 5,133 completed 12 years of school (9.6%), 11,639 completed some college, short continuing education or equivalent (21.8%), 28,983 attained a college degree, bachelor, or masters (54.3%), and 5,497 obtained a PhD/doctorate (10.3%); 24,367 participants indicated that they are isolated (45.8%), 62—isolated in medical facility or similar location (0.1%), 25,867—life carries on with minor changes (48.7%), and 2,862—life carries on as usual (5.4%) (see Supplementary Material for participants’ detailed descriptives across countries and areas).
Measures and Procedure
Both the collaborating researchers of the COVIDiSTRESS project and the participants shared the invitation to participate in the study through various online channels (e.g. social media and private communication). The questionnaire was administered via Qualtrics. Information on all the components of the survey can be found in the COVIDiSTRESS Global Survey Research Network (2020), while in this paper, we report only the measures relevant to the analyses conducted.
The levels of perceived stress were measured with the widely used Perceived Stress Scale (PSS; Cohen et al., 1983; Cohen & Williamson, 1988). It includes 10 items (four reverse‐scored) concerning various subjective feelings related to problems, behaviours, and ways of coping, on a 5‐point scale from 0 (never) to 4 (very often). Cronbach’s alphas for each of the language versions of the scale are presented in the Supplementary Material (https://osf.io/cznr8/).
Statistical Analyses
A detailed description of the cleaning process of the database can be found in the Supplementary Material. Furthermore, we transformed the continuous variables of interest that violated the normal distribution (i.e. the number of adults, dependants, children, and a measure of COVID‐19 severity), by square rooting them.1 As a robustness check, in a separate model, we have also included these variables in a raw form: the results remained almost the same.
We began our analyses by testing the cross‐cultural equivalence of the PSS‐10. To this end, in a multi‐group factor analysis in R (R Core Team, 2014), using the following packages: lavaan (Rosseel, 2012), and semTools (Jorgensen et al., 2019), we compared the models assuming a two‐factor structure (positive and negative, with the latter consisting of reversed items; Roberti et al., 2006) across 26 countries and areas (configural invariance), with a model with factor loadings and latent correlations constrained to be equal (metric invariance), and items' intercepts to be the same in all groups (scalar invariance). When evaluating the model fit, we relied on the usually applied criteria (Hu & Bentler, 1999), in which a comparative fit index (CFI) and Tucker Lewis Index (TLI) above .90 indicate adequate fit, whereas a root‐mean‐square error of approximation (RMSEA) below .08 and a standardised root‐mean‐square residual (SRMR) below .06 indicates no misfit. When evaluating measurement equivalence, we compared the configural invariance model with the metric invariance model, and then the metric invariance model with the scalar invariance model. As these models were characterised by a growing complexity (each subsequent model was nested within the previous one), while assessing models’ superiority we relied on the cut‐off criteria recommended for testing measurement invariance: a change of CFI (∆CFI) less than .01 (∆CFI < .01), a change of RMSEA of less than .015 (∆RMSEA < .015), and a change of SRMR less than .01 (∆SRMR < .01) would indicate that the two models compared do not differ in terms of model fit (Chen et al., 2008; Cheung & Rensvold, 2002).
After computing Cronbach’s alpha for the PSS‐10 for all the countries and areas separately and jointly, we tested multilevel regression models, using Jamovi (Version 1.2.16.). In these models, participants were nested within countries, and the number of adults, children, marital status, education level, and gender were group‐mean‐standardised. In the first model, we regressed a composite score of perceived stress levels on the number of adults and the number of children, as well as participants’ gender, age, marital status, education level, COVID‐19 severity (the number of confirmed cases at the time of data collection (World Health Organization, 2020a, 2020b), divided by the number of the country’s citizens (Worldometer, 2020; Level‐2 variable), and culture (i.e. collectivistic vs. individualistic: Level‐2 variable). The first model we created included a random intercept of stress on a country level. The second model differed, as we additionally estimated random slopes of the number of children. The third model tested for potential curvilinear links by adding two variables in a squared form (i.e. the number of adults and children). The final, fourth model included random intercept for the stress level as well as all random slopes for age, gender, and marital status.
|
|
title
|
METHODS
|
|
sec
|
Participants
Participants were volunteers, recruited online between 30 March and 6 April 2020, using a snowballing technique. The initial sample included 54,245 Internet users from 27 countries and areas who participated in the COVIDiSTRESS study. However, based on the reliability analysis, we decided to exclude one country (Kosovo). The final sample consisted of 53,524 participants from 26 countries and areas (see the Results section for more details). Participation was not remunerated. The numbers for each country with their respective individualism scores (based on the Hofstede index; Hofstede, 2001; Hofstede et al., 2010), along with detailed descriptions of all participants from each country are presented in the Supplementary Material, Table S1 (https://osf.io/cznr8/). As stated in the pre‐registration, we only included countries and areas with N > 300 participants.
This sample consisted of 34,475 women (64.6%), 18,288 men (34.2%) and 645 people who responded with “other/would rather not say” (1.2%). The age in the sample ranged between 18 and 110 years (M = 38.17, SD = 13.79). The sample included 26,429 married or cohabiting participants (49.4%), 20,465 singles (38.2%), 3,626 divorced or widowed (6.8%), and 3,004 people who would rather not answer (5.6%); 629 individuals did not attain any education (1.2%), 709 completed 6 years of school (1.3%), 741 completed 9 years of school (1.4%), 5,133 completed 12 years of school (9.6%), 11,639 completed some college, short continuing education or equivalent (21.8%), 28,983 attained a college degree, bachelor, or masters (54.3%), and 5,497 obtained a PhD/doctorate (10.3%); 24,367 participants indicated that they are isolated (45.8%), 62—isolated in medical facility or similar location (0.1%), 25,867—life carries on with minor changes (48.7%), and 2,862—life carries on as usual (5.4%) (see Supplementary Material for participants’ detailed descriptives across countries and areas).
|
|
title
|
Participants
|
|
p
|
Participants were volunteers, recruited online between 30 March and 6 April 2020, using a snowballing technique. The initial sample included 54,245 Internet users from 27 countries and areas who participated in the COVIDiSTRESS study. However, based on the reliability analysis, we decided to exclude one country (Kosovo). The final sample consisted of 53,524 participants from 26 countries and areas (see the Results section for more details). Participation was not remunerated. The numbers for each country with their respective individualism scores (based on the Hofstede index; Hofstede, 2001; Hofstede et al., 2010), along with detailed descriptions of all participants from each country are presented in the Supplementary Material, Table S1 (https://osf.io/cznr8/). As stated in the pre‐registration, we only included countries and areas with N > 300 participants.
|
|
p
|
This sample consisted of 34,475 women (64.6%), 18,288 men (34.2%) and 645 people who responded with “other/would rather not say” (1.2%). The age in the sample ranged between 18 and 110 years (M = 38.17, SD = 13.79). The sample included 26,429 married or cohabiting participants (49.4%), 20,465 singles (38.2%), 3,626 divorced or widowed (6.8%), and 3,004 people who would rather not answer (5.6%); 629 individuals did not attain any education (1.2%), 709 completed 6 years of school (1.3%), 741 completed 9 years of school (1.4%), 5,133 completed 12 years of school (9.6%), 11,639 completed some college, short continuing education or equivalent (21.8%), 28,983 attained a college degree, bachelor, or masters (54.3%), and 5,497 obtained a PhD/doctorate (10.3%); 24,367 participants indicated that they are isolated (45.8%), 62—isolated in medical facility or similar location (0.1%), 25,867—life carries on with minor changes (48.7%), and 2,862—life carries on as usual (5.4%) (see Supplementary Material for participants’ detailed descriptives across countries and areas).
|
|
sec
|
Measures and Procedure
Both the collaborating researchers of the COVIDiSTRESS project and the participants shared the invitation to participate in the study through various online channels (e.g. social media and private communication). The questionnaire was administered via Qualtrics. Information on all the components of the survey can be found in the COVIDiSTRESS Global Survey Research Network (2020), while in this paper, we report only the measures relevant to the analyses conducted.
The levels of perceived stress were measured with the widely used Perceived Stress Scale (PSS; Cohen et al., 1983; Cohen & Williamson, 1988). It includes 10 items (four reverse‐scored) concerning various subjective feelings related to problems, behaviours, and ways of coping, on a 5‐point scale from 0 (never) to 4 (very often). Cronbach’s alphas for each of the language versions of the scale are presented in the Supplementary Material (https://osf.io/cznr8/).
|
|
title
|
Measures and Procedure
|
|
p
|
Both the collaborating researchers of the COVIDiSTRESS project and the participants shared the invitation to participate in the study through various online channels (e.g. social media and private communication). The questionnaire was administered via Qualtrics. Information on all the components of the survey can be found in the COVIDiSTRESS Global Survey Research Network (2020), while in this paper, we report only the measures relevant to the analyses conducted.
|
|
p
|
The levels of perceived stress were measured with the widely used Perceived Stress Scale (PSS; Cohen et al., 1983; Cohen & Williamson, 1988). It includes 10 items (four reverse‐scored) concerning various subjective feelings related to problems, behaviours, and ways of coping, on a 5‐point scale from 0 (never) to 4 (very often). Cronbach’s alphas for each of the language versions of the scale are presented in the Supplementary Material (https://osf.io/cznr8/).
|
|
sec
|
Statistical Analyses
A detailed description of the cleaning process of the database can be found in the Supplementary Material. Furthermore, we transformed the continuous variables of interest that violated the normal distribution (i.e. the number of adults, dependants, children, and a measure of COVID‐19 severity), by square rooting them.1 As a robustness check, in a separate model, we have also included these variables in a raw form: the results remained almost the same.
We began our analyses by testing the cross‐cultural equivalence of the PSS‐10. To this end, in a multi‐group factor analysis in R (R Core Team, 2014), using the following packages: lavaan (Rosseel, 2012), and semTools (Jorgensen et al., 2019), we compared the models assuming a two‐factor structure (positive and negative, with the latter consisting of reversed items; Roberti et al., 2006) across 26 countries and areas (configural invariance), with a model with factor loadings and latent correlations constrained to be equal (metric invariance), and items' intercepts to be the same in all groups (scalar invariance). When evaluating the model fit, we relied on the usually applied criteria (Hu & Bentler, 1999), in which a comparative fit index (CFI) and Tucker Lewis Index (TLI) above .90 indicate adequate fit, whereas a root‐mean‐square error of approximation (RMSEA) below .08 and a standardised root‐mean‐square residual (SRMR) below .06 indicates no misfit. When evaluating measurement equivalence, we compared the configural invariance model with the metric invariance model, and then the metric invariance model with the scalar invariance model. As these models were characterised by a growing complexity (each subsequent model was nested within the previous one), while assessing models’ superiority we relied on the cut‐off criteria recommended for testing measurement invariance: a change of CFI (∆CFI) less than .01 (∆CFI < .01), a change of RMSEA of less than .015 (∆RMSEA < .015), and a change of SRMR less than .01 (∆SRMR < .01) would indicate that the two models compared do not differ in terms of model fit (Chen et al., 2008; Cheung & Rensvold, 2002).
After computing Cronbach’s alpha for the PSS‐10 for all the countries and areas separately and jointly, we tested multilevel regression models, using Jamovi (Version 1.2.16.). In these models, participants were nested within countries, and the number of adults, children, marital status, education level, and gender were group‐mean‐standardised. In the first model, we regressed a composite score of perceived stress levels on the number of adults and the number of children, as well as participants’ gender, age, marital status, education level, COVID‐19 severity (the number of confirmed cases at the time of data collection (World Health Organization, 2020a, 2020b), divided by the number of the country’s citizens (Worldometer, 2020; Level‐2 variable), and culture (i.e. collectivistic vs. individualistic: Level‐2 variable). The first model we created included a random intercept of stress on a country level. The second model differed, as we additionally estimated random slopes of the number of children. The third model tested for potential curvilinear links by adding two variables in a squared form (i.e. the number of adults and children). The final, fourth model included random intercept for the stress level as well as all random slopes for age, gender, and marital status.
|
|
title
|
Statistical Analyses
|
|
p
|
A detailed description of the cleaning process of the database can be found in the Supplementary Material. Furthermore, we transformed the continuous variables of interest that violated the normal distribution (i.e. the number of adults, dependants, children, and a measure of COVID‐19 severity), by square rooting them.1 As a robustness check, in a separate model, we have also included these variables in a raw form: the results remained almost the same.
|
|
footnote
|
1 As a robustness check, in a separate model, we have also included these variables in a raw form: the results remained almost the same.
|
|
label
|
1
|
|
p
|
As a robustness check, in a separate model, we have also included these variables in a raw form: the results remained almost the same.
|
|
p
|
We began our analyses by testing the cross‐cultural equivalence of the PSS‐10. To this end, in a multi‐group factor analysis in R (R Core Team, 2014), using the following packages: lavaan (Rosseel, 2012), and semTools (Jorgensen et al., 2019), we compared the models assuming a two‐factor structure (positive and negative, with the latter consisting of reversed items; Roberti et al., 2006) across 26 countries and areas (configural invariance), with a model with factor loadings and latent correlations constrained to be equal (metric invariance), and items' intercepts to be the same in all groups (scalar invariance). When evaluating the model fit, we relied on the usually applied criteria (Hu & Bentler, 1999), in which a comparative fit index (CFI) and Tucker Lewis Index (TLI) above .90 indicate adequate fit, whereas a root‐mean‐square error of approximation (RMSEA) below .08 and a standardised root‐mean‐square residual (SRMR) below .06 indicates no misfit. When evaluating measurement equivalence, we compared the configural invariance model with the metric invariance model, and then the metric invariance model with the scalar invariance model. As these models were characterised by a growing complexity (each subsequent model was nested within the previous one), while assessing models’ superiority we relied on the cut‐off criteria recommended for testing measurement invariance: a change of CFI (∆CFI) less than .01 (∆CFI < .01), a change of RMSEA of less than .015 (∆RMSEA < .015), and a change of SRMR less than .01 (∆SRMR < .01) would indicate that the two models compared do not differ in terms of model fit (Chen et al., 2008; Cheung & Rensvold, 2002).
|
|
p
|
After computing Cronbach’s alpha for the PSS‐10 for all the countries and areas separately and jointly, we tested multilevel regression models, using Jamovi (Version 1.2.16.). In these models, participants were nested within countries, and the number of adults, children, marital status, education level, and gender were group‐mean‐standardised. In the first model, we regressed a composite score of perceived stress levels on the number of adults and the number of children, as well as participants’ gender, age, marital status, education level, COVID‐19 severity (the number of confirmed cases at the time of data collection (World Health Organization, 2020a, 2020b), divided by the number of the country’s citizens (Worldometer, 2020; Level‐2 variable), and culture (i.e. collectivistic vs. individualistic: Level‐2 variable). The first model we created included a random intercept of stress on a country level. The second model differed, as we additionally estimated random slopes of the number of children. The third model tested for potential curvilinear links by adding two variables in a squared form (i.e. the number of adults and children). The final, fourth model included random intercept for the stress level as well as all random slopes for age, gender, and marital status.
|
|
sec
|
RESULTS
The PSS‐10 was found to be highly reliable (overall Cronbach’s alpha: .869), with the exception of Kosovo (Cronbach’s alpha: .688). Confirmatory factor analysis indicated that the overall model fit the data well in the case of configural invariance (CFI = .996, TLI = .995, SRMR = .028, RMSEA = .028), but not metric invariance (CFI = .986, TLI = .985, SRMR = .044, RMSEA = .046). Based on the poor reliability of Kosovo, we decided to exclude data from this country and re‐estimated the multi‐group CFA model. After excluding data from Kosovo, the model slightly improved: configural (CFI = .996, TLI = .995, SRMR = .028, RMSEA = .028), metric (CFI = .987, TLI = .996, SRMR = .043, RMSEA = .044), and scalar invariance (CFI = .950, TLI = .955, SRMR = .073, RMSEA = .081). However, as the scalar invariance was not reached (a change in fit model indices was above recommended criteria, i.e. ΔCFI and ΔSRMR> .01), in the next step, we tested for a partial scalar invariance (Byrne et al., 1989; Steenkamp & Baumgartner, 1998). The final model fit the data well, and partial scalar invariance was reached (CFI = .983, TLI = .983, SRMR = .047, RMSEA = .050). Table 1 depicts mean stress levels across countries, and Table 2 presents correlations across variables of interest. The mean stress level in the present study in, for instance, the USA (M = 17.50) was significantly higher than the mean stress level (M = 13.19) indicated by the PSS‐10 norms (Cohen et al., 1994; 13.19), t(3896) = 19.45, p < .001, d = 0.63.
TABLE 1 Mean Levels of Perceived Stress Across Countries and Areas
Country Number of participants Mean stress level (SD) Country Number of participants Mean stress level (SD)
Argentina 2286 17.53 (7.38) Indonesia 938 18.00 (5.75)
Belgium 339 16.38 (7.27) Italy 525 16.04 (6.89)
Bosnia and Herzegovina 660 18.95 (6.73) Japan 3190 20.08 (5.74)
Bulgaria 2413 18.78 (7.10) Lithuania 4450 15.32 (6.88)
Canada 307 17.81 (6.66) Mexico 5781 17.33 (7.32)
Croatia 900 19.55 (6.75) Poland 2094 20.19 (7.28)
Czech Republic 461 17.54 (6.95) Portugal 425 18.81 (7.25)
Denmark 5200 14.22 (7.21) Spain 363 16.80 (7.19)
Finland 7464 14.85 (7.50) Switzerland 719 13.93 (6.66)
France 9395 15.71 (7.38) Taiwan 869 15.97 (6.85)
Germany 616 16.52 (6.87) Turkey 683 21.71 (6.67)
Greece 304 17.43 (6.41) United Kingdom 886 17.53 (7.50)
Hungary 745 17.60 (6.06) United States 1511 17.50 (7.36)
Note
Mean stress scores on 10 items on a 5‐point Likert scale, ranging from 0 – almost never, to 4 – always.
John Wiley & Sons, Ltd This article is being made freely available through PubMed Central as part of the COVID-19 public health emergency response. It can be used for unrestricted research re-use and analysis in any form or by any means with acknowledgement of the original source, for the duration of the public health emergency.
TABLE 2 Correlation Matrix on Main Variables of Interest
1 2 3 4 5 6 7
1. Perceived stress —
2. Gender .09*** —
3. Age −.24*** .04*** —
4. Adults a .06*** ‐.04*** ‐.14*** —
5. Children b .04*** .06*** ‐.03*** .14*** —
6. People c .07*** .00 ‐.13*** .90*** .56*** —
7. Dependents d −.02*** .06*** .20*** .08*** .53*** .31*** —
8. Education e −.04*** −.03*** ‐.01 ‐.01** .03*** .00 .03***
Note:
aAdults–the number of adults staying with the participants during the relative isolation. bChildren–the number of children staying with the participants during the relative isolation. cPeople–the number of people staying with the participants during the relative isolation. dDependants–the number of dependants staying with the participants during the relative isolation. eEducation–education status, with a higher number representing a higher completed education level. *p < .05; **p < .01; ***p < .001.
John Wiley & Sons, Ltd This article is being made freely available through PubMed Central as part of the COVID-19 public health emergency response. It can be used for unrestricted research re-use and analysis in any form or by any means with acknowledgement of the original source, for the duration of the public health emergency. Having established configural, metric, and partly scalar invariance, we proceeded with multilevel models to examine whether the number of people (adults and children) our participants were staying with during the pandemic was associated with their perceived stress level, controlling for gender, age, education level, marital status, scores on the individualism measure, and COVID‐19 severity. Taking into consideration the four models we described in the Statistical Analyses section, based on the Bayesian Information Criterion, the fourth model (with random effects for both slopes and intercepts) showed a superior fit when compared to the others (i.e. ΔBIC > 10; Raftery, 1999; Δ between the first and the fourth model BIC = 432, the second and fourth ΔBIC = 431, the third and fourth ΔBIC = 450). The model that included the squared number of adults and children surrounding the participants in their isolation had substantially worse fit than the second model (ΔBIC = −19); thus, we infer that there is a linear rather than a curvilinear relationship between the number of adults and children, and stress.
As illustrated in Table 3, younger people perceived higher levels of stress compared with older people (β = −0.076, SE = 0.016, p < .001); the number of children an individual was staying with was positively related to perceived stress levels (β = 0.021, SE = 0.003, p < .001); women were more stressed than men (β = 0.065, SE = 0.012, p < .001); married or cohabiting people were less stressed than singles (β = −0.132, SE = 0.048, p < .01); education was negatively related to stress levels (β = −0.022, SE = 0.003, p < .001), meaning that the higher the completed education, the less stressed a participant was, and participants from countries and areas with higher COVID‐19 severity were more stressed (β = 0.090, SE = 0.044, p < .05). The countries’ and areas’ score on the individualism–collectivism continuum was not related to perceived stress (β = −0.087, SE = 0.045, p = .063), and similarly the number of adults the participant was staying with (β = 0.001, SE = 0.003, p = .806).
TABLE 3 Results of a Multilevel Model, with Participants Nested within Countries and Areas, and Random Effects for Both Intercepts and Slopes for Age, Gender, and Marital Status
Fixed effects β SE p
Culturea −0.087 0.045 .063
Age −0.076 0.016 <.001
Adultsb 0.000 0.003 .806
Childrenc 0.021 0.003 <.001
Gender (Men vs. Women) 0.065 0.012 <.001
Educationd −0.022 0.003 <.001
COVID‐19 severitye 0.090 0.044 .050
Marital status (Single vs. Married)f −0.132 0.048 .006
Random effects g SD LRT p
Age 0.077 816.83 <.001
Gender (Men vs. Women) 0.054 366.59 <.001
Marital status (Married vs. Single)f 0.238 368.13 <.001
Note
aCulture–Hofstede’s score of a country's individualism (i.e. the higher the number, the more individualistic the country). bAdults–the number of adults staying with the participants during the relative isolation. cChildren–the number of children staying with the participants during the relative isolation. dEducation–education status, with a higher number representing a higher completed education level: 6–PhD/Doctorate; 5–College degree, bachelor, master; 4–Some College, short continuing education or equivalent; 3–Up to 12 years of school; 2–Up to 9 years of school; 1–Up to 6 years of school; 0–None. eCOVID‐19 severity–the number of confirmed cases divided by the number of countries’ citizens. fMarital status–single individuals are coded as 0, and married or cohabiting individuals as 1. gRandom effects–the likelihood ratio test for each random effect, representing variability of stress levels and the association between stress levels and other variables of interest between countries. For the intercept: SD = 0.261, ICC = 0.130. Significant results are bolded.
John Wiley & Sons, Ltd This article is being made freely available through PubMed Central as part of the COVID-19 public health emergency response. It can be used for unrestricted research re-use and analysis in any form or by any means with acknowledgement of the original source, for the duration of the public health emergency.
|
|
title
|
RESULTS
|
|
p
|
The PSS‐10 was found to be highly reliable (overall Cronbach’s alpha: .869), with the exception of Kosovo (Cronbach’s alpha: .688). Confirmatory factor analysis indicated that the overall model fit the data well in the case of configural invariance (CFI = .996, TLI = .995, SRMR = .028, RMSEA = .028), but not metric invariance (CFI = .986, TLI = .985, SRMR = .044, RMSEA = .046). Based on the poor reliability of Kosovo, we decided to exclude data from this country and re‐estimated the multi‐group CFA model. After excluding data from Kosovo, the model slightly improved: configural (CFI = .996, TLI = .995, SRMR = .028, RMSEA = .028), metric (CFI = .987, TLI = .996, SRMR = .043, RMSEA = .044), and scalar invariance (CFI = .950, TLI = .955, SRMR = .073, RMSEA = .081). However, as the scalar invariance was not reached (a change in fit model indices was above recommended criteria, i.e. ΔCFI and ΔSRMR> .01), in the next step, we tested for a partial scalar invariance (Byrne et al., 1989; Steenkamp & Baumgartner, 1998). The final model fit the data well, and partial scalar invariance was reached (CFI = .983, TLI = .983, SRMR = .047, RMSEA = .050). Table 1 depicts mean stress levels across countries, and Table 2 presents correlations across variables of interest. The mean stress level in the present study in, for instance, the USA (M = 17.50) was significantly higher than the mean stress level (M = 13.19) indicated by the PSS‐10 norms (Cohen et al., 1994; 13.19), t(3896) = 19.45, p < .001, d = 0.63.
|
|
table-wrap
|
TABLE 1 Mean Levels of Perceived Stress Across Countries and Areas
Country Number of participants Mean stress level (SD) Country Number of participants Mean stress level (SD)
Argentina 2286 17.53 (7.38) Indonesia 938 18.00 (5.75)
Belgium 339 16.38 (7.27) Italy 525 16.04 (6.89)
Bosnia and Herzegovina 660 18.95 (6.73) Japan 3190 20.08 (5.74)
Bulgaria 2413 18.78 (7.10) Lithuania 4450 15.32 (6.88)
Canada 307 17.81 (6.66) Mexico 5781 17.33 (7.32)
Croatia 900 19.55 (6.75) Poland 2094 20.19 (7.28)
Czech Republic 461 17.54 (6.95) Portugal 425 18.81 (7.25)
Denmark 5200 14.22 (7.21) Spain 363 16.80 (7.19)
Finland 7464 14.85 (7.50) Switzerland 719 13.93 (6.66)
France 9395 15.71 (7.38) Taiwan 869 15.97 (6.85)
Germany 616 16.52 (6.87) Turkey 683 21.71 (6.67)
Greece 304 17.43 (6.41) United Kingdom 886 17.53 (7.50)
Hungary 745 17.60 (6.06) United States 1511 17.50 (7.36)
Note
Mean stress scores on 10 items on a 5‐point Likert scale, ranging from 0 – almost never, to 4 – always.
John Wiley & Sons, Ltd This article is being made freely available through PubMed Central as part of the COVID-19 public health emergency response. It can be used for unrestricted research re-use and analysis in any form or by any means with acknowledgement of the original source, for the duration of the public health emergency.
|
|
label
|
TABLE 1
|
|
caption
|
Mean Levels of Perceived Stress Across Countries and Areas
|
|
p
|
Mean Levels of Perceived Stress Across Countries and Areas
|
|
table
|
Country Number of participants Mean stress level (SD) Country Number of participants Mean stress level (SD)
Argentina 2286 17.53 (7.38) Indonesia 938 18.00 (5.75)
Belgium 339 16.38 (7.27) Italy 525 16.04 (6.89)
Bosnia and Herzegovina 660 18.95 (6.73) Japan 3190 20.08 (5.74)
Bulgaria 2413 18.78 (7.10) Lithuania 4450 15.32 (6.88)
Canada 307 17.81 (6.66) Mexico 5781 17.33 (7.32)
Croatia 900 19.55 (6.75) Poland 2094 20.19 (7.28)
Czech Republic 461 17.54 (6.95) Portugal 425 18.81 (7.25)
Denmark 5200 14.22 (7.21) Spain 363 16.80 (7.19)
Finland 7464 14.85 (7.50) Switzerland 719 13.93 (6.66)
France 9395 15.71 (7.38) Taiwan 869 15.97 (6.85)
Germany 616 16.52 (6.87) Turkey 683 21.71 (6.67)
Greece 304 17.43 (6.41) United Kingdom 886 17.53 (7.50)
Hungary 745 17.60 (6.06) United States 1511 17.50 (7.36)
|
|
tr
|
Country Number of participants Mean stress level (SD) Country Number of participants Mean stress level (SD)
|
|
th
|
Country
|
|
th
|
Number of participants
|
|
th
|
Mean stress level (SD)
|
|
th
|
Country
|
|
th
|
Number of participants
|
|
th
|
Mean stress level (SD)
|
|
tr
|
Argentina 2286 17.53 (7.38) Indonesia 938 18.00 (5.75)
|
|
td
|
Argentina
|
|
td
|
2286
|
|
td
|
17.53 (7.38)
|
|
td
|
Indonesia
|
|
td
|
938
|
|
td
|
18.00 (5.75)
|
|
tr
|
Belgium 339 16.38 (7.27) Italy 525 16.04 (6.89)
|
|
td
|
Belgium
|
|
td
|
339
|
|
td
|
16.38 (7.27)
|
|
td
|
Italy
|
|
td
|
525
|
|
td
|
16.04 (6.89)
|
|
tr
|
Bosnia and Herzegovina 660 18.95 (6.73) Japan 3190 20.08 (5.74)
|
|
td
|
Bosnia and Herzegovina
|
|
td
|
660
|
|
td
|
18.95 (6.73)
|
|
td
|
Japan
|
|
td
|
3190
|
|
td
|
20.08 (5.74)
|
|
tr
|
Bulgaria 2413 18.78 (7.10) Lithuania 4450 15.32 (6.88)
|
|
td
|
Bulgaria
|
|
td
|
2413
|
|
td
|
18.78 (7.10)
|
|
td
|
Lithuania
|
|
td
|
4450
|
|
td
|
15.32 (6.88)
|
|
tr
|
Canada 307 17.81 (6.66) Mexico 5781 17.33 (7.32)
|
|
td
|
Canada
|
|
td
|
307
|
|
td
|
17.81 (6.66)
|
|
td
|
Mexico
|
|
td
|
5781
|
|
td
|
17.33 (7.32)
|
|
tr
|
Croatia 900 19.55 (6.75) Poland 2094 20.19 (7.28)
|
|
td
|
Croatia
|
|
td
|
900
|
|
td
|
19.55 (6.75)
|
|
td
|
Poland
|
|
td
|
2094
|
|
td
|
20.19 (7.28)
|
|
tr
|
Czech Republic 461 17.54 (6.95) Portugal 425 18.81 (7.25)
|
|
td
|
Czech Republic
|
|
td
|
461
|
|
td
|
17.54 (6.95)
|
|
td
|
Portugal
|
|
td
|
425
|
|
td
|
18.81 (7.25)
|
|
tr
|
Denmark 5200 14.22 (7.21) Spain 363 16.80 (7.19)
|
|
td
|
Denmark
|
|
td
|
5200
|
|
td
|
14.22 (7.21)
|
|
td
|
Spain
|
|
td
|
363
|
|
td
|
16.80 (7.19)
|
|
tr
|
Finland 7464 14.85 (7.50) Switzerland 719 13.93 (6.66)
|
|
td
|
Finland
|
|
td
|
7464
|
|
td
|
14.85 (7.50)
|
|
td
|
Switzerland
|
|
td
|
719
|
|
td
|
13.93 (6.66)
|
|
tr
|
France 9395 15.71 (7.38) Taiwan 869 15.97 (6.85)
|
|
td
|
France
|
|
td
|
9395
|
|
td
|
15.71 (7.38)
|
|
td
|
Taiwan
|
|
td
|
869
|
|
td
|
15.97 (6.85)
|
|
tr
|
Germany 616 16.52 (6.87) Turkey 683 21.71 (6.67)
|
|
td
|
Germany
|
|
td
|
616
|
|
td
|
16.52 (6.87)
|
|
td
|
Turkey
|
|
td
|
683
|
|
td
|
21.71 (6.67)
|
|
tr
|
Greece 304 17.43 (6.41) United Kingdom 886 17.53 (7.50)
|
|
td
|
Greece
|
|
td
|
304
|
|
td
|
17.43 (6.41)
|
|
td
|
United Kingdom
|
|
td
|
886
|
|
td
|
17.53 (7.50)
|
|
tr
|
Hungary 745 17.60 (6.06) United States 1511 17.50 (7.36)
|
|
td
|
Hungary
|
|
td
|
745
|
|
td
|
17.60 (6.06)
|
|
td
|
United States
|
|
td
|
1511
|
|
td
|
17.50 (7.36)
|
|
table-wrap-foot
|
Note
Mean stress scores on 10 items on a 5‐point Likert scale, ranging from 0 – almost never, to 4 – always.
|
|
title
|
Note
|
|
footnote
|
Mean stress scores on 10 items on a 5‐point Likert scale, ranging from 0 – almost never, to 4 – always.
|
|
p
|
Mean stress scores on 10 items on a 5‐point Likert scale, ranging from 0 – almost never, to 4 – always.
|
|
table-wrap
|
TABLE 2 Correlation Matrix on Main Variables of Interest
1 2 3 4 5 6 7
1. Perceived stress —
2. Gender .09*** —
3. Age −.24*** .04*** —
4. Adults a .06*** ‐.04*** ‐.14*** —
5. Children b .04*** .06*** ‐.03*** .14*** —
6. People c .07*** .00 ‐.13*** .90*** .56*** —
7. Dependents d −.02*** .06*** .20*** .08*** .53*** .31*** —
8. Education e −.04*** −.03*** ‐.01 ‐.01** .03*** .00 .03***
Note:
aAdults–the number of adults staying with the participants during the relative isolation. bChildren–the number of children staying with the participants during the relative isolation. cPeople–the number of people staying with the participants during the relative isolation. dDependants–the number of dependants staying with the participants during the relative isolation. eEducation–education status, with a higher number representing a higher completed education level. *p < .05; **p < .01; ***p < .001.
John Wiley & Sons, Ltd This article is being made freely available through PubMed Central as part of the COVID-19 public health emergency response. It can be used for unrestricted research re-use and analysis in any form or by any means with acknowledgement of the original source, for the duration of the public health emergency.
|
|
label
|
TABLE 2
|
|
caption
|
Correlation Matrix on Main Variables of Interest
|
|
p
|
Correlation Matrix on Main Variables of Interest
|
|
table
|
1 2 3 4 5 6 7
1. Perceived stress —
2. Gender .09*** —
3. Age −.24*** .04*** —
4. Adults a .06*** ‐.04*** ‐.14*** —
5. Children b .04*** .06*** ‐.03*** .14*** —
6. People c .07*** .00 ‐.13*** .90*** .56*** —
7. Dependents d −.02*** .06*** .20*** .08*** .53*** .31*** —
8. Education e −.04*** −.03*** ‐.01 ‐.01** .03*** .00 .03***
|
|
tr
|
1 2 3 4 5 6 7
|
|
th
|
1
|
|
th
|
2
|
|
th
|
3
|
|
th
|
4
|
|
th
|
5
|
|
th
|
6
|
|
th
|
7
|
|
tr
|
1. Perceived stress —
|
|
td
|
1. Perceived stress
|
|
td
|
—
|
|
tr
|
2. Gender .09*** —
|
|
td
|
2. Gender
|
|
td
|
.09***
|
|
td
|
—
|
|
tr
|
3. Age −.24*** .04*** —
|
|
td
|
3. Age
|
|
td
|
−.24***
|
|
td
|
.04***
|
|
td
|
—
|
|
tr
|
4. Adults a .06*** ‐.04*** ‐.14*** —
|
|
td
|
4. Adults a
|
|
td
|
.06***
|
|
td
|
‐.04***
|
|
td
|
‐.14***
|
|
td
|
—
|
|
tr
|
5. Children b .04*** .06*** ‐.03*** .14*** —
|
|
td
|
5. Children b
|
|
td
|
.04***
|
|
td
|
.06***
|
|
td
|
‐.03***
|
|
td
|
.14***
|
|
td
|
—
|
|
tr
|
6. People c .07*** .00 ‐.13*** .90*** .56*** —
|
|
td
|
6. People c
|
|
td
|
.07***
|
|
td
|
.00
|
|
td
|
‐.13***
|
|
td
|
.90***
|
|
td
|
.56***
|
|
td
|
—
|
|
tr
|
7. Dependents d −.02*** .06*** .20*** .08*** .53*** .31*** —
|
|
td
|
7. Dependents d
|
|
td
|
−.02***
|
|
td
|
.06***
|
|
td
|
.20***
|
|
td
|
.08***
|
|
td
|
.53***
|
|
td
|
.31***
|
|
td
|
—
|
|
tr
|
8. Education e −.04*** −.03*** ‐.01 ‐.01** .03*** .00 .03***
|
|
td
|
8. Education e
|
|
td
|
−.04***
|
|
td
|
−.03***
|
|
td
|
‐.01
|
|
td
|
‐.01**
|
|
td
|
.03***
|
|
td
|
.00
|
|
td
|
.03***
|
|
table-wrap-foot
|
Note:
aAdults–the number of adults staying with the participants during the relative isolation. bChildren–the number of children staying with the participants during the relative isolation. cPeople–the number of people staying with the participants during the relative isolation. dDependants–the number of dependants staying with the participants during the relative isolation. eEducation–education status, with a higher number representing a higher completed education level. *p < .05; **p < .01; ***p < .001.
|
|
title
|
Note:
|
|
footnote
|
aAdults–the number of adults staying with the participants during the relative isolation. bChildren–the number of children staying with the participants during the relative isolation. cPeople–the number of people staying with the participants during the relative isolation. dDependants–the number of dependants staying with the participants during the relative isolation. eEducation–education status, with a higher number representing a higher completed education level. *p < .05; **p < .01; ***p < .001.
|
|
p
|
aAdults–the number of adults staying with the participants during the relative isolation. bChildren–the number of children staying with the participants during the relative isolation. cPeople–the number of people staying with the participants during the relative isolation. dDependants–the number of dependants staying with the participants during the relative isolation. eEducation–education status, with a higher number representing a higher completed education level. *p < .05; **p < .01; ***p < .001.
|
|
p
|
Having established configural, metric, and partly scalar invariance, we proceeded with multilevel models to examine whether the number of people (adults and children) our participants were staying with during the pandemic was associated with their perceived stress level, controlling for gender, age, education level, marital status, scores on the individualism measure, and COVID‐19 severity. Taking into consideration the four models we described in the Statistical Analyses section, based on the Bayesian Information Criterion, the fourth model (with random effects for both slopes and intercepts) showed a superior fit when compared to the others (i.e. ΔBIC > 10; Raftery, 1999; Δ between the first and the fourth model BIC = 432, the second and fourth ΔBIC = 431, the third and fourth ΔBIC = 450). The model that included the squared number of adults and children surrounding the participants in their isolation had substantially worse fit than the second model (ΔBIC = −19); thus, we infer that there is a linear rather than a curvilinear relationship between the number of adults and children, and stress.
|
|
p
|
As illustrated in Table 3, younger people perceived higher levels of stress compared with older people (β = −0.076, SE = 0.016, p < .001); the number of children an individual was staying with was positively related to perceived stress levels (β = 0.021, SE = 0.003, p < .001); women were more stressed than men (β = 0.065, SE = 0.012, p < .001); married or cohabiting people were less stressed than singles (β = −0.132, SE = 0.048, p < .01); education was negatively related to stress levels (β = −0.022, SE = 0.003, p < .001), meaning that the higher the completed education, the less stressed a participant was, and participants from countries and areas with higher COVID‐19 severity were more stressed (β = 0.090, SE = 0.044, p < .05). The countries’ and areas’ score on the individualism–collectivism continuum was not related to perceived stress (β = −0.087, SE = 0.045, p = .063), and similarly the number of adults the participant was staying with (β = 0.001, SE = 0.003, p = .806).
|
|
table-wrap
|
TABLE 3 Results of a Multilevel Model, with Participants Nested within Countries and Areas, and Random Effects for Both Intercepts and Slopes for Age, Gender, and Marital Status
Fixed effects β SE p
Culturea −0.087 0.045 .063
Age −0.076 0.016 <.001
Adultsb 0.000 0.003 .806
Childrenc 0.021 0.003 <.001
Gender (Men vs. Women) 0.065 0.012 <.001
Educationd −0.022 0.003 <.001
COVID‐19 severitye 0.090 0.044 .050
Marital status (Single vs. Married)f −0.132 0.048 .006
Random effects g SD LRT p
Age 0.077 816.83 <.001
Gender (Men vs. Women) 0.054 366.59 <.001
Marital status (Married vs. Single)f 0.238 368.13 <.001
Note
aCulture–Hofstede’s score of a country's individualism (i.e. the higher the number, the more individualistic the country). bAdults–the number of adults staying with the participants during the relative isolation. cChildren–the number of children staying with the participants during the relative isolation. dEducation–education status, with a higher number representing a higher completed education level: 6–PhD/Doctorate; 5–College degree, bachelor, master; 4–Some College, short continuing education or equivalent; 3–Up to 12 years of school; 2–Up to 9 years of school; 1–Up to 6 years of school; 0–None. eCOVID‐19 severity–the number of confirmed cases divided by the number of countries’ citizens. fMarital status–single individuals are coded as 0, and married or cohabiting individuals as 1. gRandom effects–the likelihood ratio test for each random effect, representing variability of stress levels and the association between stress levels and other variables of interest between countries. For the intercept: SD = 0.261, ICC = 0.130. Significant results are bolded.
John Wiley & Sons, Ltd This article is being made freely available through PubMed Central as part of the COVID-19 public health emergency response. It can be used for unrestricted research re-use and analysis in any form or by any means with acknowledgement of the original source, for the duration of the public health emergency.
|
|
label
|
TABLE 3
|
|
caption
|
Results of a Multilevel Model, with Participants Nested within Countries and Areas, and Random Effects for Both Intercepts and Slopes for Age, Gender, and Marital Status
|
|
p
|
Results of a Multilevel Model, with Participants Nested within Countries and Areas, and Random Effects for Both Intercepts and Slopes for Age, Gender, and Marital Status
|
|
table
|
Fixed effects β SE p
Culturea −0.087 0.045 .063
Age −0.076 0.016 <.001
Adultsb 0.000 0.003 .806
Childrenc 0.021 0.003 <.001
Gender (Men vs. Women) 0.065 0.012 <.001
Educationd −0.022 0.003 <.001
COVID‐19 severitye 0.090 0.044 .050
Marital status (Single vs. Married)f −0.132 0.048 .006
Random effects g SD LRT p
Age 0.077 816.83 <.001
Gender (Men vs. Women) 0.054 366.59 <.001
Marital status (Married vs. Single)f 0.238 368.13 <.001
|
|
tr
|
Fixed effects β SE p
|
|
th
|
Fixed effects
|
|
th
|
β
|
|
th
|
SE
|
|
th
|
p
|
|
tr
|
Culturea −0.087 0.045 .063
|
|
td
|
Culturea
|
|
td
|
−0.087
|
|
td
|
0.045
|
|
td
|
.063
|
|
tr
|
Age −0.076 0.016 <.001
|
|
td
|
Age
|
|
td
|
−0.076
|
|
td
|
0.016
|
|
td
|
<.001
|
|
tr
|
Adultsb 0.000 0.003 .806
|
|
td
|
Adultsb
|
|
td
|
0.000
|
|
td
|
0.003
|
|
td
|
.806
|
|
tr
|
Childrenc 0.021 0.003 <.001
|
|
td
|
Childrenc
|
|
td
|
0.021
|
|
td
|
0.003
|
|
td
|
<.001
|
|
tr
|
Gender (Men vs. Women) 0.065 0.012 <.001
|
|
td
|
Gender (Men vs. Women)
|
|
td
|
0.065
|
|
td
|
0.012
|
|
td
|
<.001
|
|
tr
|
Educationd −0.022 0.003 <.001
|
|
td
|
Educationd
|
|
td
|
−0.022
|
|
td
|
0.003
|
|
td
|
<.001
|
|
tr
|
COVID‐19 severitye 0.090 0.044 .050
|
|
td
|
COVID‐19 severitye
|
|
td
|
0.090
|
|
td
|
0.044
|
|
td
|
.050
|
|
tr
|
Marital status (Single vs. Married)f −0.132 0.048 .006
|
|
td
|
Marital status (Single vs. Married)f
|
|
td
|
−0.132
|
|
td
|
0.048
|
|
td
|
.006
|
|
tr
|
Random effects g SD LRT p
|
|
td
|
Random effects g
|
|
td
|
SD
|
|
td
|
LRT
|
|
td
|
p
|
|
tr
|
Age 0.077 816.83 <.001
|
|
td
|
Age
|
|
td
|
0.077
|
|
td
|
816.83
|
|
td
|
<.001
|
|
tr
|
Gender (Men vs. Women) 0.054 366.59 <.001
|
|
td
|
Gender (Men vs. Women)
|
|
td
|
0.054
|
|
td
|
366.59
|
|
td
|
<.001
|
|
tr
|
Marital status (Married vs. Single)f 0.238 368.13 <.001
|
|
td
|
Marital status (Married vs. Single)f
|
|
td
|
0.238
|
|
td
|
368.13
|
|
td
|
<.001
|
|
table-wrap-foot
|
Note
aCulture–Hofstede’s score of a country's individualism (i.e. the higher the number, the more individualistic the country). bAdults–the number of adults staying with the participants during the relative isolation. cChildren–the number of children staying with the participants during the relative isolation. dEducation–education status, with a higher number representing a higher completed education level: 6–PhD/Doctorate; 5–College degree, bachelor, master; 4–Some College, short continuing education or equivalent; 3–Up to 12 years of school; 2–Up to 9 years of school; 1–Up to 6 years of school; 0–None. eCOVID‐19 severity–the number of confirmed cases divided by the number of countries’ citizens. fMarital status–single individuals are coded as 0, and married or cohabiting individuals as 1. gRandom effects–the likelihood ratio test for each random effect, representing variability of stress levels and the association between stress levels and other variables of interest between countries. For the intercept: SD = 0.261, ICC = 0.130. Significant results are bolded.
|
|
title
|
Note
|
|
footnote
|
aCulture–Hofstede’s score of a country's individualism (i.e. the higher the number, the more individualistic the country). bAdults–the number of adults staying with the participants during the relative isolation. cChildren–the number of children staying with the participants during the relative isolation. dEducation–education status, with a higher number representing a higher completed education level: 6–PhD/Doctorate; 5–College degree, bachelor, master; 4–Some College, short continuing education or equivalent; 3–Up to 12 years of school; 2–Up to 9 years of school; 1–Up to 6 years of school; 0–None. eCOVID‐19 severity–the number of confirmed cases divided by the number of countries’ citizens. fMarital status–single individuals are coded as 0, and married or cohabiting individuals as 1. gRandom effects–the likelihood ratio test for each random effect, representing variability of stress levels and the association between stress levels and other variables of interest between countries. For the intercept: SD = 0.261, ICC = 0.130. Significant results are bolded.
|
|
p
|
aCulture–Hofstede’s score of a country's individualism (i.e. the higher the number, the more individualistic the country). bAdults–the number of adults staying with the participants during the relative isolation. cChildren–the number of children staying with the participants during the relative isolation. dEducation–education status, with a higher number representing a higher completed education level: 6–PhD/Doctorate; 5–College degree, bachelor, master; 4–Some College, short continuing education or equivalent; 3–Up to 12 years of school; 2–Up to 9 years of school; 1–Up to 6 years of school; 0–None. eCOVID‐19 severity–the number of confirmed cases divided by the number of countries’ citizens. fMarital status–single individuals are coded as 0, and married or cohabiting individuals as 1. gRandom effects–the likelihood ratio test for each random effect, representing variability of stress levels and the association between stress levels and other variables of interest between countries. For the intercept: SD = 0.261, ICC = 0.130. Significant results are bolded.
|
|
sec
|
DISCUSSION
The results of the present study, based on analyses of data from 53,524 respondents from 26 countries and areas, provide evidence that higher levels of stress are reported by women, single people, people of younger age, people staying with more children, the less educated, and from countries and areas with a more severe COVID‐19 situation. The level of the country's individualism and the number of adults the individual was staying with were not related to stress levels. What is worth highlighting is that the variable which was the most strongly associated with stress levels was marital status, while the standardised coefficients of other variables showed very small effect sizes. Moreover, the analysis of random effects revealed that countries and areas differed in both stress levels and the association between stress levels and marital status, gender, and age. Such differences and their underlying mechanisms warrant further investigations.
H1. The Number of People Together—Do People Living Alone Experience the Highest Level of Stress? No
Even though almost half of our sample was stuck at home (46.1%), while 20.2 per cent lived alone, and 77.5 per cent stayed with one or more adults, we found very little evidence for the association between the number of adults and perceived stress levels (beyond marital status). The results of our study contradict previous findings, which suggested that either living alone or having too many people around can negatively affect one’s well‐being (Fuller et al., 1996). It is possible that living alone may be a preference for many people in our sample, and that they are able to provide for themselves and to pay the rent for a single apartment. Many of those who live alone still maintain high levels of connections with their non‐household relatives or friends (De Vaus & Qu, 2015). Previous studies have shown that people use social media to connect with others (Kowal et al., 2020), and these valuable (and, most importantly, intentional) connections and interactions are not hindered by the current situation, as they can be kept online.
H2. Marital Status—Do Single People Experience More Stress Than Married People? Yes
The results of our study provided support to our hypothesis that married (or cohabiting) individuals experience lower levels of stress than single individuals. These findings fall in line with previous studies highlighting the benefits of being in an intimate relationship (e.g. Braithwaite et al., 2010). Dyadic coping can make up for the potential shortcomings of being in a relationship, and overall, provide more benefits than harm (Merz et al., 2014). Such a protective role may be even more pronounced during difficult times, for instance during financial hardship (Helms et al., 2014). Within the COVID‐19 context, our findings are in line with those of Odriozola‐González et al. (2020), who provided evidence that single (as compared with married) individuals score higher on measures of anxiety, depression, and stress.
On the other hand, our results differ from Wang et al. (2020) and Tian et al. (2020), who found no differences related to marital status on perceived stress, and higher levels of distress among single (vs. married) Chinese citizens, respectively. However, the sample analysed by Wang et al. (2020) may have been underpowered to detect the effect of being in a relationship, as it had 1,210 participants, and tested numerous relationships between various variables. Moreover, Tian et al. (2020) did not control for other relevant factors that could have accounted for the differences (e.g. age, gender). Our study used a high‐powered sample from different countries and areas as well as multivariate analysis techniques to control for several confundings.
H3. Children—Do People with Children Experience More Stress? Yes
Our data confirm the hypothesis that individuals with children report increased stress levels during the COVID‐19 pandemic compared with people living alone or with adults. Moreover, the levels of perceived stress increased with the growing number of children at home. The presence of children during relative isolation or quarantine, probably due to increased home strain and need for childcare, might put demands on adults living with them. In addition, concerns over a child being sick with COVID‐19 have been previously shown to be associated with greater severity of stress and anxiety symptoms in parents (Wang et al., 2020). Having children has also been associated with decreased financial satisfaction (Stanca, 2012). Future studies should examine whether the ages or gender of children living at home moderate the relationship between the number of children and adults’ stress levels.
H4. Age—Do Younger People Experience More Stress than Older People? Yes
The results of our study provided evidence for a negative linear relationship between age and stress levels—younger people experience more stress than older people—contrasting with the existing literature suggesting an inverted U‐shaped relationship between age and stress (Stone et al., 2010). The findings of the present study are in contrast with a recent study of the COVID‐19 outbreak, which showed no association between age and stress levels in a Chinese sample (Wang et al., 2020). Rather, our results support the findings of Limcaoco (2020), who, in a sample of people in quarantine across 41 countries, found age to be significantly related to stress levels. Our evidence warns that we cannot neglect the mental health condition of the younger generations as they seem to be the most stressed during the COVID‐19 outbreak.
H5. Culture—Do People from Individualistic Cultures Experience More Stress? No
In contrast to our hypothesis that people from individualistic cultures might experience more stress than those from collectivistic cultures, we observed no differences in perceived stress levels between countries with varying levels of individualism. Collectivism demands that people feel the obligation and responsibility towards in‐group members. Therefore, due to living in traditional families (Persike & Seiffge‐Krenke, 2012) and feeling more guilt and remorse for seeking help (Kim et al., 2008), people from collectivistic cultures may feel more stressed over their financial burdens than people from individualistic cultures, whereas people from individualistic cultures may treat the current situation as a threat to their need for self‐expression and freedom (Kim & Sherman, 2007) which translates into equal levels of stress.
H6. Gender—Do Women Experience More Stress than Men? Yes
Our data provide evidence that women experience significantly higher stress levels than men during quarantine. These findings support the existing literature on the relationship between gender and stress levels (e.g. Bergdahl & Bergdahl, 2002; Stone et al., 2010), but are contradictory to recent studies on the COVID‐19 outbreak, which found gender to be unrelated to stress levels (Li et al., 2020; Limcaoco et al., 2020; Wang et al., 2020). However, conclusions from these studies ought to be drawn with caution, as they either focused exclusively on samples within one country (e.g. Wang et al., 2020) or included a small number of participants (e.g. Limcaoco et al., 2020). Women are found to be more family‐oriented and provide a supportive role (Tigani et al., 2011); however, during this highly unprecedented time of quarantine, they are overwhelmed by the need to support their families, and such a commitment is beyond their coping capabilities. Moreover, during the relative isolation, it is women who are more likely to not only take care of their work responsibilities, but also take care of the children and home (Minello, 2020).
Limitations
Although the present study contributes to a better understanding of the relationship between the presence of others and perceived stress during the COVID‐19 pandemic, it is not without limitations. First, in order to reach out to participants that are currently under lockdown, it was necessary to use self‐reports and to make use of the individuals’ internet access, although it might limit our findings to a population that is wealthier and has a higher educational background than the general population. Moreover, perceived stress level was tested using only a single self‐reported measure (10‐item Perceived Stress Scale), which warrants some caution. Future studies should also include other measures of stress to expand our knowledge of various aspects of well‐being during the pandemic.
Second, it would not have been possible to use an experimental design, so all our findings are correlative, and we cannot put forward causal relationships. Shortly before the lockdown, people might have chosen to move away from their families because they work in health care and did not want to infect their family members. People that lived on their own before the pandemic might have agreed to isolate with others to combat their loneliness or to help with care work in the family. Those examples are likely rare and might only mildly affect our findings. Even though we were not able to control for confounding variables, some pre‐existing factors, such as age and gender, may at least indicate a direction of effect.
Third, longitudinal effects of the pandemic were not measured. We present the data of the first week when the COVIDiSTRESS Global Survey went online, and participants were only allowed to participate once. It would be beneficial for future studies to measure the long‐term effects and the development of stress within a sample.
Fourth, the pandemic had been spreading in countries and areas at various rates during data collection. Therefore, people from different countries and different regions within a country possibly were under different levels of threat as well as different time spans of exposure. It is important for future studies to further investigate whether there are dose–response relationships between stress and the level of threat in the given countries.
|
|
title
|
DISCUSSION
|
|
p
|
The results of the present study, based on analyses of data from 53,524 respondents from 26 countries and areas, provide evidence that higher levels of stress are reported by women, single people, people of younger age, people staying with more children, the less educated, and from countries and areas with a more severe COVID‐19 situation. The level of the country's individualism and the number of adults the individual was staying with were not related to stress levels. What is worth highlighting is that the variable which was the most strongly associated with stress levels was marital status, while the standardised coefficients of other variables showed very small effect sizes. Moreover, the analysis of random effects revealed that countries and areas differed in both stress levels and the association between stress levels and marital status, gender, and age. Such differences and their underlying mechanisms warrant further investigations.
|
|
sec
|
H1. The Number of People Together—Do People Living Alone Experience the Highest Level of Stress? No
Even though almost half of our sample was stuck at home (46.1%), while 20.2 per cent lived alone, and 77.5 per cent stayed with one or more adults, we found very little evidence for the association between the number of adults and perceived stress levels (beyond marital status). The results of our study contradict previous findings, which suggested that either living alone or having too many people around can negatively affect one’s well‐being (Fuller et al., 1996). It is possible that living alone may be a preference for many people in our sample, and that they are able to provide for themselves and to pay the rent for a single apartment. Many of those who live alone still maintain high levels of connections with their non‐household relatives or friends (De Vaus & Qu, 2015). Previous studies have shown that people use social media to connect with others (Kowal et al., 2020), and these valuable (and, most importantly, intentional) connections and interactions are not hindered by the current situation, as they can be kept online.
|
|
title
|
H1. The Number of People Together—Do People Living Alone Experience the Highest Level of Stress? No
|
|
p
|
Even though almost half of our sample was stuck at home (46.1%), while 20.2 per cent lived alone, and 77.5 per cent stayed with one or more adults, we found very little evidence for the association between the number of adults and perceived stress levels (beyond marital status). The results of our study contradict previous findings, which suggested that either living alone or having too many people around can negatively affect one’s well‐being (Fuller et al., 1996). It is possible that living alone may be a preference for many people in our sample, and that they are able to provide for themselves and to pay the rent for a single apartment. Many of those who live alone still maintain high levels of connections with their non‐household relatives or friends (De Vaus & Qu, 2015). Previous studies have shown that people use social media to connect with others (Kowal et al., 2020), and these valuable (and, most importantly, intentional) connections and interactions are not hindered by the current situation, as they can be kept online.
|
|
sec
|
H2. Marital Status—Do Single People Experience More Stress Than Married People? Yes
The results of our study provided support to our hypothesis that married (or cohabiting) individuals experience lower levels of stress than single individuals. These findings fall in line with previous studies highlighting the benefits of being in an intimate relationship (e.g. Braithwaite et al., 2010). Dyadic coping can make up for the potential shortcomings of being in a relationship, and overall, provide more benefits than harm (Merz et al., 2014). Such a protective role may be even more pronounced during difficult times, for instance during financial hardship (Helms et al., 2014). Within the COVID‐19 context, our findings are in line with those of Odriozola‐González et al. (2020), who provided evidence that single (as compared with married) individuals score higher on measures of anxiety, depression, and stress.
On the other hand, our results differ from Wang et al. (2020) and Tian et al. (2020), who found no differences related to marital status on perceived stress, and higher levels of distress among single (vs. married) Chinese citizens, respectively. However, the sample analysed by Wang et al. (2020) may have been underpowered to detect the effect of being in a relationship, as it had 1,210 participants, and tested numerous relationships between various variables. Moreover, Tian et al. (2020) did not control for other relevant factors that could have accounted for the differences (e.g. age, gender). Our study used a high‐powered sample from different countries and areas as well as multivariate analysis techniques to control for several confundings.
|
|
title
|
H2. Marital Status—Do Single People Experience More Stress Than Married People? Yes
|
|
p
|
The results of our study provided support to our hypothesis that married (or cohabiting) individuals experience lower levels of stress than single individuals. These findings fall in line with previous studies highlighting the benefits of being in an intimate relationship (e.g. Braithwaite et al., 2010). Dyadic coping can make up for the potential shortcomings of being in a relationship, and overall, provide more benefits than harm (Merz et al., 2014). Such a protective role may be even more pronounced during difficult times, for instance during financial hardship (Helms et al., 2014). Within the COVID‐19 context, our findings are in line with those of Odriozola‐González et al. (2020), who provided evidence that single (as compared with married) individuals score higher on measures of anxiety, depression, and stress.
|
|
p
|
On the other hand, our results differ from Wang et al. (2020) and Tian et al. (2020), who found no differences related to marital status on perceived stress, and higher levels of distress among single (vs. married) Chinese citizens, respectively. However, the sample analysed by Wang et al. (2020) may have been underpowered to detect the effect of being in a relationship, as it had 1,210 participants, and tested numerous relationships between various variables. Moreover, Tian et al. (2020) did not control for other relevant factors that could have accounted for the differences (e.g. age, gender). Our study used a high‐powered sample from different countries and areas as well as multivariate analysis techniques to control for several confundings.
|
|
sec
|
H3. Children—Do People with Children Experience More Stress? Yes
Our data confirm the hypothesis that individuals with children report increased stress levels during the COVID‐19 pandemic compared with people living alone or with adults. Moreover, the levels of perceived stress increased with the growing number of children at home. The presence of children during relative isolation or quarantine, probably due to increased home strain and need for childcare, might put demands on adults living with them. In addition, concerns over a child being sick with COVID‐19 have been previously shown to be associated with greater severity of stress and anxiety symptoms in parents (Wang et al., 2020). Having children has also been associated with decreased financial satisfaction (Stanca, 2012). Future studies should examine whether the ages or gender of children living at home moderate the relationship between the number of children and adults’ stress levels.
|
|
title
|
H3. Children—Do People with Children Experience More Stress? Yes
|
|
p
|
Our data confirm the hypothesis that individuals with children report increased stress levels during the COVID‐19 pandemic compared with people living alone or with adults. Moreover, the levels of perceived stress increased with the growing number of children at home. The presence of children during relative isolation or quarantine, probably due to increased home strain and need for childcare, might put demands on adults living with them. In addition, concerns over a child being sick with COVID‐19 have been previously shown to be associated with greater severity of stress and anxiety symptoms in parents (Wang et al., 2020). Having children has also been associated with decreased financial satisfaction (Stanca, 2012). Future studies should examine whether the ages or gender of children living at home moderate the relationship between the number of children and adults’ stress levels.
|
|
sec
|
H4. Age—Do Younger People Experience More Stress than Older People? Yes
The results of our study provided evidence for a negative linear relationship between age and stress levels—younger people experience more stress than older people—contrasting with the existing literature suggesting an inverted U‐shaped relationship between age and stress (Stone et al., 2010). The findings of the present study are in contrast with a recent study of the COVID‐19 outbreak, which showed no association between age and stress levels in a Chinese sample (Wang et al., 2020). Rather, our results support the findings of Limcaoco (2020), who, in a sample of people in quarantine across 41 countries, found age to be significantly related to stress levels. Our evidence warns that we cannot neglect the mental health condition of the younger generations as they seem to be the most stressed during the COVID‐19 outbreak.
|
|
title
|
H4. Age—Do Younger People Experience More Stress than Older People? Yes
|
|
p
|
The results of our study provided evidence for a negative linear relationship between age and stress levels—younger people experience more stress than older people—contrasting with the existing literature suggesting an inverted U‐shaped relationship between age and stress (Stone et al., 2010). The findings of the present study are in contrast with a recent study of the COVID‐19 outbreak, which showed no association between age and stress levels in a Chinese sample (Wang et al., 2020). Rather, our results support the findings of Limcaoco (2020), who, in a sample of people in quarantine across 41 countries, found age to be significantly related to stress levels. Our evidence warns that we cannot neglect the mental health condition of the younger generations as they seem to be the most stressed during the COVID‐19 outbreak.
|
|
sec
|
H5. Culture—Do People from Individualistic Cultures Experience More Stress? No
In contrast to our hypothesis that people from individualistic cultures might experience more stress than those from collectivistic cultures, we observed no differences in perceived stress levels between countries with varying levels of individualism. Collectivism demands that people feel the obligation and responsibility towards in‐group members. Therefore, due to living in traditional families (Persike & Seiffge‐Krenke, 2012) and feeling more guilt and remorse for seeking help (Kim et al., 2008), people from collectivistic cultures may feel more stressed over their financial burdens than people from individualistic cultures, whereas people from individualistic cultures may treat the current situation as a threat to their need for self‐expression and freedom (Kim & Sherman, 2007) which translates into equal levels of stress.
|
|
title
|
H5. Culture—Do People from Individualistic Cultures Experience More Stress? No
|
|
p
|
In contrast to our hypothesis that people from individualistic cultures might experience more stress than those from collectivistic cultures, we observed no differences in perceived stress levels between countries with varying levels of individualism. Collectivism demands that people feel the obligation and responsibility towards in‐group members. Therefore, due to living in traditional families (Persike & Seiffge‐Krenke, 2012) and feeling more guilt and remorse for seeking help (Kim et al., 2008), people from collectivistic cultures may feel more stressed over their financial burdens than people from individualistic cultures, whereas people from individualistic cultures may treat the current situation as a threat to their need for self‐expression and freedom (Kim & Sherman, 2007) which translates into equal levels of stress.
|
|
sec
|
H6. Gender—Do Women Experience More Stress than Men? Yes
Our data provide evidence that women experience significantly higher stress levels than men during quarantine. These findings support the existing literature on the relationship between gender and stress levels (e.g. Bergdahl & Bergdahl, 2002; Stone et al., 2010), but are contradictory to recent studies on the COVID‐19 outbreak, which found gender to be unrelated to stress levels (Li et al., 2020; Limcaoco et al., 2020; Wang et al., 2020). However, conclusions from these studies ought to be drawn with caution, as they either focused exclusively on samples within one country (e.g. Wang et al., 2020) or included a small number of participants (e.g. Limcaoco et al., 2020). Women are found to be more family‐oriented and provide a supportive role (Tigani et al., 2011); however, during this highly unprecedented time of quarantine, they are overwhelmed by the need to support their families, and such a commitment is beyond their coping capabilities. Moreover, during the relative isolation, it is women who are more likely to not only take care of their work responsibilities, but also take care of the children and home (Minello, 2020).
|
|
title
|
H6. Gender—Do Women Experience More Stress than Men? Yes
|
|
p
|
Our data provide evidence that women experience significantly higher stress levels than men during quarantine. These findings support the existing literature on the relationship between gender and stress levels (e.g. Bergdahl & Bergdahl, 2002; Stone et al., 2010), but are contradictory to recent studies on the COVID‐19 outbreak, which found gender to be unrelated to stress levels (Li et al., 2020; Limcaoco et al., 2020; Wang et al., 2020). However, conclusions from these studies ought to be drawn with caution, as they either focused exclusively on samples within one country (e.g. Wang et al., 2020) or included a small number of participants (e.g. Limcaoco et al., 2020). Women are found to be more family‐oriented and provide a supportive role (Tigani et al., 2011); however, during this highly unprecedented time of quarantine, they are overwhelmed by the need to support their families, and such a commitment is beyond their coping capabilities. Moreover, during the relative isolation, it is women who are more likely to not only take care of their work responsibilities, but also take care of the children and home (Minello, 2020).
|
|
sec
|
Limitations
Although the present study contributes to a better understanding of the relationship between the presence of others and perceived stress during the COVID‐19 pandemic, it is not without limitations. First, in order to reach out to participants that are currently under lockdown, it was necessary to use self‐reports and to make use of the individuals’ internet access, although it might limit our findings to a population that is wealthier and has a higher educational background than the general population. Moreover, perceived stress level was tested using only a single self‐reported measure (10‐item Perceived Stress Scale), which warrants some caution. Future studies should also include other measures of stress to expand our knowledge of various aspects of well‐being during the pandemic.
Second, it would not have been possible to use an experimental design, so all our findings are correlative, and we cannot put forward causal relationships. Shortly before the lockdown, people might have chosen to move away from their families because they work in health care and did not want to infect their family members. People that lived on their own before the pandemic might have agreed to isolate with others to combat their loneliness or to help with care work in the family. Those examples are likely rare and might only mildly affect our findings. Even though we were not able to control for confounding variables, some pre‐existing factors, such as age and gender, may at least indicate a direction of effect.
Third, longitudinal effects of the pandemic were not measured. We present the data of the first week when the COVIDiSTRESS Global Survey went online, and participants were only allowed to participate once. It would be beneficial for future studies to measure the long‐term effects and the development of stress within a sample.
Fourth, the pandemic had been spreading in countries and areas at various rates during data collection. Therefore, people from different countries and different regions within a country possibly were under different levels of threat as well as different time spans of exposure. It is important for future studies to further investigate whether there are dose–response relationships between stress and the level of threat in the given countries.
|
|
title
|
Limitations
|
|
p
|
Although the present study contributes to a better understanding of the relationship between the presence of others and perceived stress during the COVID‐19 pandemic, it is not without limitations. First, in order to reach out to participants that are currently under lockdown, it was necessary to use self‐reports and to make use of the individuals’ internet access, although it might limit our findings to a population that is wealthier and has a higher educational background than the general population. Moreover, perceived stress level was tested using only a single self‐reported measure (10‐item Perceived Stress Scale), which warrants some caution. Future studies should also include other measures of stress to expand our knowledge of various aspects of well‐being during the pandemic.
|
|
p
|
Second, it would not have been possible to use an experimental design, so all our findings are correlative, and we cannot put forward causal relationships. Shortly before the lockdown, people might have chosen to move away from their families because they work in health care and did not want to infect their family members. People that lived on their own before the pandemic might have agreed to isolate with others to combat their loneliness or to help with care work in the family. Those examples are likely rare and might only mildly affect our findings. Even though we were not able to control for confounding variables, some pre‐existing factors, such as age and gender, may at least indicate a direction of effect.
|
|
p
|
Third, longitudinal effects of the pandemic were not measured. We present the data of the first week when the COVIDiSTRESS Global Survey went online, and participants were only allowed to participate once. It would be beneficial for future studies to measure the long‐term effects and the development of stress within a sample.
|
|
p
|
Fourth, the pandemic had been spreading in countries and areas at various rates during data collection. Therefore, people from different countries and different regions within a country possibly were under different levels of threat as well as different time spans of exposure. It is important for future studies to further investigate whether there are dose–response relationships between stress and the level of threat in the given countries.
|
|
sec
|
CONCLUSIONS
The present study contributes to a better understanding of the relationship between the presence of others and perceived stress during the COVID‐19 lockdown. Moreover, our results add to the limited literature on people during the pandemic. These insights highlight the need to focus on neglected groups (i.e. women, younger, single people, with children, and with lower education status) both in research and in public health interventions, as they may be especially susceptible to experiencing elevated levels of stress, with their psychological well‐being being compromised.
|
|
title
|
CONCLUSIONS
|
|
p
|
The present study contributes to a better understanding of the relationship between the presence of others and perceived stress during the COVID‐19 lockdown. Moreover, our results add to the limited literature on people during the pandemic. These insights highlight the need to focus on neglected groups (i.e. women, younger, single people, with children, and with lower education status) both in research and in public health interventions, as they may be especially susceptible to experiencing elevated levels of stress, with their psychological well‐being being compromised.
|
|
sec
|
Conflict of interest
The authors declare no conflict of interest.
|
|
title
|
Conflict of interest
|
|
p
|
The authors declare no conflict of interest.
|
|
sec
|
Supporting information
Table S1. Descriptive statistics of gender, marital status, age, the number of adults and children under 12 individuals were staying with during the data collection, and country’s score on Hofstede individualism scale.
Table S2. Means, standard deviations and Cronbach’s alphas of the Perceived Stress Scale across countries.
Table S3. Results of a Multilevel Model, With Participants Nested Within Countries, and Random Effects for Both Intercepts and Slopes for age, gender, isolation, and marital status with variables in a raw form.
Table S4. Responses on the question: ‘What best describes your current situation?’, with respect to countries.
Table S5. The number of citizens across countries and the number of confirmed cases across days of data collection (with an early launch in Denmark – March 28 and 29).
Table S6. The number of participants with the given education level (the highest completed education) with respect to countries.
Figure S1. Three scatter plots, depicting relationships between perceived stress and the number of people (A), children (B), and adults (C) an individual was staying with during the relative isolation.
Click here for additional data file.
|
|
title
|
Supporting information
|
|
caption
|
Table S1. Descriptive statistics of gender, marital status, age, the number of adults and children under 12 individuals were staying with during the data collection, and country’s score on Hofstede individualism scale.
Table S2. Means, standard deviations and Cronbach’s alphas of the Perceived Stress Scale across countries.
Table S3. Results of a Multilevel Model, With Participants Nested Within Countries, and Random Effects for Both Intercepts and Slopes for age, gender, isolation, and marital status with variables in a raw form.
Table S4. Responses on the question: ‘What best describes your current situation?’, with respect to countries.
Table S5. The number of citizens across countries and the number of confirmed cases across days of data collection (with an early launch in Denmark – March 28 and 29).
Table S6. The number of participants with the given education level (the highest completed education) with respect to countries.
Figure S1. Three scatter plots, depicting relationships between perceived stress and the number of people (A), children (B), and adults (C) an individual was staying with during the relative isolation.
|
|
p
|
Table S1. Descriptive statistics of gender, marital status, age, the number of adults and children under 12 individuals were staying with during the data collection, and country’s score on Hofstede individualism scale.
|
|
p
|
Table S2. Means, standard deviations and Cronbach’s alphas of the Perceived Stress Scale across countries.
|
|
p
|
Table S3. Results of a Multilevel Model, With Participants Nested Within Countries, and Random Effects for Both Intercepts and Slopes for age, gender, isolation, and marital status with variables in a raw form.
|
|
p
|
Table S4. Responses on the question: ‘What best describes your current situation?’, with respect to countries.
|
|
p
|
Table S5. The number of citizens across countries and the number of confirmed cases across days of data collection (with an early launch in Denmark – March 28 and 29).
|
|
p
|
Table S6. The number of participants with the given education level (the highest completed education) with respect to countries.
|
|
p
|
Figure S1. Three scatter plots, depicting relationships between perceived stress and the number of people (A), children (B), and adults (C) an individual was staying with during the relative isolation.
|
|
caption
|
Click here for additional data file.
|
|
p
|
Click here for additional data file.
|
