2.2 Statistical analyses
Descriptive statistics, χ2 tests, and univariate analysis of variance were computed to analyse the data from the survey conducted from 14th September 2021 to 14th November 2021 (t2).
Respondents to the survey conducted at t2 were matched with respondents from the survey conducted from 3rd February 2021 to 28th February 2021 (t1) using propensity score matching (Ho et al., 2011; Rosenbaum & Rubin, 1983). A propensity score is assigned to each participant, representing the probability of belonging to one of the two groups, given a vector of observed covariates; in this case age, gender, region and school type. This score is then included as a covariate in all future analyses (Bartak et al., 2009). Multiple matching techniques were assessed via jitter plots, histograms and percent balance improvement and ‘nearest neighbour’ was selected as having the best matching.
After matching the two groups (t1 and t2) were compared, with the mental health measures as the dependent variables and time point and propensity score as fixed factors, using linear models for continuous measures and general linear models with the binomial family for assessing the likelihood of being over the cut-off score.
Effect sizes are shown as Cohens d . P values were 2-tailed, and statistical significance was set at P  = .05.
Results
A total of 1505 adolescents participated in the survey in autumn 2021 (t2), and 1173 (77.9%) were female. Their mean (SD) age was 16.3 (1.4) years, and 249 (16.5%) had a migration background.
Measures of self-reported psychological health of the whole sample by gender are summarized in Table 1. All variables were significantly affected by gender (all p  < .0001), with worst scores in non-binary students. Girls showed worse mental health compared to boys. The cut-off for clinically relevant depressive symptoms (ie, PHQ-9 score, ≥11) was exceeded by 61.9% girls, 38.1% boys and 94.1% non-binary students. 49.3% girls, 28.8% boys and 70.6% non-binary students had clinically relevant anxiety symptoms (ie, GAD-7 score, ≥11). Clinically relevant moderate insomnia (ie, ISI score, ≥15) was reported by 27.5% girls, 16.7% girls and 43.1% students with non-binary gender. The prevalence of suicidal ideation (item 9 of the PHQ-9) within the last 2 weeks was 46.8% in girls, 32.0% in boys and 90.2% in students with non-binary gender identity.
Of 1505 high school students participating at t2, 1257 could be matched to participants from t1, leading to a total sample size of N=2514. The matched sample at t1 (N=1257) comprised 77.2% (N=971) girls, 20.4% (N=256) boys and 2.4% (N=30) non-binary adolescents with a mean age of 16.3 (SD= 1.38) years. 16.63% (N=209) had a migration background. The matched sample at t2 (N=1257) comprised 78% (N=981) girls, 19.3% (N=242) boys and 2.7% (N=34) non-binary adolescents, with a mean age of 16.3 (SD=1.4) years, and 16.63% (N=209) had a migration background.
Table 2 summarizes the results of the mental health measurements and cut-off values by gender and time point. Table 3 shows the statistical results. Matched-sample analyses showed a mean (SD) change from time 1 to time 2 for girls’ self-ratings of well-being (WHO-5: 37.7 [20.1] to 34.82 (20.2)), depressive symptoms (PHQ-9: 11.6 (6.43) to 13.0 (6.32)), and insomnia (ISI: 10.2 (5.47) to 10.8 (5.7); all p  < .05). For boys and adolescents with non-binary gender, no differences in mean scores were observed between t1 and t2. At t2 the number of girls exceeding the cut-off for clinically relevant depression (61.8%), and insomnia (26.4%) increased compared to t1 (53.2% depression and 21.0% insomnia; all p< 0.01). The prevalence of suicidal ideation as assessed by a single-item from the PHQ-9 was higher in girls (47.2%) and boys (33.5%) at t2 as compared to t1 (girls: 35.3%, boys: 30.6%) withp < 0.0001. Effect sizes for the significant findings were small, with a range from 0.05 to 0.23 (Table 3). Although girls had worsened mean scores and increased prevalences for most variables investigated at t2 compared with t1, only one significant difference (suicidal ideas) was observed for boys and no significant differences were observed for adolescents with non-binary gender.
Discussion
The findings of this study suggest that from February 2021 to autumn 2021 girls’ mental health has worsened in the areas of well-being, depression, suicidal thoughts, and sleep. Although little change has occurred in boys’ mental health during this time, suicidal thoughts have increased significantly. Likewise, non-binary adolescents showed similar scores in early and late 2021. The lack of change in non-binary youth’s mental health may be due to a ceiling effect as their scores were already very poor in the first survey. In addition, the non-binary and male samples were smaller than the female sample, and significant results are more probable in larger samples.
At this point in the pandemic, it appears that the mental health of high school students is not directly linked to school openings or lockdowns. Furthermore, as new data emerge, it seems possible that factors other than school and restrictions may also be having detrimental effects on youth mental health. Özlü-Erkilic et al. (2021) investigated 15-25-year-olds in Turkey and Austria at two time points (May-June 2020 and September-October 2020) and found a significant decrease in mental health between the two time points. As the pandemic progressed, the estimated severity of COVID-19 infection, the fear of the individual and/or a family member being infected and ruminations about COVID-19 were rated higher across all groups. Therefore, rather than demonstrating resilience and coping, young people seem to be experiencing deteriorating mental health as the pandemic continues.
Similarly in adults, compared to during the first lockdown, mental health did not improve soon after the lockdown or six months after the end of the first lockdown (Pieh, Budimir, et al., 2021; Pieh, Probst, et al., 2021). Furthermore, an Austrian study of 12 waves between April and December 2020 found depressive symptoms accumulated over time, worsening with each lockdown but not improving in the periods of reduced restrictions in between, and this was particularly the case for young people (16-29 years; Niederkrotenthaler et al., 2022).
As such, it seems that although school closures and restrictions likely play a role in mental health (Humer et al., 2021; Riiser et al., 2021), more general pandemic factors are also involved, such as rises in daily COVID cases (Özlü-Erkilic et al., 2021), slow vaccination rates in Austria, exhaustion/depletion of mental resources (Orgilés et al., 2020), and concerns about the future (Salari et al., 2020). Further studies are required to better understand the underlying reasons for the persistent mental health burden in adolescents.
The cross-sectional nature, the exclusive use of self-rating instruments, as well as the possibility of a self-selection bias due to the online implementation of the study are important limitations to note.
Overall, these results highlight the need to implement timely health promotion and prevention strategies to mitigate the mental health burden in young people caused by the COVID-19 pandemic and associated measures.
Author Contributions: Conceptualization, C.P., E.H., T.P.; methodology, E.H., C.P., formal analysis, R.D., E.H.; investigation, E.H., C.P.; data curation, E.H.; writing—original draft preparation, R.D., A.J. writing—review and editing, E.H., T.O.; project administration, E.H.; All authors have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.
Institutional Review Board Statement: The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Ethics Committee of the Danube University Krems (protocol code EK GZ 41/2018–2021).
Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.
Data Availability Statement: The raw data supporting the conclusion of this article will be made available by the authors upon reasonable request.
Acknowledgments: Authors thank the participating counselors.
Conflicts of Interest: The authors declare no conflict of interest.