Worry and Psychosomatic Problems Among Adolescents in Sweden in the Wake of the COVID-19 Pandemic: Unequal Patterns Among Sociodemographic Groups?

Purpose To analyze potential associations between sociodemographic factors and adolescents' worry about family finances and other types of worries during the COVID-19 pandemic and the potential associations between different worry themes and psychosomatic problems. Methods From December 2020 to March 2021, 3,068 16–17 years old students in Sweden completed a questionnaire about the impact of the COVID-19 pandemic on their schooling and everyday life. From April to June 2020, 70% of these students also responded to questions about their psychosomatic health. Logistic regression was applied to analyze the associations between sociodemographic factors and different worry themes and estimate the association between worry and psychosomatic problems. Results Across all 13 worry themes, girls reported worry to a higher degree than boys. In most of the cases, adolescents with a lower parental educational background or immigrant background reported more frequent worrying. This was most evident for worry about family finances. The differences between adolescents with a Swedish background and those with a foreign background were large except for worry about climate change. All worry themes were associated with the students' psychosomatic health. Those who often or always experienced worry were more likely to report psychosomatic problems. Worry about family finances showed the highest odds ratio and marginal effect. Discussion Worry about family finances during the COVID-19 pandemic turns out to be a strong and socially structured stressor that may widen the health inequalities among young people. This underlines the need for preventive and promoting measures aimed at supporting socioeconomic disadvantaged groups.


IMPLICATIONS AND CONTRIBUTION
The study demonstrates the significance of addressing sociodemographic inequalities in adolescents' responses to the COVID-19 pandemic. Adolescents' worry about family finances is a strong and socially structured stressor, potentiating risks of increased health inequalities in the wake of the COVID-19 pandemic. The need for financial support aimed at supporting socioeconomic disadvantaged families is highlighted as well as the need for preventive and health-promoting measures targeting socioeconomically disadvantaged groups of young people.
While young people are less affected by COVID-19 in terms of severe illness and death when compared to older age groups, in a worldwide perspective, their living conditions and everyday life have been profoundly affected by the COVID-19 pandemic [1]. Public health controls to limit the spread of the pandemic were implemented, e.g., schools were closed and classroom learning was replaced with distance learning.
During the early phase of the pandemic in the spring of 2020, 150 countries fully closed their schools [2]. In March to May 2020, up to 1.5 billion children and young people were out of school [3]. In an international context, Sweden stands out as an exception by not adopting the universal strategy of school closure as a general way to confine the transmission of the corona virus. In the compulsory school system, including lower and secondary levels, the schools in Sweden remained opened, while in upper secondary schools and universities, classroom teaching was replaced by distance learning.
COVID-19 has increased the risk of mental health problems among young people with many studies showing that mental health and wellbeing of young people is negatively affected by disease containment measures [4]. A systematic review conducted by Viner and colleagues [3] identified 27 studies concerning mental health revealing considerable impacts across emotional, behavioral, and restlessness/inattention problems. Viner and colleagues note that these studies do not allow for distinguishing the impacts of school closures from impacts of other social lockdown measures. Reports about declines in mental health are also conveyed by other systematic reviews [5,6] and large follow-up surveys (such as that completed by NHS Digital [7]). These show that depression, anxiety, stress, worry, social isolation, and loneliness have increased. Meta-analyses of 29 studies showed that the global prevalence of clinically elevated symptoms of depression and anxiety in youth during the first year of the COVID-19 pandemic were twice as big as prepandemic rates [8].
While a preponderance of literature supports the negative psychological impact the COVID-19 pandemic has on adolescent mental health, there is also research that counters this finding. For example, according to a report published by the Swedish Corona Commission, there are no indications of declines in mental health among upper secondary school students following the pandemic [9].That said, it is possible that methodological issues impact the interpretability of the findings.
Although there is an increasing number of studies pointing to negative impacts of the COVID-19 pandemic on young people's mental health, it is hard to draw firm conclusions because of a scarcity of high quality empirical studies [6]. These methodological shortcomings include lack of pre-pandemic measures and issues relating to sampling [10]. The deficiencies and uncertainties in many studies are summarized by a recent systematic review concluding that "the results demonstrated an overall trend in declining mental health for youth, likely due to the pandemic" [11]. Among the 19 studies which were included in the review, 17 were cross-sectional and two were cohort studies. Almost all of these studies entailed non-probability samples using Internet and social media for recruitment of participants. Additionally, there are few studies that have assessed associations between the COVID-19 pandemic and psychosomatic health in youth. Based on three waves of data collections among adolescents 11e17 years old in Germany, a recent study concludes that low health-related quality of life, mental health problems, and psychosomatic complaints increased during the pandemic, from May 2020 to October 2021 [12].
Moreover, general descriptions of the consequences of the pandemic may suffer from oversimplifications that obscure huge variations across countries as well as within countries. Highincome countries have managed to mitigate the negative impacts of some confinement measures by capitalizing on digital technologies. In schools, internet and digital devices have been frequently used for distance learning. However, students from lower-income families are, to a greater extent, affected by school closures because of poorer access to internet and digital devices [13]. In a global context, there are large inequalities in digital connectivity. Among young people in high-income countries, 87% have internet access at home compared to 6% in low-income countries [14]. Even within high-income countries, inequalities may occur with respect to space for homework and access to parental support.
In considering the mental health impact of the loss of learning for young people, the time perspective is important and potential long-term effects should be kept in mind. School failure is a major risk factor for mental health problems with both the shortterm and long-term negative impacts of school closures and distance learning on education presenting causes for concern that need to be evaluated [3,15].
School closures and distance learning entail not only a loss or diminishment of education but additionally provoked big changes in the social relations, with more loneliness, more sedentary behavior and fewer physical activities for many students [16]. In addition to school closures many children have also been affected by devastating family events with an estimated one million children globally having lost a parent because of deaths associated with COVID-19 [17]. These challenges are further highlighted for public health in the UNICEF's flagship report "The State of the World's Children 2021" [18].
The impacts of COVID-19 on young people have been shown to vary across sociodemographic backgrounds. A UK study found that deterioration of mental health following the pandemic was particularly noticeable among adolescents in one-parent, onechild, and low-income households [19]. Distress may increase in families experiencing financial adversity and challenges due to home schooling [10], and there are also indications that COVID-19 in families has increased the gap in prosocial behaviors between low and high SES students [20]. While some studies indicate that young people with mental health conditions are at increased risk of mental ill-health because of the COVID-19 pandemic [21,22], the aforementioned UK study reports an opposite pattern. Among adolescents with prepandemic better-than-median mental health, internalizing and externalizing problems increased. In contrast, among those with prepandemic worse-than-median mental health, these problems decreased [19].
Downturns in the economy with increasing unemployment rates during the pandemic are affecting family finances and increasing the risk of economic stress [23,24]. Worry about family finances has been shown to be a significant stressor among adolescents during the pandemic [25]. While economic stress may be considered as a socioeconomic stressor per see that is more pronounced among lower socioeconomic groups, in fact worry about family finances may occur across socioeconomic groups, specifically during economic crisis.
The aim of the current study is to analyze the potential associations between sociodemographic factors and worry about family finances and other types of worries during the COVID-19pandemic and the potential associations between different worry themes and psychosomatic problems.

Research questions
Is worry about personal, family and global issues during the COVID-19 pandemic socially structured, i.e., unequally distributed among sociodemographic groups?
How is worry about personal, family and global issues during the COVID-19 pandemic related to psychosomatic problems?

Data collection
The survey data were collected as a part of the cohortsequential study "The Evaluation Through Follow-up (ETF)" which is an infradata structure hosted by the University of Gothenburg, Sweden. The ETF is primarily used for evaluation and research about schools and education in Sweden. The participants in the current study belong to the 2004 birth cohort, the latest of 10 cohort samples. The cohort was sampled by Statistics Sweden when the students attended grade 3. To ensure a nationally representative sample, a two-stage sampling procedure was applied, drawing a stratified sample of schools among which classes were selected. The current study uses survey data collected by Statistics Sweden among the 2004 cohort students at two different time points.
Time point 1. Data collected from April to June 2020 when the majority of the students were in grade 9 (15e16 years old). During this data collection period, the students were exposed to the COVID-19 pandemic and general containment measures, but were not subjected to school closures because the compulsory schools (up to and including grade 9) remained open.
Three thousand and thirty-three grade-9 students (31.7% of the sample) completed a questionnaire about their school and life situation and their psychosomatic health (60% completing paper and pencil; 40% web). The grade-9 questionnaire did not include the worry questions used in the survey conducted in 2020/2021 in year 1 of upper secondary school. The survey in school grade 9 was approved by the Swedish Ethical Review Authority (decision 2020-04-01, dnr 2019-06033).
Time point 2. Data collected from December 2020 to March 2021 when the students were in the first year of upper secondary school (16e17 years old). During this data collection period, the students were subjected to school closures because the upper secondary schools were mainly closed and distance learning was implemented.
Three thousand and sixty-eight students in the first year of upper secondary school completed an online questionnaire about the impact of the COVID-19 pandemic on their schooling, worry and everyday life. The response rate was 32%. The survey conducted in 2020/2021 in year 1 of upper secondary school did not include questions about the students' health, and was therefore not submitted for ethical review. However, the same form for informed consent was used as stipulated for the 2020 spring survey by Statistics Sweden to meet requirements by the Ethical Review Authority in its decision 2020-04-01. While 3,068 students participated in the survey in grade 1 of upper secondary school, 2,149 of these students also participated in the survey in grade 9, corresponding to 70% of the entire set of data collected in grade 1 of upper secondary school.
In order to analyze the associations between sociodemographic factors and worry, sociodemographic data retrieved from national registers administered by Statistics Sweden were linked to the survey data collected in year 1 of upper secondary school.

Variables and measures
Worry. Six questions were directly related to the COVID-19 pandemic. Are you worried that...
."you could become ill with COVID-19?" ."your parents could become ill with COVID-19" ."another close relative (for example, grandparents) could become ill with COVID-19?" ."your family's finances could get worse due to the corona pandemic?" ."your parents could become unemployed due to the corona pandemic?" ."the healthcare system will not be able to take care of everyone who is ill with COVID-19 and needs care?" Another seven questions were aimed to tap information about other types of worry, not directly related to the COVID-19 pandemic: "Are you worried about...
.being subjected to robbery or violence? .being bullied?" ."gang crime?" ."terrorist attacks in Sweden?" ."war and other threats to world peace?" ."climate change?" ."an uncertain future?" All questions about worry included five response categories: Never; Seldom; Sometimes; Often; Always. In the analyses, these categories were trichotomized by collapsing the never-seldom and often-always response categories.
Socio-demographic factors. Sex was a binary variable. Detailed data about parental education were condensed into a three-category variable: Lower: No upper secondary school education or upper secondary school education from a vocational program; Mid: Education from a theoretical program at upper secondary school or higher education less than two years; Higher: Two or more years higher education. Data on children's and parents' country of birth were used to construct a composite variable on national background consisting of two categories: Swedish background (child and at least one of the parents born in Sweden, or child born outside Sweden and both parents born in Sweden.); foreign background (child born in Sweden and both parents born outside Sweden, or child and at least one parent born outside of Sweden).
Psychosomatic problems. Psychosomatic health was measured with the PSP-Scale which is a standardized measure constructed through summation of the students' responses to eight items.
The response categories for all these items, which are in the form of questions, are: always (1); often (2); sometimes (3); seldom (4); never (5). The recall period concerned the last 6 months. This ordering implies that the lower the value, the higher the degree of psychosomatic problems.
Some of these eight items may be viewed as indicators of somatic problems, others as indicators of psychological problems. Taken together in a composite measure, somatic and psychological items have been shown to reflect a common dimension, which has been labeled psychosomatic [26]. In the present paper, the concept of psychosomatic problems is used in a generic way without making any presumptions about aetiology. This approach is similar to those used in previous public health studies among adolescents [12,27e31].
The psychometric properties of the PSP scale were examined by applying Rasch Measurement Theory [32] on the entire grade 9 sample. At a general level, the PSP-Scale met the measurement requirements for invariance and proper categorization of the items. Analyses of Differential Item Functioning across each of the three sociodemographic factors indicated some evidence of sex-DIF across one item (sad), i.e., this item did not function in the same way for boys and girls. This sex DIF was considered to have minor impact on the scale as a whole. The reliability was good, indicated by a high value (0.86) on the Person Separation Index, an analogue to Cronbach's alpha. The overall results of the Rasch analysis are consistent with previous reported analyses of the PSP scale [26,33].

Analysis
The survey data were weighted to take account of potential bias caused by systematic deviations of the nonresponses from the original nationally representative sample. For this purpose, estimated calibration weights [34] were applied, which were based on an algorithm developed by Statistics Sweden and including four factors: sex, parental education, immigration and school marks.
Multinomial logistic regression was applied for analysis of the associations between socio-demographic factors and 13 different worry themes. For three selected worry themes, marginal effects are reported in addition to relative risk ratios.
Binary logistic regression was used to analyze the association between worry and psychosomatic problems. The person location values generated by the Rasch analysis mentioned above were divided into two categories using the 10th percentile as a cutoff point for higher degree of psychosomatic problems. Marginal effects are reported based on multivariate analysis including three sociodemographic factors and three selected variables representing different worry themes.
Contingency table analysis and the logistic regression analysis were conducted using the STATA 15.1 software, applying its svy design tool for complex survey data. The user written STATA program SPost13 was used for some analyses of marginal effects. The Rasch model analysis was performed with the software RUMM2030Plus.

Results
In Figure 1 the representativeness of the survey data is shown across socio-demographic groups, comparing unweighted survey data and weighted survey data with register data (percent). Figure 1 shows that the collected survey data clearly deviate from the original cohort sample drawn with respect to three sociodemographic variables: Girls and students with Swedish background and higher parental education are overrepresented. The figure also shows that the differences between the register data and the unweighted survey data are more or less eliminated after weighting of the data.
In Table 1 the proportions (percentage) of students experiencing worry often or always, distributed across 13 worry themes and three socio-demographic groups are shown using weighted data.  Table 1 shows that there is a wide range of high frequency responses to the worry questions. Worry about a close relative shows the highest proportion (>50%) and worry being bullied is showing the lowest (<10%). Across the sample, sex differences are large with girls reporting higher frequencies on all worry questions. Across almost all themes, adolescents with lower parental education backgrounds show higher proportions of frequent worrying than those from families with higher parental education backgrounds. Worry about climate change differs noticeably, showing an inverse pattern. National background shows a consistent pattern: Except for climate change, adolescents with a foreign background report higher proportions of frequent worrying compared to Swedish adolescents.
In Table 2 the outcomes from multinomial logistic regression analysis of the associations between socio-demographic factors and worry are reported as relative risk ratios, regressing each of the 13 worry themes on three sociodemographic factors using weighted data.
For all worry themes, there are significant associations with sex ( Table 2). The relative risk ratio of worry often/always compared to seldom/never are higher for adolescents with lower parental education than for those with higher parental education for most of the COVID-related themes. For about half of the worry themes, adolescents with a foreign background show significant relative risks of worry, while there are no significant associations for the other half of worry themes. Figure 2 presents the outcomes from multinomial logistic regression analysis of the associations between socioeconomic factors and three selected worry themes. These are reported as predictive percentages of worry often or always, regressing each of the three worry themes on three sociodemographic factors.
Worries about becoming ill, family finances and climate change are displayed in Figure 2. This reveals higher predictive percentages for girls relative to boys for all three themes. Worry about family finances shows an obvious pattern with respect to parental education background: About one of 10 young people in families with higher educational background experience such worry, while the corresponding figure is one in four among those in families with lower educational background. Worry about climate change shows an inverse pattern: Those with higher educational background report the highest percentage of worry. Students with foreign background report higher percentages than Swedish adolescents do on all three worry themes.
In Figure 3 the outcomes from binary logistic regression analysis of the associations between the three worry themes and psychosomatic problems are reported as predicted percentages and changes of marginal probability values, regressing psychosomatic problems on three worry themes, simultaneously controlling for three socio-demographic factors. Figure 3 shows that all three worry themes are associated with psychosomatic problems. In particular, the association is evident for worry about family finances, which is indicated by significantly higher predictive percentages for highly frequent worry compared to less frequent worry. Among the three selected worry themes, worry about climate change shows the weakest association with psychosomatic problems.

Discussion
Across the 13 worry themes, the associations between sociodemographic factors and worry are complex and diversified but socially structured.
Not surprisingly given the gender differences in adolescent mental health [35], the sex patterns here are consistent. Girls report worry to a much higher degree than boys. In contrast, the relationships between parental educational background and worry are more heterogenous with the strength and direction depending on which worry themes are subjected to analysis: some worry themes show no association with parental educational and Swedish/foreign background, while a number of themes show clear patterns. In most of those cases, adolescents with a low parental educational background or foreign national background show more frequent worrying. In particular, this pattern is evident for worry about family finances reflecting differences in living condition and vulnerability. Although the Swedish government has made efforts to mitigate the economic consequences for companies and families during the pandemic, serious concerns about family finances seem inevitable and they occur among young people in socioeconomic disadvantaged families. Among the worry themes, worry about climate change comprises a noticeable exception: adolescents from higher educated families report such worry to a much higher extent than those from lower educated families. The differences between adolescents with a Swedish background and those with a foreign background are large except for worry about climate change.
The strong association between adolescent worry about family finances and psychosomatic problems is in line with results presented in previous studies. In a study conducted during the economic crisis in Sweden during the 1990's, among 10 different personal, family and global worry themes, worry about family finances turned out to show the strongest association with psychosomatic problems [37]. In addition, trend analyses of adolescent data showed that increases in economic stress explained the entire increase of psychosomatic problems in the wake of the 1990's recession in Sweden [38].
Studies on the consequences of the COVID-19 pandemic on adolescents' lives have until now mainly focused on the impacts of the containment measures that have been implemented, while it has been difficult to estimate the effects from specific measures, e.g., school closures [3]. Less attention has been paid to the consequences of economic and broader societal changes taking place simultaneously and the extent to which such consequences are unequally distributed among sociodemographic groups. By focusing on, and contrasting a range of adolescent worry themes and their associations with sociodemographic factors and psychosomatic health, the current study adds to the literature on the impacts of the pandemic on adolescents' lives and mental health.
A strength of the present study is that the sociodemographic bias in the survey data caused by nonresponse has been adjusted by using register data to weight the survey data.
Given the scarcity of COVID-19 adolescent surveys reporting on somatic and psychosomatic problems, the current study adds to previous research among adolescents by inclusion of a measure of psychosomatic health that taps information about somatic as well as psychological problems.

Limitations
The current study was not designed for assessing the impacts of the COVID-19 pandemic on young people's psychosomatic health and there were no prepandemic measures available. At the time of the data collection in grade 9, which included measures of psychosomatic problems, there was already an ongoing spread of COVID-19 in the society and containment measures had been implemented. Similarly, in interpreting the associations between worry and psychosomatic problems, the direction of the relationships cannot be determined. Worry may cause psychosomatic problems but the association could also work the other way around, e.g. students experiencing worry may do that because of psychosomatic problems. While the study does not allow for conclusions about causality, the results indicate that worry about family finances turns out to be a strong and socially structured stressor, which support hypotheses about widening health inequalities among young people in the wake of the pandemic. Such hypotheses raise issues regarding preventive measures aimed at supporting the financial situation of socioeconomic disadvantaged groups, as well as universal measures for social and emotional training. Thirteen worry themes regressed on sex, parental education, and national background simultaneously. Relative risk ratios for worry often/always versus seldom/never. 95% Confidence Intervals: Lower Level (LL) and Upper Level (UL). Weighted data. Figure 2. Predictive percentages of worry often or always for three types of worry themes across three sociodemographic groups. Ninety-five percent confidence intervals of the predictive percentages are within parentheses. Computations of probabilities based on multinomial regression including sex, parental education and national background as independent variables. Weighted data. Figure 3. Predicted percentages of a higher degree of psychosomatic problems across three different worry themes and changes in the probability when comparing adolescents reporting worry never/seldom and always/often. 95% confidence intervals values are reported within parentheses. Computations of probabilities based on multinomial regression simultaneously including all three worry themes, sex, parental education and national background as independent variables. Weighted data.