Sleep duration during COVID-19 lockdown: systematic review and meta-analysis

In attempts to slow down the spread of COVID-19 pandemic, countries took different measures including lockdowns. Several studies reported the effects of lockdowns on sleep mostly suggesting an increase in sleep duration during lockdowns. This is an interesting outcome as life in the modern world were associated with insu�cient sleep. The present study is a systematic review and meta-analysis of work regarding sleep duration during lockdowns. We searched PubMed using the terms “COVID-19”, “SARS-CoV-2”, “lockdown”, “stay at home”, “quarantine” and “sleep” and found 262 papers. After applying exclusion criteria, we identi�ed 26 papers (with 35 discrete studies) for the meta-analysis. A �xed effect model was used with effect sizes estimated using Cohen’s d and heterogeneity of effect sizes tested using the Cochran’s Q test statistics. Most studies (31/35) report an increase in sleep duration during lockdown periods with a mean (± STD) of 31.5 ± 21.2 minutes. Effect sizes (Cohen’s d) ranged between 0.08 and 3.34, which are considered between very small and huge. Q statistics was 4859.3 (p < 0.0001) and with no indication for publication bias (Egger’s test, p = 0.45). Eight studies also reported measures of “sleep quality” with 6/8 reporting worsening of sleep quality with small to large effect sizes Cohen’s d range between 0.09 and 1.86). The study demonstrates increased sleep duration during lockdowns. Increased sleep duration is a positive change, but effects on well-being during lockdowns might be overshadowed by reduced sleep quality, increased stress, and lowered mood. It is interesting whether longer sleep duration habits may survive to post-COVID-19 days.


Introduction
Starting December 2019 our world experienced a major change with the appearance and fast spread of the COVID-19 virus.In attempts to slow down the spread of the disease, many countries took different measures at different times imposing social distancing regulations and mobility restrictions such as lockdowns (Caristia et al., 2020).Clearly, these regulations affected human behavior including sleep and eating habits, travel and social activities, with signi cant damaging consequences (Kumar et al., 2020;Schippers, 2020).One aspect of change that might be of signi cant importance is sleep, including sleep duration, timing and quality and indeed a signi cant number of studies in the last year examined sleep parameters in the context of the COVID-19 regulations [e.g.(Altena et al., 2020;Cellini, Canale, Mioni, & Costa, 2020)].
Whereas many studies emphasized the adverse effects of lockdowns on aspects of human health and behavior, including increased anxiety and stress (Oved et al., 2021;Violant-Holz et al., 2020), reduced activity (Violant-Holz et al., 2020) and reduced socialization, the results regarding sleep are more complex.Several studies concentrated on sleep disturbances during the pandemic.A recent metaanalysis based on 44 papers, involving a total of 54,231 participants from 13 countries reports that the prevalence of sleep problems during the COVID-19 pandemic is up to 40% of people from the general population (Jahrami et al., 2021), whereas another meta-analysis based on 6 studies and limited to India reports a lower prevalence of 26.7% (Singh, Bajpai, & Kaswan, 2021).These numbers may appear high, but it is important to remember that many studies show that even during "normal" days there is a high rate of sleep disturbances in the general population.For example, a meta-analysis of sleep quality in students (conducted before the COVID-19 outbreak) shows a 25% prevalence of subjects reporting low sleep quality (Chen, Zhang, & Zhou, 2020).
In general, it is repeatedly suggested that people in the modern society do not sleep enough (Chattu et al., 2018) and that disturbed and insu cient sleep can lead to a wide range of problems and pathologies ranging from increased chances for heart disease (Shi et al., 2021) to heightened emotional susceptibility (Hairston & Cohen-Zion, 2020).Within the range of sleep problems and disturbances, life in the modern world were repeatedly associated with insu cient sleep (CDC, 2012;Chattu et al., 2018) and insu cient sleep and other circadian and sleep disturbances had been repeatedly associated with a large variety of signi cant adverse effects ranging from somatic to mental diseases including (but not limited to) diabetes, hypercholesterolemia, heart disease, depression and more (Chattu et al., 2018;Shi et al., 2021;Zimmet et al., 2019).In that context, it is possible to suggest that amongst a range of adverse effects, lockdowns may also have at least one positive consequence, increased sleep duration.Yet, the results in different studies vary signi cantly and it is not simple to make general conclusions based on one or even a few studies.
To further explore the effects of COVID-19 lockdowns on sleep duration, the current study presents a systematic review and meta-analysis of recent studies that report changes in the duration of sleep in the general population.

Search and selection of papers
We searched PubMed (https://pubmed.ncbi.nlm.nih.gov/) using the terms "COVID-19", "SARS-CoV-2", "lockdown", "stay at home", "quarantine" and "sleep" and found 262 papers.Their abstracts were screened and only papers written in English were selected.Additional exclusion criteria included studies of individuals a icted with any disorder or hospitalized or institutionalized.Studies of speci c age groups were not excluded as long as they were performed in the general population.Following this initial screening, we further searched for studies that reported direct measures of sleep duration.After applying these exclusion criteria, we identi ed 26 reports (with 35 discrete studies) to be included in the nal metaanalysis.
Of these studies, four experiments were conducted in children (of various ages), 3 in youth or late teens and the rest of the experiments in adults at various ages.In four of these studies (including eight discrete experiments) data collection was performed using wearable devices whereas in all other studies the data were collected using self-reported surveys and questionnaires.

Statistical analysis
A xed effect model was used in the meta-analysis.The xed effect model meta-analysis takes into account only one level of sampling as all studies are sampled from a population with effect size and therefore only within studies sampling error needs to be considered (Borenstein, Hedges, Higgins, & Rothstein, 2010).
The effect size for each experiment was estimated using Cohen's d, an unbiased measure of the difference between two means (Cohen, 1992).Cohen's d is calculated by dividing the difference between the two groups by their pooled standard deviations.Cohen's d is an advantageous measurement for effect-size.First, this test is gradually becoming the standard and therefore it allows comparison with a large number of published studies in a variety of research areas.Moreover, Cohen's (Cohen, 1992) suggestion that effect sizes of 0.20 are small, 0.50 are medium, and 0.80 or more are large enables comparison to some standard measures (Rosnow, Rosenthal, & Rubin, 2000).
Heterogeneity of effect sizes within each comparison was tested using the Cochran's Q test statistics (Higgins, Thompson, Deeks, & Altman, 2002).A test for heterogeneity examines the null hypothesis that all studies are evaluating the same effect.The usual test statistic (Cochran's Q) is computed by summing the squared deviations of each study's estimate from the overall meta-analytic estimate, weighting each study's contribution in the same manner as in the meta-analysis.The p values are obtained by comparing the statistic with Chi square distribution with k − 1 degrees of freedom (DerSimonian & Laird, 2015).
Calculations of Cohen's d and Cochran's Q were performed using Wolfram Mathematica 11.

Results
As shown in Table 1, 31/35 studies showed a signi cantly longer sleep duration during lockdown compared with pre-lockdown periods, whereas non-signi cant effects were demonstrated in four studies.
Mean increase in sleep duration across all studies was 31.5 ± 3.6 (mean ± SEM) minutes and the range was 1.0 min to 96.0 minutes.Effect sizes (Cohen's d) for the studies with signi cant effects ranged between 0.08 and 3.34 (Fig. 1).Mean effect size across all experiments comparing pre-lockdown to lockdown periods was 0.58 ± 0.11 (mean ± SEM) which is considered a medium effect size (Rosnow et al., 2000).Q statistics was 4859 (p < 0.001).There was no indication for publication bias in these studies (Egger's test, p = 0.22).Seven of the studies examined children and youth (Table 1).Teenagers are expected to sleep longer than adults regardless of lockdown conditions (Kocevska et al., 2021;Ohayon, Carskadon, Guilleminault, & Vitiello, 2004) and this was the case here with the measures of sleep duration before lockdown where mean (± SEM) for adult studies was 446.3 ± 7.5 min and for the children and youth studies 522.7 ± 20.7 min [t-test: t(33) = 5.91, p < 0.001].However, despite the differences in baseline, no differences between young and adult populations were demonstrated for the effects of lockdown on sleep duration.Six of the studies in young populations showed signi cant elongation of sleep duration during lockdown.Effect sizes in these studies ranged between 0.08 and 1.53, not different than effect sizes for the studies conducted in adults [t-test: t(33) = 0.18, p = 0.86].
Eight of the 35 studies also reported direct measures for sleep quality using either the PSQI questionnaire or measure of sleep e ciency.As shown in Table 2, six out of these papers report worse sleep quality during lockdown compared with pre-lockdown periods with a medium mean effect size (Cohen's = 0.68).

Discussion
Short sleep duration and sleep disturbances are common in modern life and had been associated with a large scope of problems ranging from reduced functionality, well-being and increased susceptibility to many modern world diseases (CDC, 2012;Chattu et al., 2018;Shi et al., 2021;Zimmet et al., 2019).It is therefore interesting to note that amongst the multiple consequences of COVID-19 related lockdowns (Kumar et al., 2020;Schippers, 2020), one change that emerges as possibly positive is an increase in sleep duration.This change is emphasized in the current systematic review where nearly all studies demonstrated signi cant increase in sleep time.
Unfortunately, the possible advantages of increased sleep in these conditions may not be clear because several studies also report a reduction in sleep quality (Table 2).Accordingly, it is not clear if the advantage of sleeping more is not overshadowed by the reduction in the quality of sleep.Yet, some optimism could be appropriate as at least one study suggest that sleep quality recovers when the lockdown ends (Beck, Leger, Cortaredona, Verger, & Peretti-Watel, 2020).
One question regarding the studies of sleep is how well they re ect the actual situation rather than the subjective feeling of subjects.Most of the studies examined in this meta-analysis are based on selfreported questionnaires which suggests at least some level of inaccuracy in reporting past and present sleep duration (Matricciani, 2013).Only three papers (that included reports of 7 discrete studies) were based on data obtained using wearable devices or smartphone apps or both.A separate analysis of these few studies resulted in a smaller sleep duration increase (mean = 15.1 min) compared with the analysis of survey based studies (mean = 31.3).These differences may suggest that while answering sleep questionnaires, individuals may overestimate the changes that happen during lockdowns.However, as the number of studies that used wearable devices is small it is not possible to make any strong conclusion on this issue and it should be further investigated.Interestingly, a recent study that compared self-reports (using a dedicated smartphone app) with data from wearable devices during the COVID-19 lockdown reported shorter subjective sleep duration (app) than objective sleep duration (wearable devices) both before and during lockdown but the increase in sleep during lockdown was similar for subjective and objective measures (Oved et al., 2021).This same report (Oved et al., 2021) suggests differential effects of lockdown in sub-populations.Speci cally, the study shows that both age and chronotype are factors that in uence change in sleep duration under lockdown conditions.For the older age group, the increase in sleep duration was mainly expressed in early chronotypes, whereas for the younger age group, it was mainly expressed in late chronotypes (Oved et al., 2021).These data are important as they demonstrate that whereas most studies address the general population, further exploration of the consequences of lockdowns and other measures taken during the COVID-19 epidemic should really explore speci c sub-populations.
In summary, the current meta-analysis suggests that amongst many other effects, lockdowns result in increased sleep duration but that this increase varies between studies.It is an open question how life in these odd times will affect future habits and human behavior and whether the habit of sleeping more will stay with us after the epidemic will end.