Did Stress Prevalence Among Adolescents in Scandinavia Change from 2000 to 2019? A literature review

As the age constituting adolescence is not standardized, and with a lack of knowledge of stress prevalence in age groups outside of 16-24 and 16-29 in Scandinavia, this literature review used PubMed’s definition of adolescence, age 13-18 (19). A cut-off was set, where maximum 10 % of the participants could be outside this age range for the included studies. The inclusion criteria were: 1) The studies were conducted in the time period 2000-2019; 2) The study population was either Swedish, Danish or Norwegian; 3) The study population was between 13-18 years of age; 4) Stress was measured through self-reported questions/questionnaires; 5) The questions/questionnaires could be compared with other studies; 6) The study was published in a peerreview journal. In cases where several studies conducted research on the same population, the study with a comparable stress question/ questionnaire was preferred.

Exclusion criteria were: 1) The studies had clinical populations; 2) The studies only measured school-related stress; 3) The study was an abstract, a systematic review or a protocol.

The surveys and reports included were identified through a grey literature search. Relevant surveys and reports were found through an extensive search on Google and governmental web pages, for the respective Scandinavian countries. A chain search was also conducted, identifying surveys through citations in other literature. All of the criteria for inclusion of the surveys and reports followed those of the studies, except the criterion of publication in a peer-review journal – which was not applicable.

For surveys and reports categorizing their study population by grade levels in school, grade levels were translated into approximate age. This takes into consideration that the Scandinavian countries have different grading systems (i.e., in Sweden students start in 9th grade the year they become 15 years old, compared to 10th grade in Norway). Also, to better reflect the actual age of the study populations, the translation into approximate age is based on the time the survey was taken (i.e., spring semester versus fall semester).

The papers included in this review, were identified through systematic searches of the PubMed and PsycInfo databases on April 20, 2020. The grey literature search was conducted during the same period. Therefore, the papers, surveys, and reports included in the analysis are not affected by possible contributions to stress caused by the Corona pandemic. The search strategy was produced to best target stress self-reported by adolescents between 13-18 years of age in Norway, Sweden and Denmark, from year 2000-2019. A description of the search string for PubMed and PsycInfo, can be found in Supplemental Appendix A.

(a) Study selection: The searches in PubMed and PsycInfo, resulted in 4651 identified papers. After removing duplicates, 3893 unique papers were screened through reading titles and abstracts. Among these, 132 papers with study characteristics possibly being in line with the inclusion criteria, were assessed in full text. Finally, nine papers met the criteria for inclusion.

After searching Google, websites of national health authorities, and conducting a chain search, 25 reports and surveys were assessed in full-text, and nine met the criteria for inclusion. All the included papers, surveys, and reports were assessed for risk of bias (see Supplemental Appendix B), by the use of the Newcastle-Ottawa quality assessment scale (NOS) (see Supplemental Appendix C).

There was a gender difference in stress prevalence, where girls scored higher than boys in all the papers, surveys, and reports that provided stress scores stratified on gender. Data from Ungdata 18 and Ungdata 19, suggest that it is especially from age 13-14 to age 14-15 (approximate ages for grade 8 and grade 9 in the Norwegian school system), that the gender difference seems to occur (see Supplemental Table 10). The results from Child-ULF 2008-2014 show that the gender difference in stress scores is higher in the age group 16-18 compared to 13-15 (see Supplemental Table 3). The same pattern is also seen in Children and Young in Scania, where the gender difference in stress is higher among pupils aged 17-18 than for the 15-16-year-olds (see Supplemental Table 5). Also, in all the studies conducted by Moksnes and different co-authors (i.e. (36,39)), girls scored significantly higher than boys on a majority of the stress domains (see Supplemental Table 11 and Supplemental Table 12). These findings are strengthened by the use of a validated stress questionnaire (ASQ-N) in all the studies. Thus, the data show that there is pervasive gender difference in stress among adolescents, aged 13-18, in Scandinavia and those differences increase with age.

All papers, surveys, and reports that provided data stratified on age among 13-18-year-old Scandinavian adolescents, show that stress scores increase with age. For instance, in Child-LNU from 2010 the percentage of 13-15-year-olds that were stressed at least once a week was 44,3 %, compared to 55,4 % of the 16-18-year-olds (see Supplemental Table 2). The results from Ungdata 18 and Ungdata 19 (see Supplemental Table 10), show that stress scores were highest among the oldest participants (approximate age 18-19; final year of upper secondary school), which may be due to the importance of final school grades for further educational opportunities (43). Further, the age difference in stress is influenced by gender as well, in that boys seem to have lower and more stable stress scores than girls. Overall, the findings indicate that stress may increase with age within the age group 13-18, and more so for the girls.

A general finding from the included papers, surveys, and reports on stress among 13-18-year-old adolescents in Scandinavia, is the use of nonvalidated stress questions/questionnaires, possible nonresponse bias, and populations that are not representative on the national scale.

The use of non-validated stress questions is an issue in all the Swedish surveys. Therefore, one cannot conclude that they measured Swedish adolescents’ real experience of stress. The low response rate and possible non-response bias in the Child-ULF surveys from 2008-2014 and 2016-2019 (41,44), are issues concerning the representativeness of the study population and generalizability of the results. The surveys Children and Young in Scania from 2012 and 2016, had high numbers of participants and high response rates – i.e. 82 % in grade 9 (approximate age 15-16) in 2012 (25). This is a strength for the representativeness of the results for adolescents aged 15-16 and 17-18, who attend lower secondary and upper secondary school, in Scania County. Even though the results showed an increase in stress from 2012 to 2016, the use of a nonvalidated stress question is a concern for the validity of the results. Further, the results cannot be generalized outside of Scania County, as other counties in Sweden may differ in socio-economic factors related to stress among adolescents. Thorsen et al. (28) and von Rosen et al. (29) used a validated questionnaire to measure stress (Cohen’s Perceived Stress Scale), and the study populations were similar in age, which makes a comparison of stress scores more appropriate. However, the low participation rate in Thorsen et al. (28) is a challenge regarding the representativeness of the study population, and as the two study populations come from different, specific parts of Sweden, the results cannot be generalized to a national scale for Swedish adolescents.

Regarding risk of bias in the Danish literature, all surveys used non-validated stress questions. The inclusion of pain-related discomfort in the MULD survey’s stress-question prompts a critical examination of whether the survey genuinely measured stress. While the survey’s broad scope encompassing various aspects of adolescent well-being is commendable, the incorporation of pain in the stress question may introduce ambiguity into the stress measurement. In addition, the use of different sampling strategies in the MULD surveys is a challenge regarding both the representativeness of the study populations and comparisons of stress scores between the MULD surveys. The Danish National Youth Studies from 2014 and 2019, may be considered representative for Danish 15-18-year-olds attending upper secondary school in 2014 and 2019, as they sampled a representative study population and had adequate response rates, however certain school programs were not well-represented (33,45). Overall, due to the limitations in the literature, one cannot conclude that the Danish surveys ranging from year 2003 to 2019, did capture the real experience of stress among 13-18-year-old adolescents.

The large number of participants, high response rates and wide representativeness of Norwegian adolescents bolster our confidence in the results from Ungdata 18 and Ungdata 19 (7,35). However, the use of a non-validated stress question and lack of information on non-respondents, makes it unclear whether the real experience of stress has been captured and whether non-respondents differ in characteristics that could be related to stress. The use of a validated stress questionnaire is a strength of the studies conducted by Moksnes and co-authors. As the studies were conducted in Mid-Norway, the results cannot necessarily be generalized to the entire Norwegian or Scandinavian adolescent population. Also, age differences and geographical factors regarding the study populations – i.e., if the students attended urban versus rural schools – may have influenced the stress scores.

A majority of the papers, surveys, and reports did not report characteristics of non-respondents. This is an issue, since non-respondents could differ in characteristics related to stress prevalence. For instance, Moksnes et al. (40) provide information on the response rate in their 2016 study population, which was 65.1 %, but do not give any information on the non-respondents. Therefore, one cannot exclude the possibility that non-respondents differ in characteristics related to prevalence of stress. This is probably a larger issue in Child-ULF 2008-2014, and 2016-2019. The participants in Child-ULF were children of participants in the adult Swedish Living Conditions Surveys (ULF) for the respective years (41). As non-respondents in the ULF surveys were more likely to have a non-Western background, lower income and lower education, the adolescents who did not participate in Child-ULF may also have these characteristics, that could be related to higher levels of stress (41). Further, the total share of nonrespondents in the Child-ULF and ULF surveys increased from 60 % in 2010 to 72 % in 2017 (41). As the majority of the papers, surveys, and reports do not provide sufficient information on nonrespondents, one cannot exclude the possibility that they differ in characteristics related to prevalence of stress. Overall, this is a challenge in relation to the representativeness and generalizability of the results among Scandinavian adolescents, aged 13-18, in the time period of the years 2000-2019.

A general finding in the Scandinavian literature is a gender difference in stress levels, and an increase in stress levels with age from 13 to 18 years. This is similar to the pattern found in self-reported psychological symptoms for this age group (35), and for the personality trait Neuroticism (46), which is a risk factor for developing mental illness (47). Girls scored higher on stress in all the papers, surveys, and reports included in the literature review that stratified scores on gender. Results from Ungdata 18 and Ungdata 19 (7,35) suggest that the gender difference arises from age 13-14 to age 14-15. This is in line with findings regarding the onset of gender difference in depressive symptoms (48). The identified gender differences in adolescent stress levels, highlight complex interplays of sociocultural, biological, and psychosocial factors. Societal expectations, hormonal fluctuations, and coping mechanisms contribute to varied stress experiences. Differences in peer and parental relationships, along with distinct educational pressures, also play a role. Understanding these nuanced dynamics is crucial for tailored interventions to support the mental well-being of adolescents across genders. Further research is needed to explore these factors comprehensively. Stress increased with age, both for girls and boys. However, the increase was steeper for the girls. These findings imply that levels of stress increase from age 13 to 18 in Scandinavia. This underlines the importance of viewing 13-18-year-olds as a heterogeneous group in relation to stress. Therefore, comparisons across- and between studies with large age-spans for adolescence, should be undertaken with caution and under consideration of the age and gender composition of the different populations.

An increase in stress levels was found for specific age groups in certain years. For instance, among 13-15-year-old girls and 16-18-year old boys in the Swedish 2012-2013 Child-ULF survey (23) (see Supplemental Table 3). Also, there was an increase in stress among pupils aged 15-16 and 17-18 in Scania County in 2016, compared to 2012 (25) (see Supplemental Table 5). However, due to the use of non-validated stress questions in both surveys, and a possible nonresponse bias in the Child-ULF surveys, the results may not be considered reliable to reflect an increase in stress prevalence. The most valid information on stress levels measured over time, comes from the studies conducted by Moksnes and different coauthors in Norway (36–40). By using a validated stress questionnaire specifically designed for Norwegian adolescents (Adolescent Stress Questionnaire – ASQ-N), the studies are more likely to capture the real experience of stress. As the ASQ-N has several subscales, the questionnaire can also help to address which domains of life are most relevant for stress among adolescents. This is important, as it can inform the design of stressreducing interventions. However, even these studies’ results are difficult to interpret with respect to a possible increase in stress levels over time, as the populations differ slightly, both in age and rural versus urban composition. As girls scored higher than boys on all subscales, and as stress scores were higher among the older study populations, the results support the general finding of gender differences in stress and an increase of stress levels with age. Again, this exemplifies the importance of not viewing adolescents, aged 13-18, as a homogenous group when it comes to stress. In future research, the use of the ASQ would be helpful to gain a more valid and detailed picture of stress among Scandinavian adolescents. Additionally, it would be relevant to include questions regarding stress related to body image, sports/exercise and social media – as these domains were related to negative stress among adolescents in the Ungdata surveys (7,35) and may become more important stress factors among adolescents due to increased use of Internet and smart phones (49).

As the papers, surveys, and reports included in the literature review were cross-sectional, it is not possible to examine whether other factors have influenced the stress scores. For instance, in Children and Young in Scania, increased stress scores were mirrored by increased scores in self-reported psychological symptoms from 2012 to 2016 (25). As depression and stress covary (50), higher rates of stress may be due to more adolescents experiencing depressive symptoms and possibly mental illness. It might therefore be speculated that changes in other factors could covary with or possibly cause higher stress scores. This also relates to the issue with the widespread use of non-validated stress questions, as it is uncertain whether the questions have measured the experience of stress itself or possibly other stress-related phenomena, such as depressive symptoms.

Two of the included studies and almost all the included surveys, used single-item measures to assess stress. This could make a comparison of stress scores across the studies and surveys possible. However, due to differences in formulations and time frames in the stress questions, this was not considered appropriate. For instance, the “Child-ULF” surveys from 2008-2014, asked: «During the past 6 months, how often have you had the following problems?» «I felt stressed». “Ungdata” from 2018 and 2019, asked whether the person had experienced so many demands during the last week that they had trouble coping with it. This can lead to a different understanding and response to questions that try to address the same phenomenon, which further makes a comparison challenging. Also, as few papers, surveys and reports measured stress over longer periods of time (i.e., 10-15 years) it is difficult to address possible changes in stress prevalence within the entire range of the study, namely the years 2000-2019. Therefore, based on the present evidence, one cannot entirely exclude the possibility that stress levels may have increased from 2000 – 2019. However, neither do the data support such a conclusion.

Finally, the chosen search strategy can pose a limitation on the number of papers included in the analysis. The two databases, PubMed and PsycInfo, were chosen on the basis that they are among the largest databases within medical literature and mental health. However, one cannot exclude the possibility that adding more databases to the search strategy would lead to findings of more papers relevant to include in the analysis.

Overall, the findings of the present literature review are of relevance both when comparing stress prevalence among adolescents, aged 13-18 across time and for suggesting methodological directions for further research. It is of utmost importance that future research addressing this question uses validated stress items or questionnaires.

A large sample of 13-18-year-olds, with equal representation of each age and gender, recruited from a nationally representative population, would be preferable in future studies. Also, as The Danish National Youth Studies from 2014 and 2019 (34,42), and Ungdata 18 and Ungdata 19 (7,35) show that surveys conducted in schools have high response rates, this may be a good strategy for attaining high response rates and representative populations.

If these conditions are met, it will be possible to investigate possible changes in stress prevalence over time. This can provide information that can be relevant in addressing factors that can affect adolescents’ mental health, and in designing preventive measures.