Theoretical background

The onset of the COVID-19 pandemic triggered multiple restrictions at a structural level and personal level (e.g., school closures, contact restrictions). For many children and adolescents this period was characterized by increased homeschooling and isolation, resulting in symptoms such as anxiety and depression [1]. Thus, about two thirds of adolescents felt distressed by the extraordinary situation of the pandemic [2]. Several studies showed associations between mental health and sleep in times of COVID-19 among youths. A meta-analysis by Sharma et al. [3] focused on sleep during the COVID-19 restrictions and emphasized that a large proportion of children and adolescents did not follow recommendations regarding sleep duration and delayed individual bedtimes. The authors also pointed out an alarmingly high prevalence of sleep disturbances during the pandemic. Although negative changes of mental health seem to predominate, some adolescents report positive changes related to pandemic circumstances. A study by Socarras et al. [4] showed sleep improvement for many adolescents in terms of longer sleep duration, better sleep quality, and decreased daytime sleepiness, especially on weekdays.

Individual resources appear to play a distinctive role in coping within the pandemic. Jiang et al. [5] showed that resilience, defined as the ability to endure adversity, acted as a protective factor for depressive symptoms and sleep disturbances in adolescents before and during the pandemic. Their definition of resilience includes positive thinking in a crisis situation, in addition to family support. In an adult sample, personal resources such as positive affect and self-care behavior also predicted subjective sleep quality as well as more positive retrospective changes in sleep [6].

The present study investigates whether adolescents with negative and positive COVID-19-related sleep changes and without CSC differ in terms of their mental health profiles (sleep, stress, emotional and behavioral problems, quality of life, and resources). CSC refers to the subjective extent to which adolescents have noticed a change in their sleep since the onset of the pandemic. First, it is expected that adolescents with positive CSC during the pandemic have more functional profiles in terms of mental health facets than adolescents without CSC, and particularly compared to adolescents with negative CSC. Second, these three groups will be examined with respect to possible group-dependent changes in mental health facets before and during the pandemic.

Methods

Sample and procedure

Data collection was conducted as part of a longitudinal research project at Bielefeld University involving both adolescents with and without sleep disturbances. First, either a 2-week sleep diary or an online questionnaire battery was completed at home in randomized order. Next, a sleep interview was conducted at the university, and further questionnaires were filled out. The data collection procedure was similar for both measurement timepoints, with approximately 9 months in between (standard deviation [SD] = 1.8). Data for the first timepoint (T1) were collected before the start of the COVID-19 pandemic and the first lockdown in Germany from July 2019 (week 29) to the beginning of March 2020 (week 10). The second timepoint (T2) followed from mid-June 2020 (week 25) to the end of February 2021 (week 8). The resulting sample aged 11 to 17 years (N = 92; mean [M] = 14.43; SD = 1.69; 54% female) attended the 5th to 12th grade of a Gymnasium (46%), a secondary school (53%), or a vocational school (1%). The subjects were rewarded with € 30 vouchers for each timepoint. The study was approved by the university ethics committee, and participants and their caregivers gave informed consent.

Instruments

Sleep disturbances were assessed by the Sleep Disturbance Scale for Children (SDSC [7]; α = 0.83Footnote 1). In addition, the presence of an insomnia diagnosis was determined with a sleep interview based on the criteria of the International Classification of Sleep Disorders [8]. Average total sleep time, sleep onset latency (SOL), sleep quality, and sleep efficiency were captured via a 2-week sleep diary. Evening pre-sleep arousal was assessed using the Pre-Sleep Arousal Scale (α = 0.91) [9], and daytime sleepiness by the Epworth Sleepiness Scale (α = 0.78) [10]. The sleep parameters were collected at both T1 and T2. At T2 only, the perceived COVID-19-related sleep change (CSC) was assessed by asking whether sleep had changed since the beginning of the COVID-19 pandemic. The item “I have sleep disturbances (e.g., falling asleep or sleeping through the night, nightmares)” was answered on a five-point Likert scale ranging from much less (−2) to unchanged (0) to much more (2). Based on the responses, three groups were distinguished: adolescents with subjectively positive CSC (responses of −2 or −1), adolescents without CSC (responses of 0), and those with negative CSC (responses of 1 or 2). The level of stress experience was assessed using the Adolescent Stress Questionnaire—Short version (α = 0.91) [11]. COVID-19-related distress was measured at T2 via the item “The current corona situation causes feelings of uncertainty or anxiety in many people. How much do you feel stressed by the current situation?” The response options varied on a five-point Likert scale ranging from not at all (0) to very much (4). Internalizing/externalizing symptoms and health-related quality of life (HQoL) were measured by the Youth Self Report (αInternalizing = 0.91; αExternalizing = 0.84) [12] and the Questionnaire for the Assessment of Health-Related Quality of Life in Children and Adolescents (KINDL; α = 0.86) [13]. Social (α = 0.91) and personal resources (α = 0.88) were assessed by the Questionnaire on Resources in Childhood and Adolescence [14]. The sleep-related self-efficacy scale measured one’s own coping skills for potential sleep disturbances (adapted from Kanis [15]; 10 items, response scale: I can’t manage that at all (0) to I am completely sure that I will be able to manage it (100), α = 0.91).

Statistical analysis

Missing values (2%) were imputed using MICE in R with variables utilized correlating ≤ 0.10 (R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria). The 10 resulting multiple datasets were aggregated, and total scores were calculated.

To address the first research question, a profile analysis was conducted by CSC group to test for profile differences at T2. The profile analysis was calculated using profileR. In case of significant profile deviations, these were subsequently explored using multivariate analysis of variance (MANOVA) and post-hoc ANOVAs with pairwise comparisons. The dichotomous variables sex and insomnia were examined for group differences using χ2 tests.

Regarding the second research question, a time (2) × group (3) MANCOVA with repeated measures and post-hoc ANCOVAs and pairwise comparisons were calculated, with age and sex as covariates. Partial η2 indicates a small (> 0.01), medium (> 0.06), and large effect (> 0.14), respectively.

Results

Regarding CSC, 18 adolescents were assigned to the group with a positive CSC (nine with slight and nine with strong improvements in sleep). At T2, 61% of them had clinically relevant sleep disturbances (SDSC sum score ≥ 39 [9]). For 62 adolescents, sleep had remained unchanged since the onset of the pandemic (no CSC), of whom 69% had clinically relevant sleep disturbances at T2. Twelve adolescents reported a negative CSC, with all adolescents in this group reporting mild worsening of sleep and 92% within the range of relevant sleep disturbances at T2.

Differences in mental health profiles during the COVID-19 pandemic (T2)

The profile analysis by group (positive/no/negative CSC) of health-related parameters (sleep indicators, stress, psychopathology, HQoL, resource-related parameters, age, and sex) showed group depending profile patters, e.g. their differed their mental health profiles, interaction group × mental health parameters, F(34, 148) = 2.08, p = 0.001, Pillai trace = 0.65. This finding was examined in more detail using MANOVA and ANOVAs with contrasts. The MANOVA indicted a main group effect, F(32, 150) = 2.10, p < 0.01, Pillai trace = 0.62, partial η2 = 0.31, revealing significant differences for age, sleep disturbances, SOL, pre-sleep arousal, stress experiences, COVID-19-related distress, HQoL, personal resources, and sleep-related self-efficacy (all p < 0.05; Table 1). Adolescents with positive compared to negative CSC differed regarding the abovementioned parameters (except SOL and sleep-related self-efficacy). Adolescents with positive compared to without CSC had lower COVID-19-related distress and higher personal resources (both p < 0.05).

Table 1 Means (M), standard deviations (SD) and comparisons for the three groups during COVID-19
Full size table

Differences in mental health before (T1) compared to during (T2) the COVID-19 pandemic

The time (T1/T2) × group (positive/no/negative CSC) MANCOVA with repeated measures revealed neither a significant main effect of time, F(14, 74) = 1.55, p = 0.11, Pillai trace = 0.23, partial η2 = 0.23, nor an interaction effect of time × group, F(28, 150) = 0.82, p = 0.73, Pillai trace = 0.27, partial η2 = 0.13. Consequently, there was neither a time effect nor a group-specific difference between the two measurement timepoints before the pandemic onset (T1) and after the first wave (T2) in terms of the mental health parameters studied in this sample. However, a significant group effect, F(28, 150) = 1.68, p = 0.03, Pillai trace = 0.48, partial η2 = 0.24, supports the above finding that adolescents with positive and/or without CSC showed more functional scores on mental health at both T1 and T2. The post-hoc ANOVAS again showed group differences for stress experiences, sleep disturbances, sleep onset latency, pre-sleep arousal, sleep-related self-efficacy, and personal resources (all p < 0.05), but not for HQoL.

Discussion

Several studies have highlighted an increasing prevalence of insomnia and mental health problems since the COVID-19 pandemic [2, 16]. Others found sleep improvements (e.g., increased sleep duration [4]). This study is specifically interested in the mental health profiles of adolescents with a perceived improvement or no change in sleep parameters since the onset of the pandemic to identify associated resources and relevant entry points for future prevention efforts. The results revealed a perceived positive change or no change in sleep to be associated with better overall sleep, lower stress, higher sleep-related self-efficacy, and higher perceived HQoL. Compared to adolescents without CSC, adolescents with positive CSC are characterized by more pronounced personal resources and less experienced COVID-19-related distress. It should be noted, however, that these results are partly due to group differences that already existed before the pandemic (as is shown by the longitudinal analysis).

Regarding sleep, as expected, shorter SOL, lower pre-sleep arousal, and fewer sleep disturbances overall were present among adolescents with positive CSC and/or without CSC. Surprisingly, significant differences in daytime sleepiness were not found. The flexibility provided by online class schedules could possibly provide compensation also for adolescents with negative CSC due to individually adjusted sleep times and durations and the opportunity to rest more frequently [17].

In line with other findings [18], adolescents with positive CSC/without CSC showed lower stress experiences as well as specifically lower COVID-19-related distress and higher HQoL. However, in contrast to the findings of Rosen et al. [19], where sufficient sleep in adolescents was associated with less psychopathology during the pandemic, this study found no difference in internalizing and externalizing symptoms. In our sample, 61% of the adolescents with a positive sleep change still reported relevant amounts of sleep disturbances and might therefore still suffer from internalizing/externalizing symptoms, which could explain these findings.

Most importantly, personal resources such as optimism and self-esteem were found to be higher in adolescents with positive CSC compared to adolescents with negative CSC or without CSC. It can therefore be assumed that either personal resources have suffered from subjectively poorer sleep and related difficulties since the pandemic or, conversely, that more pronounced personal resources play a protective role in the context of sleep disturbances. An association between personal resources and sleep in children and adolescents outside a pandemic situation has also been confirmed in several previous studies [20, 21]. As possible mechanisms for this link, Sadeghi-Bahmani and Brand [22] hypothesize that higher (psychological) resilience is associated with more efficient coping strategies as well as better emotion-regulation skills, which, in turn, might facilitate sleep.

It is interesting to note that sleep-related self-efficacy was highest in the group without CSC and not, as expected, in the group with positive CSC. It is conceivable that adolescents with positive CSC might attribute their improved sleep to external circumstances (such as online classes at home or more flexible daily structure) rather than to their own competence (self-efficacy).

Although the present study showed that various aspects of mental health are intertwined with positive CSC, the subsequent longitudinal analysis revealed that their extent did not differ before (T1) and during (T2) the pandemic (regardless of the group). Thus, adolescents who stated that they had slept better since the pandemic seem to have already been less burdened before the pandemic than those who reported negative CSC. Consequently, adolescents who were already less burdened before the pandemic thus seem to suffer less from sleep-related pandemic impacts. A similar assumption is made by Bujard et al. [23], who assume that a good mental and physical condition before the pandemic reduces the risk of developing depressive symptoms during the pandemic.

Limitations

Although the inclusion of cross-sectional and longitudinal data is a particular strength of this study, the findings should be interpreted in light of some limitations: 1) the sample size, particularly in connection with the unbalanced group sizes, is relatively small for the analyses, although post-hoc power analyses showed satisfactory power for all calculations performed; 2) the time interval from June 2020 to February 2021 includes different phases of the COVID-19 pandemic (e.g., a relaxation phase in late summer 2020 and a rise of incidences in autumn/winter 2020/21), which may affect interpretation of the results; 3) the assessment of CSC as a single-item indicator allows for limited reliable capture of the construct and of various facets of sleep, although a retrospective estimation can be an appropriate estimation method for indicating the perceived intensity [24]; 4) other relevant behaviors (e.g., exercise, media consumption) or traumatizing experiences (e.g., experiences of violence, illness, or death from COVID-19 within the family) should be considered in future studies within the framework of a multifactorial model of sleep and mental health.

Conclusion

In conclusion, the present study particularly emphasizes the relevance of strengthening personal resources (e.g., optimism) and sleep-related self-efficacy expectations in the context of sleep parameters, especially during times of crisis. Thus, the findings suggest a resource-oriented view of sleep disturbances and related domains of stress and HQoL as potentially helpful in clinical practice for prevention and intervention—both after a pandemic and to equip for possible future challenges.