Sleep during COVID‐19‐related school lockdown, a longitudinal study among high school students

Summary There has been great concern about the impact of coronavirus disease 2019 (COVID‐19)‐related school lockdown on adolescent health. The aim of the present study was to compare sleep patterns before and during COVID‐19‐related school lockdown, in a large sample of high school students. The present study is based a prospective, longitudinal survey on adolescent sleep health. Phase 1 was conducted in 2019, whereas phase 2 was conducted in 2020 (response rate 60.2%), during the last 10 days of a 60‐day long school lockdown. Main outcomes comprised sleep parameters from the Munich ChronoType Questionnaire (MCTQ). A total of 2,022 students provided valid responses to MCTQ in both survey phases. Results showed later sleep timing on schooldays in 2020 compared to 2019 (36 min later bedtimes, Cohen’s d = 0.56; 1:35 hr later rise times, Cohen’s d = 1.44). Time spent in bed on schooldays increased from 8:20 to 9:19 hr (Cohen’s d = 0.78), and sleep duration increased by 45 min (Cohen’s d = 0.49). The proportion of adolescents obtaining the recommended ≥8 hr of sleep on schooldays increased from 13.4% (2019) to 37.5% during the lockdown. Social jetlag was reduced from 2:37 hr (2019) to 1:53 hr (2020, Cohen’s d = 0.59). Results points to a potential advantageous effect of school lockdown in terms of increased school day sleep duration and reduced social jetlag. As sleep is important for mental and somatic health, it is conceivable that increased sleep duration offered some protection against harmful aspects of the COVID‐19 pandemic and associated social restrictions. Future studies should address possible associations between sleep changes and health during COVID‐19‐related school lockdown.

during the COVID-19 pandemic (Becker et al., 2021). Also studies in older and more diverse samples have shown increased sleep duration and reduced social jetlag during the COVID-19 pandemic, with the largest changes seen in younger people (Korman et al., 2020;Leone et al., 2020;Wright et al., 2020). Still, large scale prospective studies specifically addressing the impact of COVID-19 on sleep among adolescent high school students, the group who is likely to benefit the most from reduced social time pressure, are lacking from the literature. Furthermore, to our knowledge, no studies have to date addressed the effects of school lockdown on adolescent sleep in relation to circadian typology.
The aim of the present study was to compare adolescent sleep patterns between pre-pandemic times (spring 2019) and the initial phase of the COVID-19 pandemic when schools were in lockdown (spring 2020), overall and in relation to day (schooldays versus free days) and circadian typology (morning types versus intermediate types versus evening types), based on data from a longitudinal survey on sleep health in Norwegian high school students (the Western Norway Adolescent Longitudinal Sleep Study [WALOSS]). We hypothesised that school day sleep would be later and longer, and that social jetlag would be reduced during the lockdown as compared to spring 2019. We further hypothesised that these changes would be more pronounced in evening types compared to the other circadian types.

| Procedures
In spring 2019, all first year high school students aged ≥16 years in Hordaland and Rogaland counties in Norway were invited to participate in the WALOSS. The initial phase was facilitated by the county school authorities (schools were encouraged to allocate one school hour to respond to the questionnaire) and had a response rate of 42% (11,574 invited, 4,863 responders). Detailed procedures for phase 1 has previously been described elsewhere (Saxvig et al., 2020). The WALOSS cohort comprised the 3,736 responders who also consented to data linkage with the county school authorities.
For phase 2, invitations were sent through Short Message Service (SMS) and private email, and participants were required to respond to the questionnaire in their spare time. As an incentive to participate, all who completed phase 2 took part in a lottery where the price was five iPhone 11 Pros and 100 gift cards, each with a value of 500 Norwegian Kroner (NOK), ~$60 (American dollars). A total of 2,249 students responded in phase 2, yielding a response rate of 60.2%. The survey was technically administered by SurveyXact (Rambøll Management Consulting AS, www.surve yxact.no). The study was funded through a postdoctoral grant provided by the Western Norway Regional Health Authority (Helse Vest RHF). The funding source had no other involvement in the study.

| Settings
Phase 1 was conducted from April 10 through to June 20, 2019, whereas phase 2 was conducted from April 30 through to May 10, 2020, which represent the last 10 days of a 60-day long lockdown of Norwegian high schools (March 12 through to May 10). During the school lockdown period, students were required to attend remote classes, by means of digital media. Other coronavirus mitigation measures during this period included travelling restrictions and closure of arenas for sports, culture and recreation, and the population was encouraged to practice social distancing.

| Background information
The survey included items on date of birth, sex and parental education, and students were asked to report their normal school start times (phase 1) or start time for remote teaching (phase 2). In phase 2 they were asked if they were still high school students.

| Sleep
The sleep habits questionnaire included the MCTQ (Roenneberg et al., 2003), assessing habitual sleep for schooldays and free days separately. The Norwegian version of the MCTQ was adapted for children/adolescents for the purpose of the present study, based on the English version of the MCTQ for children/adolescents (www. thewep.org). In addition, the questionnaire included an item on wakefulness during the sleep period (WASO): "For how long are you awake during the night on schooldays/free days?". All responses were provided on drop-down menus, time items on 15-min interval scales and latency/duration items on 5-min interval scales. In the present study, we report bedtime, rise time, time spent in bed (interval from bedtime to rise time) and sleep duration (time in bed minus wakefulness in bed), for schooldays and free days, respectively, as well as social jetlag (calculated as the midpoint of sleep on free days [MSF] minus the midpoint of sleep on schooldays). For more details on how these parameters were computed, see Saxvig et al. (2020). In addition, as a measure of biological timing for sleep (chronotype), we calculated sleep corrected MSF (MSFsc) according to the formula: MSFsc = MSF − (sleep period on free days -average weekly sleep period)/2 (Roenneberg et al., 2019). The MSFsc was estimated only for those who did not use an alarm clock on free days, and when the sleep period on free days exceeded the sleep period on schooldays, see www.thewep.org for detailed description of the computation. Based on the recommendations by the USA National Sleep Foundation, sleep duration on school days was categorised into ≥8 hr (recommended), 7-8 hr (may be appropriate), and <7 hr (not recommended) (Hirshkowitz et al., 2015). Further, in order to visualise sleep duration in relation to circadian typology, the adolescents were grouped into seven categories based on their sleep duration: <5, 5-6, 6-7, 7-8, 8-9, 9-10 and ≥10 hr.

| Circadian typology
Circadian typology was measured using the short version of the Horne-Ostberg Morningness-Eveningness Questionnaire (rMEQ) (Adan & Almirall, 1991;Horne & Ostberg, 1976). The rMEQ is a brief five-item questionnaire on circadian preference that has been shown to correlate well with other measures of circadian typology (Di Milia et al., 2013). Scores range from 4 to 26, with higher scores indicating more morningness. For the present study, we used rMEQ data

| Statistics
The IBM Statistical Package for the Social Sciences (SPSS®) version 25 (IBM Corp., Armonk, NY, USA) was used for statistical analyses.
To address possible selection bias, background information (age, sex, maternal and paternal education) as well the main outcome measures (circadian typology, sleep duration, and social jetlag) were compared between respondents and non-respondents, using t tests for independent samples and chi-square analyses. Within the final sample, self-reported school start times in phase 1 were compared to start time for remote classes in phase 2 using paired samples t tests.
Alpha was set to 0.05.
Three-way analyses of variance (ANOVAs; year × day × type) were conducted for bedtime, rise time, time in bed and sleep duration to address overall differences in relation to year (2019 versus 2020), day (school day versus free day) and circadian type (morning types versus intermediate types versus evening types), whereas two-way ANOVAs (year × type) were conducted for social jetlag and chronotype. Alpha was set to 0.008 to implement Bonferroni correction for the six sleep parameters analysed (0.05/6). Significant year × day interaction effects were further explored using paired samples t tests (2019 versus 2020) for schooldays and free days separately, and the magnitude of change was explored by calculating in 2020), χ 2 (4) = 226.79, p < 0.001. School day sleep durations in 2019 and 2020 are illustrated in Figure 1. On free days, the mean bedtime was only 9 min later and rise time only 11 min later in 2020 compared to 2019 (small effect sizes) (Table 1), and time in bed and sleep duration did not change from 2019 to 2020 (Table 1).
Mirroring the large changes in school day sleep combined with small changes in free day sleep from 2019 to 2020, we found a significant effect of year for social jetlag, which was reduced by 44 min in 2020 as compared to 2019 (moderate effect size) (Table 1). We also found a significant effect of year for chronotype (calculated for 1,328 adolescents), but the delay was only 12 min (small effect size) (Table 1).
Addressing the role of circadian typology, results from the twoway ANOVA revealed year × type interaction effects for social jetlag and chronotype, whereas results from the three-way ANOVAs revealed year × day × type interaction effects for rise time, time in bed, and sleep duration (Table 1). Figure 2 illustrates social jetlag and school day sleep duration in 2019 and 2020, respectively, in relation to circadian typology. The post hoc tests revealed that ΔRise time, ΔTime in bed and ΔSleep duration on schooldays, as well as ΔSocial jetlag were of larger magnitude in evening types compared to intermediate and morning types (Table 2). Chronotype was significantly later in 2020 compared to 2019 in evening types only (p < 0.001 paired samples t tests), but ΔChronotype did not differ significantly between the circadian types.

| DISCUSS ION
The results from this longitudinal, prospective study showed later rise times, longer school day sleep duration, and reduced social jetlag in adolescent high school students during COVID-19-related school lockdown/remote teaching in 2020 compared to pre-pandemic times in 2019. Furthermore, these changes were of larger magnitude in evening types compared to the other circadian types.
The main difference in sleep from 2019 to 2020 comprised the ~1.5 hr delay in school day rise time. Although also bedtimes were delayed by ~0.5 hr, the adolescents were left with almost 1 hr longer time in bed, providing more opportunity to sleep. Accordingly, mean school day sleep duration increased by 45 min, and the proportion of adolescents who obtained the recommended ≥8 hr of sleep in-  and may not necessarily be comparable to late rise times due to late school start times under normal conditions. Furthermore, it is of note that during school lockdown in 2020, the mean rise time was only 11 min prior to the mean start time for remote teaching. Thus it appears that many adolescents attended remote classes directly upon rising (or even in bed), not spending time on breakfast or personal needs. One might worry that such behaviours may negatively affect health, well-being and learning outcomes, and we recommend this to be addressed in future studies.
In terms of circadian typology, adolescents in the present study were classified based on a questionnaire for circadian preference (rMEQ) (Adan & Almirall, 1991;Horne & Ostberg, 1976) completed in phase 1 (2019). Circadian preference is assumed to represent a relatively stable measure of circadian typology, although such preference generally changes with age (Randler et al., 2017). The positive effects of school lockdown appeared to be largest in adolescents with an evening preference (evening types), who were the ones with the latest and shortest sleep in 2019 ( Figure 2). Thus, it appears that potential positive effects of school lockdown in terms of sleep, increase with the degree of eveningness. Similar findings have been reported in a slightly older student sample, showing that those with shortest sleep duration prior to the pandemic experienced the largest increase in sleep during school lockdown (Wright et al., 2020).
It is crucial to stress that the impact of school lockdown on sleep is likely to be present only for the duration of the lockdown period. Whenever normal daily routines are resumed (i.e. re-opening of schools), the adolescents may slip back into old sleep patterns.
They may even be worse off, as suggested by recent evidence that delayed sleep and reduced social jetlag in a large Argentinian sample were accompanied by a delayed chronotype (Leone et al., 2020).
Chronotype as a concept is related to circadian typology, but reflects behaviour rather than preference (usually measured as MSFsc), and is thus more likely to change rapidly according to social demands (Leone et al., 2020;Roenneberg et al., 2019). In the present study, after 50-60 days of school lockdown/remote classes, there was only a slight delay in chronotype of ~12 min (small effect size), and free F I G U R E 1 The proportion of high school students obtaining <5, 5-6, 6-7, 7-8, 8-9, 9-10 (Cellini et al., 2020;Korman et al., 2020;Leone et al., 2020;Wright et al., 2020), but also poorer sleep quality, reduced sleep efficiency, and increased sleep problems (Benham, 2020;Cellini et al., 2020). A combination of increased difficulties initiating sleep and longer school day sleep duration was also reported by Becker et al., in their prospective study among a small sample of adolescents (Becker et al., 2021). Thus, the COVID-19-related school lockdown may seem to represent a double-edged sword, where some perish, and others prosper from loss of daily routines and social demands. Following this line of reasoning it seems important that future studies explore whether (and which) sleep markers can be useful when seeking to identify adolescents at risk for negative health outcomes in relation to the COVID-19 pandemic and associated social restrictions.

| Strengths and limitations
The strengths of the present study include its prospective longitudinal design and the relatively large, population-based sample, as well as the use of validated questionnaires to address sleep and circadian typology. Moreover, both survey phases were administered at approximately the same time of year (spring), thereby reducing the possible impact of season on sleep. Another strength is that the questionnaire allowed us to account for wakefulness in bed and thus compute actual time spent asleep, according to self-report.
Still, the present study has some limitations that should be considered when interpreting the data. First, in terms of representativeness, participation in phase 1 was facilitated by the schools, allocating 1 school hour to respond the questionnaire, whereas participants in phase 2 were required to respond to the questionnaire in their spare time. This procedure yielded a response rate in phase 2 of ~60%.
Participation in phase 2 appeared slightly biased, with responders being slightly younger, more often female, and fewer had mothers with university/college education. Responders were also less likely to be evening types, and they had longer school day sleep duration and less social jetlag at phase 1 compared to the non-responders.