3.1 Did children’s sleep behaviour and/or anxiety change at the start of lockdown?
3.1.1 Parent reported sleep
Parents were asked questions about their child’s sleep ‘during the past week (while schools were closed)’ and also ‘during a typical week (when schools were open)’ before the COVID-19 pandemic. Parents were asked What time does your child go to bed on weeknights? and What time does your child wake up on weekdays?. Options were given to the nearest hour as bedtimes are known to vary substantially28 and parents are approximate in their estimates of their children’s sleep times even over the primary school years29. Mean bed time became significantly later at the start of lockdown (mean = 8.37pm, sd = 66.6 mins) relative to a typical week of school (mean = 7.50pm, sd = 49.2 mins; t(529.8)= -9.675, p < 0.001) as did mean wake-up time (6.39am (sd = 35 mins) to 7.21am (sd = 50mins); t(409.83) = -8.755, p < 0.001) (Figure 3). However, total sleep time (TST), calculated to the nearest hour based on bed time and wake-up times, did not change: TST during a typical week was 10 hours 50 minutes (sd = 51.7 minutes), and during the surveyed lockdown week, 10 hours 43 minutes (sd = 61.2 minutes). Age in months predicted change in both bed time (B = -0.003, t = -2.13, p = 0.034) and wake-up time (B = -0.012, t = -6.57, p < 0.001), with younger children being less likely to show change in patterns of behaviour.
3.1.2 Parent reported anxiety
We explored parent-reported anxiety for both parents themselves and their children (see Table 1). A notable shift is evident in reported bedtime anxiety, with 61.8% of children being described as ‘not at all anxious’ at bedtime before the lockdown, dropping to 51.0% at the start of lockdown.
We considered whether the change in bedtime anxiety from before lockdown to the start of lockdown could be predicted by the child’s general anxiety level, the extent to which they had expressed worry about COVID-19 to their parents, days since lockdown began, the child’s age and whether anyone in the household had experienced symptoms at the time of parents’ responses. In an ordinal logistic regression, every step up in children’s expressed COVID-19 anxiety increased the proportional odds of them showing an increase in bedtime anxiety. Most notably, with a shift in a child’s COVID-19 related anxiety from ‘Quite a bit’ to ‘A lot’, children were 56.3 times more likely to show an increased level of bedtime anxiety (Table 2).
3.2 Did anxiety and sleep change over the course of lockdown?
3.2.1 Sleep
At Phase 2, mean bed time was 8.32pm (sd=64 minutes), and wake up time was 7.15am (sd = 66 minutes), with a mean TST of 10 hours 43 minutes (sd=60 minutes). Compared with the start of lockdown none of these values significantly changed (bed time t(355.62) = 0.957, p = 0.339; wake time t(353.9) = 0, p=1.000; TST t(355.79)=-1.016, p=0.310).
At both phases, parents were asked additional questions regarding their child’s sleep: How long do you think it takes your child to fall asleep after lights are turned out? (this question will hereafter be referred to as sleep onset latency; SOL); and Does your child seem sleepy during the day? (for responses see Figure 4). Finally, relating to sleep behaviour, parents were asked Are you currently worried about your child's sleep?: at Phase 1 39.2% reported yes, with the remaining 60.8% reporting no. At Phase 2, 38.0% reported that yes they were, with the remaining 62.0% reporting no.
No significant differences were evident from the start to the end of lockdown for daytime sleepiness (for n=179; Wilcoxon V = 1750.5, p=0.973) or whether parents were worried about their child’s sleep (Wilcoxon V = 644, p = 0.227) but a difference did emerge for SOL, for n=176, Wilcoxon V = 1418, p=0.024, with median response at both phases ‘10-30 minutes’ (option 3) but with a larger IQR at Phase 2 (IQR = 1 vs IQR = 1.5); as is evident in Figure 4, fewer parents reported that their children took ’30-60 minutes’ or ‘more than 60 minutes’ to fall asleep at Phase 2 compared to Phase 1.
Parents were asked if they felt their child’s sleep had changed during lockdown, 43.0% of respondents said ‘yes’, 25.7% ‘maybe’ and 31.3% ‘no’. Those parents who reported that ‘yes’ or ‘maybe’ their child’s sleep had changed (68.7% of the whole sample), were asked to select from one or more of seven possible reasons for that change: ‘Not tired in the evenings’ (35.0%); ‘Taking longer to fall asleep’ (66.7%); ‘Needing someone else in the room’ (30.1%); ‘Waking in the night’ (34.1%); ‘Nightmares/night terrors’ (22.0%); ‘Early morning waking’ (17.9%); ‘Daytime sleepiness’(18.7%); ‘Other’ (15.4%). Parents who said their child’s sleep changed over lockdown were then asked if their child’s sleep had returned to normal. Only 7.3% said ‘yes’, with 39.8% reporting ‘somewhat’ and 52.8% ‘no’.
Children themselves were asked if they would be happy to complete a few questions regarding their sleep and feelings about the pandemic. 13.2% did not consent to answer the questions and were consequently not asked to complete any. Where a percentage of ‘no response’ is given, this does not include those who opted to not respond to any questions in this section; rather, responses are detailed here for the 243 children who answered at least some questions. Children were asked questions about their sleep at both phases (Figure 5): At the moment, do you feel you get enough sleep? and At the moment, how sleepy do you feel during the day?, with responses to these questions strongly correlated, rs=0.53, p<0.001; After your lights are turned off, do you spend a long time thinking or worrying about things? Children’s responses to whether they get enough sleep did not significantly change from Phase 1 to Phase 2, based on 137 respondents (Wilcoxon V = 1509.5, p = 0.968). For child-reported SOL, 134 respondents answered at both phases, over which time median response at Phase 2 = 3 (IQR = 1.5), at Phase 1 = 3 ‘I spend some time thinking’ (IQR=1, Wilcoxon’s V=1226.5, p=0.004). The correlation between child-reported enough sleep and daytime sleepiness was again significant, rs=0.71, p<0.001.
Children were asked How many times do you think you wake up in the night? Mean response at Phase 2 (134 responses) was 1.37 (sd = 1.27), significantly less than at Phase 1, where mean = 1.63 (sd = 1.48), t(133) = -2.49, p = 0.014. Finally regarding wake after sleep onset, children were asked If you wake in the night does it take you a long time to get back to sleep?: again, this measure changed significantly from Phase 1 to Phase 2 for the 134 respondents who answered this question twice. At Phase 1 median response = “some time” (IQR = 3) to Phase 2 median response = ‘very little time’ (IQR = 2), Wilcoxon’s V = 1133.5, p=0.003.
3.2.2 Anxiety
At Phase 2, parents were asked about current bedtime anxiety and COVID-related anxiety in their child, as well as their own current COVID-19-related anxiety (Table 3). All three of these anxiety measures improved significantly from Phase 1 to Phase 2. Child bedtime anxiety improved from a median response ‘a little anxious’ (IQR = 2) to a median response ‘not at all anxious’ (IQR = 1; n = 179); Wilcoxon V = 821, p<0.001. Children’s anxiety about COVID-19 changed from a median response ‘a little anxiety’ (IQR = 1) to ‘no anxiety’ (IQR = 1), for 179 respondents, Wilcoxon’s V = 604, p < 0.001. Finally, parental anxiety about COVID-19 also decreased, from a median response ‘somewhat anxious’ (IQR = 2) to a median response ‘a little anxious’ (IQR = 2), for 179 respondents, Wilcoxon’s V=834, p<0.001.
Children were asked about anxiety: During the past week, did you feel scared or worried for no particular reason? and During the past week, did you worry about the virus?: see Table 5 for responses from those participants who responded on both occasions. General worry did not significantly change from Phase 1 to Phase 2 (Wilcoxon’s V = 1419, p=0.067), while COVID-19-specific worry did with median response at Phase 1 being I ‘sometimes’ worry, to I ‘rarely’ worry at Phase 2 (Phase 1 IQR=2, Phase 2 IQR=2): Wilcoxon’s V = 966, p<0.001 (see Table A2 in Additional File 2 for full breakdown of responses).
3.3 Did anxiety predict sleep behaviour at the start and/or end of lockdown?
3.3.1 Phase 1
We asked whether children’s TST or SOL, as reported by their parents, could be predicted by children’s anxiety levels at bedtime, along with household symptoms, age, PPB and whether the child had been diagnosed with or been referred for an NDD. Table 3 gives the results of ordinal logistic regressions, suggesting that bedtime anxiety is the best predictor of SOL during lockdown, with every step up in children’s bedtime anxiety increasing the proportional odds of it taking longer for them to fall asleep after lights out (Figure 6). Conversely, age was the most reliable predictor of TST.
As the relationships between parent-reported child anxiety and sleep were strong, we assessed the relationships between worry and sleep parameters as reported by children themselves at Phase 1 (Table 4). The correlation between worry about the virus and child-reported sleep onset latency was particularly strong, similar to the parent-reported data.
3.3.2 Phase 2
To establish whether anxiety at Phase 2 continued to predict children’s sleep parameters, we ran a second regression model predicting parent-reported SOL and TST at Phase 2, with child bedtime anxiety, and child anxiety about COVID-19 at Phase 2, along with whether the household reported symptoms at Phase 2, whether the child had a NDD, PPB and child age at Phase 2 as predictors (see Table 5). Bedtime anxiety remained the only predictor of SOL at Phase 2, as it was at Phase 1, although odds ratios were lower. For TST, age remained a significant predictor, along with a shift in bedtime anxiety from ‘somewhat anxious’ to ‘anxious’, which was associated with shorter TST.