Predictors of COVID-19 anxiety in UK university students

ABSTRACT The current research aimed to evaluate UK student mental health during the first 4 months of the COVID-19 pandemic. Specifically, we were interested in exploring factors that contribute to students’ anxiety levels about COVID-19. Demographics, mental health symptomatology and well-being-related variables were tested as predictors of COVID-19 anxiety. A cross-sectional online survey was completed by 1,164 UK university students (71.8% females). Measures included self-reported data on stress, social phobia, anxiety, depression, psychotic-like experiences, hopefulness, group membership, social identity, belonging, loneliness, COVID-19 related variables and demographics. The majority of participants were between 18 and 24 years old (92.2%), White British (57.9%) and in the first or second year of their course study (63.12%). A series of multiple linear regressions revealed that being female, experiencing more stress and anxiety, and greater worry about COVID-19 and its effects on social relationships were significant in explaining students’ pandemic-related anxiety. Students’ level of pandemic-related anxiety was independent of ethnicity, socioeconomic background and pre-existing mental health problems. Worrying about the future was the most common pandemic-related stressor, but it was not a significant predictor of COVID-19 anxiety. Our findings are in line with previous findings that females and students experiencing more stress have been disproportionately affected by the COVID-19 pandemic. Our findings could inform the delivery of targeted stress-management interventions that might prove beneficial for student wellbeing.


Introduction
The spread of the SARS-CoV-2 (COVID-19) was officially declared a pandemic on the 11 th of March 2020. By then there were 118,000 cases in 114 countries, and 4,291 deaths globally (WHO, 2020). In the UK, movement restrictions were imposed on the 23rd of March 2020 as part of a national lockdown intended to mitigate virus transmission. By that time, more than 12,000 cases and 400 deaths due to the virus were reported in the UK (GOV.UK 2021). For many, the lockdown was a period of social isolation that had detrimental effects on mental health. For example, in the UK, during the first four to 6 weeks of the lockdown, depression, stress and anxiety among working adults were significantly worse than was typical for the general population (Jia et al. 2020). Younger adults were disproportionately affected, as their mental health significantly deteriorated during to the pandemic (Banks and Xu 2020).
Early adulthood is a key risk period for the onset of mental health problems (Kim-Cohen et al. 2003;Kessler et al. 2005) and, even before the pandemic, there was evidence that mental health problems amongst students in particular were rising (Broglia, Millings, and Barkham 2018). Undergraduate students anxiety and depression levels have increased across the globe in response to the pandemic from Saudi Arabia (Elsharkawy and Abdelaziz 2020) and Turkey (Cam, Ustuner Top, and Kuzlu Ayyildiz 2021) to China (Feng, Zhang, and Ho 2021) and the US (Rogers, Ha, and Ockey 2021). Excessive worry about COVID-19 has also been linked to insomnia (Scotta, Cortez, and Miranda 2020), whilst depressive symptomatology during the pandemic has been linked to lower academic engagement (Kecojevic et al. 2020). These findings are rather unsurprising as universities were affected by the global pandemic and ensuing lockdown; by March 2020 teaching in many UK universities had been disrupted and social and academic aspects of students' lives had undergone substantial change.
The aim of the current study was to explore the role of current mental health symptomology, psychosocial and demographic variables as well as COVID-19 related concerns as risk factors for increased anxiety about COVID-19 in undergraduate students during the first few months of the pandemic in the UK. Below, we review literature on the effects of stress, depression, generalised anxiety, psychotic experiences, loneliness, hope, social identity, belonging, group membership, pre-existing mental health problems, socioeconomic status (SES), gender and sexual orientation on student wellbeing during the pandemic.
There is a lack of published data regarding the current mental health and wellbeing of undergraduate students in the UK during the pandemic. Students' stress levels and sedentary behaviour (Savage et al. 2020) as well as depression (Evans et al. 2021) increased during the lockdown, though the modest sample sizes (214 and 254, respectively) limit the generalisability of the results. Evans et al. (2021) found a third of their sample reached the threshold for clinically depressed, compared to 15% at baseline. Brazilian (Fernandez et al. 2021) and Israeli (Savitsky et al. 2020) students showed especially elevated scores in moderate and severe generalised anxiety compared to before the pandemic. Psychotic experiences among Slovakian students during the pandemic have also been investigated with higher frequencies occurring at earlier stages of the pandemic than a few months later (Hajdúk et al. 2020).
Loneliness has been identified as a significant predictor of mental distress in UK students (McIntyre et al. 2018). Increased feelings of loneliness have been associated with COVID-19 anxiety in a sample of Turkish college students, acting as a mediator between COVID-19 anxiety and negative repetitive thinking about COVID-19 (Arslan, Yıldırım, and Aytaç 2020). Conversely, higher levels of hopefulness are associated with lower COVID-19 anxiety and act as a protective factor against the negative effects of COVID-19 anxiety on students' subjective wellbeing (Genç and Arslan 2021). Despite their direct associations with students' wellbeing, the effects of loneliness and hope have not yet been investigated among UK students during the pandemic.
Another psychosocial trait that might have implications for COVID-19 anxiety in UK students is social identification. Identifying as a member of a social group has been found to be beneficial for individuals' mental health and buffer the negative effects of a life transition on wellbeing (Iyer et al. 2009). Moreover, Vignoles et al. (2021) demonstrated that social identification was a negative predictor of mental health, depression and anxiety during the first months of the lockdown in a UK community sample. Given the beneficial effects of social identity on well-being, it is important to explore whether social identity is a predictor of UK students' COVID-19 anxiety during the pandemic. Closely aligned with social identity is the sense of belonging, which is regarded a product of strong social identification that facilitates social support and is in turn protective towards positive mental health (Haslam et al. 2016). The feeling of group belonging is significantly associated with students' stress and depression levels (Stebleton, Soria, and Huesman 2014) and adolescent mental health and wellbeing indicators (Arslan, Allena, & Ryan, 2020). Group membership has also been shown to be beneficial for wellbeing during the pandemic (Alcover et al. 2020). Thus, it is possible that, like social identity, a higher sense of belonging and group membership might be protective agents for UK students' COVID-19 anxiety.
In addition to specific current mental health and psychosocial traits, adults with pre-existing mental health problems seem to be more susceptible to further deterioration after a disaster (Goldmann and Galea 2014). Therefore, it is not surprising that during the pandemic their condition is likely to worsen (Murphy et al. 2021). For example, individuals with pre-existing (past year) anxietyrelated mental health conditions in the USA and Canada experience more COVID-19-related stress compared to those with no mental health disorder (Asmundson et al. 2020).
Certain demographic factors are of importance too when it comes to COVID-19-related stress. For example, differences in the way college students have been psychologically impacted by the pandemic have been observed in US college students as it was found that females and LGBTQI+ students have been unequally stressed by the pandemic compared to their counterparts (Hoyt et al. 2021). Previous research has found that lower socioeconomic status (SES) is associated with higher rates of distress, mental health problems, and struggles after a traumatic event (Baum, Garofalo, and Yali 1999). More recent data from Hong Kong suggest that high SES is associated with greater wellbeing among students in the beginning of the pandemic (Amoah et al. 2021).
Thus, students' wellbeing has been severely impacted by the pandemic with recent evidence suggesting that there is a plethora of variables contributing to this. To address the impact of mental health, psychosocial and demographic variables on students' COVID-19-related anxiety we collected data from >1000 students at a UK university on anxiety, depression, psychotic symptoms, social phobia, stress, belonging, group membership, hope, loneliness, social identity, and demographic variables. We also created items reflecting common worries regarding COVID-19 and a single item to quantify COVID-19 anxiety. We categorised variables into four clusters and regressed the single COVID-19 anxiety item on each of these clusters. Then we regressed the COVID-19 anxiety item onto significant variables in a final model.

Participant characteristics
A total of 1,164 students completed the survey, with the majority (93.3%, n = 1085) studying towards an undergraduate degree course at the institution within which the research was being conducted. Most participants were aged 18-24 years (92.2%, n = 962), identified as female (71.8%, n = 752), and more than half described their ethnicity as White British (57.9%, n = 606). The majority were UK citizens (76.3%, n = 798) with English as their first language (78.5%, n = 821). More than half reported living at university away from home (67.4%, n = 784) and in total 720 reported having a mental health problem with or without a professional diagnosis (65.5%). Full participant demographic and academic characteristics are displayed in Supplementary Material 1.

Procedure
Data were collected through an UK online survey administered as part of a broader longitudinal investigation into student mental health during transitions to and within the university. The survey was piloted and took between 20 and 30 minutes to complete. Following the UK government's implementation of the first nationwide 'lockdown' in response to the global pandemic on 23 March 2020, additional survey questions were added to assess students' concerns and anxiety around COVID-19. Data for the present study were collected between 4 th April and 31 July 2020. At the point at which recruitment began, there were 57,111 reported cases of COVID-19 in the UK and 7,753 deaths, rising to 305,131 reported cases and 41,358 deaths on the day on which recruitment ceased (https://coronavirus.data.gov.uk/details/cases). Throughout the recruitment period, UK universities were operating undergraduate degree courses online according to government guidelines.
The survey was advertised using social media platforms to students of any academic discipline and registered with any UK university, and directly to students registered at the institution within which the research was being conducted through an online newsletter. Participants were offered the opportunity to enter a prize draw with the chance of winning £100 or £5. The study received ethical approval from the Science and Technology Cross School Research Ethics Committee of the University of Sussex (ER/OC206/1). Written informed consent was obtained from all participants.

COVID-19 anxiety.
A single question was created to assess COVID-19 anxiety. Participants were asked to indicate on a scale of 0 = not at all anxious to 100 = extremely anxious their overall level of anxiety in relation to the COVID-19 pandemic.

COVID-19 characteristics. Two items asking respondents about their experiences in relation to
COVID-19 were created. Participants indicated whether they knew anybody who had contracted COVID-19 and whether the pandemic had resulted in any changes to who they were living with.

COVID-19 concerns.
Participants were asked to select their concerns from a list of 12 items reflecting an array of daily life domains (see Supplementary Material 2). Example items included Being unwell myself with COVID-19, and Financial worries. We conducted an exploratory factor analysis of the 12 items, identifying three factors: disease concerns; relational concerns; and vocational/lifestyle concerns. A scree plot indicated a break at three factors. A four-factor solution achieved superior fit indices but resulted in a negative residual variance and had low face and factorial validity: the fourth factor was a factor comprising only two cross-loading indicators. Therefore, a three-factor solution was accepted; χ 2 = 230.66 (33), p < .00001, CFI = 0.92, TFI = 0.85, RMSEA = 0.07, SRMR = 0.06.

Current mental health measures
Cronbach's α showed good reliability within our data for all the validated scales listed below (Table 2).
Anxiety. Assessed using the Generalised Anxiety Disorder (GAD-7) questionnaire (Spitzer, Kroenke, & Williams, 2006), a seven-item self-report scale evaluating the frequency with which respondents have experienced the primary symptoms of GAD over the past 2 weeks (i.e. feeling nervous, anxious, or on edge, and worrying too much about different things). Items are rated on a 4-point Likert scale ranging from 0 = not at all to 3 = nearly every day. Scores range from 0 to 21, higher scores indicating more severe symptomatology. An eighth item asks participants to rate how difficult anxiety has made carrying out tasks at work and at home and getting along with others (0 = not at all difficult to 3 = extremely difficult). GAD-7 does not necessarily function well as a clinical diagnostic tool for GAD but can detect anxiety symptoms associated with a broad range of anxiety disorders including GAD, panic, social anxiety and post-traumatic stress disorders.
Depression. The Patient Health Questionnaire (PHQ-9; Kroenke, Spitzer, and Williams 2001) was administered to assess depression. Itis a 9-item self-report measure designed to assess depression severity. Participants rate on a four-point Likert scale (0 = not at all to 3 = nearly every day) how often they felt sad, down, hopeless, depressed, and bad about themselves over the past 2 weeks. PHQ-9 scores of 5, 10, 15, and 20 represent mild, moderate, moderately severe, and severe depression, respectively (Kroenke et al. 2010).

Psychotic-like experiences.
Proneness to psychotic-like experiences (PLEs) was assessed using the Community Assessment of Psychic Experiences-Positive Scale (Capra et al. 2017). The CAPE-P15 consists of 15 items evaluating the frequency of PLEs and associated distress in three domains: persecutory ideation, bizarre experiences, and perceptual abnormalities. The scale is based on an earlier 20-item version of the CAPE-P that has good internal validity and reliability for detecting early psychosis-like positive symptomatology in undergraduate student samples (Capra et al. 2017). Participants rated on a four-point Likert scale how often they had experienced PLEs over the past 3 months (0 = never to 3 = nearly always). Weighted scores for individuals were produced by dividing the total score by the number of items completed, and a cut-off of ≥1.47 was employed to detect students reporting PLEs. This cut-off provides 77% sensitivity and specificity of 58% (Bukenaite et al. 2017).
Social phobia. Participants completed the Social Phobia Scale (Peters et al. 2012), a six-item selfreport scale assessing anxiety associated with the performance of various tasks whilst being scrutinised by others (e.g. working, eating, drinking, writing, using public toilets). Whilst the SPS originally consisted of 20-items, the six-item version administered for the present study has demonstrated good reliability. Participants respond using a five-point Likert scale, ranging from 0 = not at all characteristic or true of me to 4 = extremely characteristic or true of me.
Stress. Stress was measured using the Perceived Stress Scale (Cohen and Williamson 1988). The PSS-10 is a widely used instrument for measuring the extent to which life situations are perceived as stressful. The ten-item self-report scale asks participants to rate how unpredictable, uncontrollable, and overloaded they find their lives to be on a 5-point Likert scale (0 = never to 4 = very often

Psychosocial measures
Cronbach's α showed good reliability within our data for all the validated scales listed below (Table 2).
Belonging. The sense of belonging at university was assessed using three items adapted from Walton and Cohen (2007). Participants rated the extent to which they agreed with each item using a 7-point Likert scale (1 = Strongly disagree to 7 = Definitely). The items were as follows: In general, I really feel like I belong at university; In general, I feel people at university accept me; and In general, people at university like me.
Group membership. The Exeter Identity Transition Scales (EXITS; Haslam et al. 2008) was used to assess group membership. Participants completed items evaluating changes in identity and group affiliations over time using a 7-point Likert scale (1 = strongly disagree to 7 = definitely agree). EXITS has been used to explore group membership and life satisfaction in older adults after stroke (Haslam et al. 2008). For the present study, three adapted subscales of four items assessed participants': membership in multiple groups before university; the extent to which students maintained their old group memberships after joining university; and experiences of belonging to new groups whilst at university.
Hope. Hope was measured using the Domain Specific Hope Scale (Sympson 1999 Loneliness. The UCLA Loneliness Scale (Russell, Peplau, and Cutrona 1980) assesses how much a person feels lonely, cut-off, or separated from others, by evaluating the discrepancy between the social interactions participants report engaging in compared to those they would like to engage in. The UCL-8 contains eight items, including two that are reverse-scored (I am an outgoing person and I can find companionship when I want it. Each item is associated with a four-level frequency score, with answer choices of 0 = never, 1 = rarely, 2 = sometimes, and 3 = always. Total scores range from 0 to 24, with higher scores indicating more feelings of loneliness. Social identity. Social identification with university was assessed using one item, adapted from Postmes, Haslam, and Jans (2013). Participants responded to the statement I identify with my university on a seven-point Likert scale (1 = Strongly disagree to 7 = Strongly agree).

Demographic and academic variables
A full list of the demographic and academic variables participants reported is provided in Supplementary Material 1.

Socioeconomic background (IMD).
We asked participants for their home postcodes and then we used those to retrieve the corresponding Index of Multiple Deprivation (IMD) for each postcode as provided in gov.uk. IMD results were provided in deciles from 1 to 10 with decile 1 signifying the most-deprived areas to 10 signifying the least-deprived areas. Finally, we grouped the IMDs in three groups; high IMD (decile 1-3), middle IMD (decile 4-7) and low IMD (decile 8-10).

Differing Likert scales and anchoring
We used previously validated measures for the constructs of interest (see above) and, therefore, retained the original Likert response scales for each individual set of items, which differ from one another. By varying the response scales within a single questionnaire battery we aimed to ensure that respondents did not feel compelled to be consistent and, consequently, select either the same or a nearby response category, irrespective of fit; also known as anchoring (Lyu and Bolt 2022). Anchoring should diminish when the item content or response-scale changes substantially, in which case the respondent is forced to adopt a different orientation in responding to the item (Gehlbach and Barge 2012). Thus, our use of different Likert scales may have reduced anchoring and led participants to answer more accurately.

Data analysis
All analyses were conducted using IBM SPSS 26. Correlation showed that all three factors of COVID-19 concerns were significantly associated (p < .001). All predictors were checked for collinearity using Pearson's bivariate correlations (Supplementary Material 3). Then a series of forced entry multipleregression analyses were conducted incorporating COVID-19 concerns, demographic, mental health, and psychosocial variables as predictors of COVID-19 anxiety in four separate models, respectively. Significant predictors from the analyses were subsequently entered into a final regression model, to examine which variables would continue to predict COVID-19 anxiety. We followed this analytic strategy to avoid complex models that include large number of predictors and possible multicollinearity issues. Differences were observed when analyses were run with missing data deleted pairwise compared to listwise (Supplementary Material 4, 5). Where necessary, a t-test was used to confirm the significance of predictors in the final model.

Mental health and psychosocial factors
Means, standard deviations, minimum and maximum values and Cronbach's α for all main study variables are shown in Table 2. Correlations for measures of mental health and psychosocial factors are shown in Supplementary Material 3.

Predictors of COVID-19 anxiety
Four different multiple-regression models using COVID-19 as an outcome measure were implemented. The first model used COVID-19 related variables as predictors (Table 3). In this model, COVID-19 disease concerns, relational concerns, vocational/lifestyle concerns were significant predictors of COVID-19 anxiety, whereas COVID-19 connection and living situation were not.
In the second multiple-regression model, demographic variables were used as predictors (Table 4). In this case, only gender was a significant predictor of COVID-19 anxiety, whilst index of multiple deprivation (IMD), ethnicity, sexuality, year of study and pre-existing mental health difficulties were not. Participants identifying as female were more likely to experience higher levels of COVID-19 anxiety than their male counterparts. The third multiple-regression model incorporated current mental health variables (Table 5). In this model perceived stress and generalised anxiety emerged as significant predictors of COVID-19 anxiety, whereas social phobia, depression, and psychotic experiences were not.
The fourth model incorporated psychosocial variables as predictors of COVID-19 anxiety (Table 6). In this model, only group membership before university and loneliness were significant predictors of COVID-19 anxiety, whilst hope (both academic and social), maintenance of old group membership, new group membership, social identity and belonging were not significant predictors. Differences were observed when we compared models in which missing data was deleted pairwise (Tables 3-6) with similar models in which missing data was deleted listwise (Supplemental Material 4).
Using the forced entry method, we then regressed COVID-19 anxiety on significant predictors from the four models in a final model that significantly predicted COVID-19 anxiety, F (8, 981) = 37.457, p < .001, adj. R 2 = .228 (Table 7). In this final model, disease concerns, relational concerns, female gender, perceived stress, and generalised anxiety were significant predictors of COVID-19 anxiety whereas vocational/lifestyle concerns, group membership before university, and loneliness were not. Generalised anxiety contributed 5% unique variance in explaining COVID-19 anxiety, feeling stressed 1.6%, and being female 1%. Differences were observed when the final model was run with missing data deleted pairwise (Table 7) compared to listwise (Supplemental Material 5). On average, females scored higher on the measure of COVID-19 anxiety (M = 58.54, SD = 23.75) than males (M = 48.19,SD = 27.48). This difference was statistically significant: t(386.84) = −5.34, p < .001.

Summary of findings
The current research aimed to explore the role of demographic, current mental health, and psychosocial variables as risk factors for increased COVID-19 anxiety in a sample of 1,164 UK university students. We found that students' self-report current mental health symptomatology of stress, anxiety, social phobia, and depression were significantly positively correlated with COVID-19 anxiety, whilst lower levels of belonging, social identity, and group membership were negatively correlated with COVID-19 anxiety. Regression analyses revealed that COVID-19 disease concerns, relational concerns, vocational/lifestyle concerns, perceived stress, generalised anxiety group membership before university and loneliness predicted students COVID-19 anxiety. In the final multivariate regression model, it was shown that disease and relational concerns, being female, stress, and anxiety positively predicted COVID-19 anxiety.
Our data provide evidence that students with stress and anxiety-related mental health symptomatology are more likely to report higher levels of COVID-19 anxiety. Our findings support the notion that female university students are more prone to be psychologically impacted by COVID-19, as demonstrated in a US sample of college students (Browning et al. 2021). Females historically show greater prevalence in anxiety-related disorders (McLean et al. 2011). Two studies among French university students also conclude that females and those experiencing greater levels of stress are more likely to be psychologically affected by the COVID-19 pandemic (Bourion-Bedes et al. 2021;Wethelet et al. 2020). Similar findings have been observed within Polish, Russian and Belarusian students (Debowska, Horeczy, Boduszek, 2020;Gritsenko et al. 2020).
Recent data suggest that students from lower income families experienced greater distress and were disproportionately affected during this pandemic (Rudenstine et al. 2020). However, IMD did not emerge as a significant predictor of COVID-19 anxiety in UK students in either the separate or final model. This could be due to the relatively low variance as for a large number of participants the home town was the area of the University where the study was conducted, which is above average in IMD terms.
Previous studies have measured COVID-19 anxiety with common psychometric tools such as the GAD and the PSS. We created items to capture the different real-life domains functioning as COVID-19 stressors, as opposed to generic measures of stress and anxiety. As a result, we explored differential associations between several COVID-19 related stressors and COVID-19 anxiety, and found that disease-related concerns such as worrying about getting infected or family/friends getting infected, and worrying about social relationships were most anxiety-provoking. This shows that issues with a more personal tone that have to do with health and illness seemed to concern students at the time of the recruitment (at the beginning of the pandemic-April to July 2020). Conversely, academic, financial and worries about the future, which are more external issues, did not seem to produce greater COVID-19 anxiety. Previous research suggests that, globally, academic or future-oriented (Aristovnik et al. 2020) and interpersonal matters are amongst the most common factors to undermine students' mental health during the pandemic (Padrón et al. 2021). Our data showed that worrying about the future was the most common COVID-19 stressor; however, it did not emerge as a significant predictor of COVID-19 anxiety. At the onset of the pandemic, students' primary concerns may be linked to personal issues such as health. As the pandemic progress students may be considering how the pandemic might affect other domains of their life.

Strengths, limitations, and directions for future research
The cross-sectional nature of our data does not allow for causal inferences to be made. Thus, it is imperative that future research explores the ongoing impact of the pandemic using longitudinal designs. However, our findings contribute to the literature in different ways. Firstly, they come to complement and strengthen the notion that being female, suffering from generalised anxiety, being stressed and worrying about the health and social consequences of the pandemic put individuals at increased risk of COVID-19 anxiety. Undergoing an educational transition period can be stressful, therefore the additional stress of a pandemic can have negative effects on student wellbeing. It would be worthwhile to develop interventions that are targeted to students and address specifically pandemic-related anxiety. Any future stress management interventions for students stressed due to COVID-19 should be designed with caution and consider being targeted to students with extremely high levels of anxiety and stress. Two recent systematic reviews revealed that stress-targeted intervention such as CBT (Amanvermez et al. 2020) and mindfulness (Dawson et al. 2020) are especially beneficial for stressed students as they were successful in improving distress and anxiety symptomatology. Given the pandemic has intensified the digitalisation of courses and services, it would be reasonable to suggest online or mhealth stress interventions for students experiencing significant stress. Despite the relative lack of our results' generalisability as the vast majority of the sample were students at the university where the study was conducted, our findings illustrate the toll that stress and anxiety can pose on students' ability to deal with a massive stressor like the pandemic.

Conclusions
Our research showed that general anxiety symptoms and stress, and being female, associated with greater COVID-19 anxiety. Although concerns about the future was the most highly endorsed individual COVID-19-related stressor, disease-related concerns and social relationship-related concerns were predictive of greater COVID-19 related anxiety.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Funding
This work was supported by the Office for Students.