Original research

Arch Dis Child: first published as 10.1136/archdischild-2020-320372 on 9 December 2020. Downloaded from http://adc.bmj.com/ on March 21, 2021 by guest. Protected by copyright.
Longitudinal increases in childhood depression
symptoms during the COVID-19 lockdown
Giacomo Bignardi  ‍ ‍, Edwin S Dalmaijer  ‍ ‍, Alexander L Anwyl-­Irvine  ‍ ‍,
Tess A Smith  ‍ ‍, Roma Siugzdaite  ‍ ‍, Stepheni Uh  ‍ ‍, Duncan E Astle  ‍ ‍

MRC Cognition and Brain ABSTRACT

Sciences Unit, University of What is already known on this topic?
Objective  There has been widespread concern that
Cambridge, Cambridge, UK
so-­called lockdown measures, including social distancing
►► Due to a lack of prospective studies with
and school closures, could negatively impact children’s
Correspondence to before-­lockdown assessments, the impacts
Dr Duncan E Astle, MRC mental health. However, there has been little direct
of physical distancing and school closures on
Cognition and Brain Sciences evidence of any association due to the paucity of
children’s mental health is unclear.
Unit, Cambridge CB2 7EF, UK; longitudinal studies reporting mental health before and
​Duncan.​Astle@m
​ rc-​cbu.​cam.​ ►► Adolescence is a developmental period where
during the lockdown. This present study provides the first
ac.u​ k mental health may be particularly vulnerable to
longitudinal examination of changes in childhood mental
reduced peer interaction and loneliness.
Received 31 July 2020 health, a key component of an urgently needed evidence
►► Existing research in adult populations suggests
Revised 12 October 2020 base that can inform policy and practice surrounding the
Accepted 5 November 2020 deteriorations in mental health during the
continuing response to the COVID-19 pandemic.
lockdown policies in different countries.
Methods  Mental health assessments on 168 children
(aged 7.6–11.6 years) were taken before and during the
UK lockdown (April–June 2020). Assessments included
self-­reports, caregiver reports, and teacher reports. Mean What this study adds?
mental health scores before and during the UK lockdown
were compared using mixed linear models. ►► Changes in three mental health scales before
Results  A significant increase in depression symptoms and during the lockdown are analysed used
during the UK lockdown was observed, as measured by mixed linear models in a UK cohort of 8–12 year
the Revised Child Anxiety and Depression Scale (RCADS) olds.
short form. CIs suggest a medium-­to-­large effect size. ►► Depression symptoms increased during
There were no significant changes in the RCADS anxiety lockdown, with CIs suggesting a medium-­to-­
subscale and Strengths and Difficulties Questionnaire large standardised mean difference, even when
emotional problems subscale. controlling for age at assessment.
Conclusions  During the UK lockdown, children’s ►► Changes in anxiety and emotional problems
depression symptoms have increased substantially, were small and not statistically significant,
relative to before lockdown. The scale of this effect suggesting that depression may be particularly
has direct relevance for the continuation of different susceptible.
elements of lockdown policy, such as complete or partial
school closures. This early evidence for the direct impact
of lockdown must now be combined with larger scale clinically meaningful deteriorations in anxiety,
mental health and well-­being during lockdown.4–7
epidemiological studies that establish which children are
most at risk and tracks their future recovery. One study suggests younger and lower income indi-
viduals have been more affected.8 However, one
longitudinal study from a large Dutch probability
sample of adults found a very small, non-­significant
BACKGROUND change in depression and anxiety, measured with
In response to the COVID-19 pandemic, the UK the five-­item Mental Health Index.9
Government implemented a national ‘lockdown’ Far less is known about how young children
involving school closures and social distancing. have coped during lockdown, though evidence is
There has been widespread concern that these rapidly emerging for adolescents. A large, longitu-
measures will negatively impact child and adoles- dinal study of 13–14 year olds in the UK reported
cent mental health.1–3 To date, however, there is a mixed pattern of changes in well-­being, depres-
© Author(s) (or their relatively little direct evidence of this. This is in sion and anxiety in April/May 2020 compared
employer(s)) 2020. Re-­use part due to the paucity of studies including pre-­ with October 2019.10 Self-­reported well-­being and
permitted under CC BY.
Published by BMJ. lockdown baseline data. Longitudinal changes anxiety slightly improved during the lockdown
within the same individuals represent the most compared with before. However, the analyses do
To cite: Bignardi G, direct way of quantifying the association between not control for age at assessment or report mean
Dalmaijer ES, Anwyl-­Irvine AL,
the onset of lockdown and children’s mental health. change scores for depression and anxiety for the
et al. Arch Dis Child Epub
ahead of print: [please There is good reason to suspect that the imple- whole sample.
include Day Month Year]. mentation of a lockdown has negatively impacted How the lockdown measures impact children’s
doi:10.1136/ children’s mental health. Early evidence from mental health may depend on a variety of factors.
archdischild-2020-320372 adult cohort studies suggests that there have been Loneliness in children is associated with subsequent
Bignardi G, et al. Arch Dis Child 2020;0:1–7. doi:10.1136/archdischild-2020-320372    1
 Original research

Arch Dis Child: first published as 10.1136/archdischild-2020-320372 on 9 December 2020. Downloaded from http://adc.bmj.com/ on March 21, 2021 by guest. Protected by copyright.
mental health problems, particularly depression.11 Reduced
Table 1  Demographic data for both subgroups
access to play and activities for young people may impair mood
homeostasis, engaging in pleasurable activities to improve School group Lab group
mood.12 There is a particular concern for children already strug- Sample size
gling with mental health issues, where access to mental health  N 114 54
services has been impacted.13 14 Using social media may also Gender
mitigate the impacts of physical distancing.3 Its also plausible  Male 58 22
that the alleviation of school-­related stressors may elicit short-­  Female 56 32
term improvements, although we are not aware of any existing Age at baseline
evidence yet.  Mean 8.7 8.5
We report results from the Resilience in Education and Devel-  SD 0.63 0.66
opment (RED) study, a small but rich dataset collected from a
Age at lockdown assessment
cohort of children living in the East of England.15 This cohort
 Mean 10.5 9.4
had been assessed via a combination of caregiver, teacher and
 SD 0.74 0.78
child reports of mental health, alongside a variety of other
Caregiver is homeowner
measures. Around 18 months after this initial assessment, these
children were subject to the national lockdown. During lock-  % (N) 73% (83) 63% (34)
down, we contacted a subsample of the families and tested for Caregiver has degree
any changes in their children’s mental health and well-­being.  % (N) 64% (67) 60% (32)
This study aims to test whether changes in emotional well-­being,  Number of responses 105 53
anxiety and depression occurred during lockdown since the Index of Multiple Deprivation
initial assessment.  Mean decile 7.9 6.9
Only children included in one of the mixed linear model analyses (with both
baseline and during lockdown mental health data) are presented here.
METHODS
Participants
The RED study comprises two groups. A larger school group
assessed in classrooms (n=567, from 22 classes, 6 schools) and and RCADS-­short form subscales for depression and anxiety.16 17
a smaller group of children (N=92) who completed the same All scales were adapted for computerised testing using contin-
and additional assessments at our laboratory. Both samples are uous slider scales.
convenience samples. Families in the lab group were recruited via Before lockdown, teachers and caregivers completed the SDQ
posters, word of mouth and online Facebook advertisements. In for the school and lab groups, respectively. At this time point,
the school groups, all children in year 3 and 4 classroom groups children in both groups completed the RCADS, along with care-
were recruited into the study using opt-­out parental consent. givers in the lab group. Children completed the RCADS on a
Due to ethical constraints, schools did not provide information custom-­ developed tablet application,15 which included audio
on the number of ‘opt-­outs’ from their schools. Children absent presentation of each question. Follow-­up testing during lock-
on the day of before-­lockdown testing (ie, due to sickness) will down took place between 29 April and 19 June 2020, with only
not have data on child-­reported mental health. Schools did not caregivers completing the assessments.
provide information on whether children have moved by the We collected several demographic variables from families.
time of lockdown. Neighbourhood deprivation was estimated using the English
Baseline assessments occurred between June 2018 and March Indices of Deprivation, a national statistics database that ranks
2019 in the school group and December 2018 and September small areas in England from most (1) to least (10) deprived
2019 in the laboratory group. Six schools were recruited to deciles.18 Free school meal eligibility, a widely used proxy for
take part in the study, with all children in eligible year 3 and 4 socioeconomic status (SES), measures whether parents are
classroom groups eligible to take part. Eligibility criteria in the eligible for a series of government benefits.19 Caregiver educa-
lab group included a medical screener for suitability to undergo tion and homeownership were also assessed in the lockdown
medical resonance imaging. questionnaire.
The mental health assessments by caregivers and teachers, both
before and during lockdown, were completed using an online
Statistical analysis
survey. Participation was incentivised with a £5 Amazon voucher
We report descriptive statistics and correlations between
for completion. We directly contacted all legal caregivers of chil-
measures in figure 1, which includes all participants. We anal-
dren in the lab group, and five schools contacted caregivers in the
ysed the impact of lockdown by combining child, teacher and
school group, to complete the survey. One school (representing
caregiver reports using linear mixed models. Coefficients esti-
84 children with baseline data) did not contact caregivers.
mated the effect of lockdown (0=before/1=during lockdown),
One hundred and sixty-­eight parents completed mental health
and responder (0=child/1=caregiver  or 0=teacher/1=care-
assessments for their children during lockdown (142 mothers and
giver), on children’s mental health, including a random inter-
26 fathers, no responders selected ‘grandparent’ or ‘other’), for
cept for participant. Children were included in a given mixed
whom prior mental health data were available. This represents
linear model only if data from before and during lockdown
29% of the contacted, eligible sample. Demographic features of
were available for a given mental health outcome. Participant’s
this sample are summarised in table 1.
age, gender and SES were controlled in sensitivity analyses. SES
was measured using a mean of: household income, homeown-
Measures ership, caregiver education and neighbourhood deprivation.
Three mental health measures were used: the Strengths and SES was scaled to have zero mean and unit variance. Interaction
Difficulties Questionnaire (SDQ), Emotional Problems subscale effects between age, gender and SES and lockdown status were
2 Bignardi G, et al. Arch Dis Child 2020;0:1–7. doi:10.1136/archdischild-2020-320372
 Original research

Arch Dis Child: first published as 10.1136/archdischild-2020-320372 on 9 December 2020. Downloaded from http://adc.bmj.com/ on March 21, 2021 by guest. Protected by copyright.
Figure 1  Correlations between mental health variables and patterns of missing data. Below diagonal: Pearson correlations between SDQ Emotional
Problems (SDQ), RCADS anxiety subscale (generalised anxiety disorder (GAD)) and RCADS depression subscale (major depressive disorder (MDD)),
before and during lockdown. On diagonal: number of observations for each variable. Above diagonal: number of observations with data on two given
variables and 95% CIs for correlations. RCADS, Revised Child Anxiety and Depression Scale; SDQ, Strengths and Difficulties Questionnaire.

examined, by multiplying lockdown status with these variables Multiple Deprivation) and free school meal eligibility weakly
and entering them into the mixed model. predicted non-­ participation (r=−0.17 and −0.18, respec-
Analyses were performed using R (V.3.6.2) and the nlme tively).18 Greater child-­reported RCADS depression (r=−0.08,
(V.3.1–144) package.20 21 Mental health variables were scored 95% CI −0.16 to 0.01) and anxiety (r=−0.09, 95% CI −0.16
using the arithmetic mean response, after recoding each item so to 0.01) symptoms at baseline also weakly predicted non-­
that a higher score indicated worse mental health. For mixed participation. The date of lockdown questionnaire completion
linear models, all three mental health outcomes were quantile-­ had weak, non-­significant correlations with the three mental
normalised to match a standard normal distribution. Normali- health outcomes during lockdown (all |r|<0.06).
sation was performed for each outcome after transforming data Correlations and 95% CIs between all variables are provided
into a ‘long’ format, with repeated measurements (including in figure 1, alongside patterns of available data and missingness.
before and during lockdown from all raters) gathered in a single All three mental health measures during lockdown were strongly
variable. Variables were converted into a percentile rank, and we correlated (0.53≤r ≤ 0.69), though correlations between mental
then applied the standard normal distribution quantile function. health reports before and during lockdown were generally low
(r≤0.34). Internal consistency of each mental health scale was
Role of the funding source above 0.74 for all measures except child-­rated depression symp-
The funders had no role in the study design, data collection, toms (Cronbach’s alpha=0.52).
data analysis, data interpretation or writing of the manuscript. Table 2 reports all main effects. From the mixed linear
The corresponding author had full access to all the data and had models with no control variables, we estimated a non-­significant
the final responsibility for the decision to submit for publication. decrease of 0.25 in SDQ emotional problems (B=−0.25, 95% CI
−0.54 to 0.05) and a decrease of 0.06 in RCADS anxiety scores
RESULTS (B=−0.06, 95% CI −0.34 to 0.23) during lockdown compared
First, the baseline data enabled us to estimate the size of any with before. Note that because all outcomes are standardised,
recruitment biases in those who responded during lockdown. the coefficients estimated from mixed linear models (B) can
There was a small bias for more affluent families to partici- be interpreted similarly to a standardised mean difference like
pate. Neighbourhood deprivation (measured using the Index of Cohen’s d (see note in table 2).22 The CI upper limits suggest
Bignardi G, et al. Arch Dis Child 2020;0:1–7. doi:10.1136/archdischild-2020-320372 3
Original research

Arch Dis Child: first published as 10.1136/archdischild-2020-320372 on 9 December 2020. Downloaded from http://adc.bmj.com/ on March 21, 2021 by guest. Protected by copyright.
Table 2  Main effects from mixed linear models
SDQ emotional problems RCADS anxiety subscale RCADS depression subscale
B 95% CI P value N B 95% CI P value N B 95% CI P value N
Model 1
 Lockdown −0.246 −0.542 0.050 0.103 298/149 −0.055 −0.335 0.225 0.699 371/162 0.736 0.458 1.014 <0.001 377/165
 Responder 0.276 −0.056 0.607 0.102 298/149 −0.796 −1.077 −0.515 <0.001 371/162 −1.331 −1.610 −1.052 <0.001 377/165
Model 2
 Lockdown −0.161 −0.537 0.215 0.398 298/149 0.059 −0.284 0.402 0.736 371/162 0.580 0.239 0.920 0.001 377/165
 Responder 0.274 −0.059 0.607 0.106 298/149 −0.796 −1.077 −0.514 <0.001 371/162 −1.302 −1.580 −1.023 <0.001 377/165
 Gender −0.025 −0.351 0.301 0.881 298/149 −0.033 −0.289 0.224 0.802 371/162 −0.328 −0.573 −0.082 0.009 377/165
 Lckdwn*Gndr −0.153 −0.552 0.245 0.449 298/149 −0.223 −0.580 0.134 0.220 371/162 0.249 −0.106 0.604 0.169 377/165
Model 3
 Lockdown −0.111 −0.505 0.284 0.580 296/148 0.108 −0.238 0.455 0.538 363/158 0.722 0.376 1.068 <0.001 369/161
 Responder 0.331 −0.006 0.668 0.055 296/148 −0.790 −1.069 −0.510 <0.001 363/158 −1.320 −1.602 −1.037 <0.001 369/161
 Age −0.013 −0.281 0.256 0.925 296/148 0.003 −0.212 0.218 0.979 363/158 0.027 −0.182 0.236 0.798 369/161
 Lockdown*age −0.215 −0.511 0.081 0.153 296/148 −0.222 −0.477 0.034 0.089 363/158 −0.031 −0.289 0.228 0.816 369/161
Model 4
 Lockdown −0.255 −0.551 0.041 0.090 298/149 −0.060 −0.340 0.220 0.674 371/162 0.733 0.455 1.011 <0.001 377/165
 Responder 0.290 −0.040 0.620 0.085 298/149 −0.792 −1.073 −0.510 <0.001 371/162 −1.329 −1.608 −1.050 <0.001 377/165
 SES −0.140 −0.301 0.020 0.086 298/149 −0.060 −0.187 0.067 0.353 371/162 −0.051 −0.174 0.071 0.409 377/165
 Lockdown*SES −0.046 −0.245 0.153 0.650 298/149 −0.054 −0.233 0.126 0.557 371/162 −0.063 −0.241 0.115 0.487 377/165
Paired t-­test
 Lockdown −0.195 −0.480 0.089 0.173 50 0.145 −0.136 0.426 0.305 51 0.713 0.432 0.994 <0.001 51
N for mixed linear models gives the (number of observations)/(number of individuals). Continuous variables of age and SES were z-­scored, and lockdown and responder are
binary variables. Lockdown is coded as before (0) or during (1) lockdown. Responder is coded teacher/child (0) or caregiver (1). Gender is coded as male (0) or female (1).
Coefficients for binary variables (eg, responder, lockdown and gender) can be interpreted mean group differences. For example, Bgender=−0.025 indicates that when accounting
for lockdown and responder, on average girls scored 0.025 less than boys. Because outcomes are standardised (M=0, SD=1), regression coefficients for binary variables can be
interpreted similarly to a standardised mean difference. Sample sizes are lower in model 3 due to missing age information for some children.
RCADS, Revised Child Anxiety and Depression Scale; SDQ, Strengths and Difficulties Questionnaire; SES, socioeconomic status.

that at most a small increase in these symptoms occurred during often seems worried’, from the SDQ) significantly changed,
lockdown. This is consistent with the proportion of children decreasing during lockdown (fewer worries during lockdown).
with SDQ emotional problem scores in the elevated range,
which changed very little, decreasing from 13% (19 children) to DISCUSSION
8% (12 children) from before to during lockdown.16 The short-­ National lockdowns with mass school closures are unprece-
form RCADS subscales do not have established cut-­offs for iden- dented, and the evidence base to guide future policymaking is
tifying elevated scores. emerging rapidly. Longitudinal data form a vital component of
In contrast, standardised RCADS depression scores were on that evidence base. This study is one of the first longitudinal
average 0.74 (95% CI 0.46 to 1.01) higher during lockdown studies and suggests that children’s depression ratings signifi-
than before (see figure 2). The CIs suggest a medium-­to-­large cantly increased during the lockdown, relative to 18 months
increase is likely. beforehand, with a medium-­ to-­
large effect. Note that this
Controlling for demographic factors separately (age, gender represents an average and not uniform change across children.
and SES) did not strongly alter our estimates for these effects. The effect of lockdown on mental health did not significantly
Interaction effects of these three factors were also estimated differ across demographic groups in moderation analyses exam-
to assess whether changes in mental health disproportionally ining children’s age, gender and family SES. However, larger
occurred in certain groups. No interaction effects were statisti- sample sizes are required to adequately statistically power
cally significant, although these estimates are highly uncertain. moderation analyses.23 24
A sensitivity analysis using only caregiver-­rated mental health
before and during lockdown was performed (see bottom table 2). Implications for policy makers and practitioners
One-­sample t-­tests were conducted on the standardised change The backdrop is that children’s mental health appears to be
scores, that is, the raw score during lockdown minus the before-­ worsening across successive cohorts, and even before lock-
lockdown score, divided by the change score SD. These analyses down, the resources for Child and Adolescent Mental Health
found similar effects as the mixed models (see table 2). Services were stretched thin.25 26 The current findings suggest
One potential limitation of using mean scale scores is that that lockdown measures will likely exacerbate this, specifically
changes during lockdown may be driven by specific items within with an increase in childhood depression symptoms, something
the scale. Therefore, changes in responses to each individual previously relatively uncommon in children of this age.27 The
question in the mental health scales were examined, using the education sector and families may bear the initial brunt of this.28
same t-­test approach outlined above (see figure 3). Four out of Indeed, one study has also reported an increase in parent’s
five of the depression questions showed significant increases psychological symptoms over lockdown.29 A key implication
during lockdown. Only one other question (‘Many worries, of the current findings is that the potential association between
4 Bignardi G, et al. Arch Dis Child 2020;0:1–7. doi:10.1136/archdischild-2020-320372
 Original research

Arch Dis Child: first published as 10.1136/archdischild-2020-320372 on 9 December 2020. Downloaded from http://adc.bmj.com/ on March 21, 2021 by guest. Protected by copyright.
Figure 2  Change in mental health ratings from before to during the lockdown measures. Panels A–C display changes in mental health ratings for
all three outcomes, respectively. Dark purple lines indicate changes in only parent-­reported mental health scores. Dashed lines indicate changes in
mental health scores from either teacher or child reports (before lockdown) to parent report (during lockdown). In each plot, we report the number of
responses before and during lockdown, by teachers, children or parents. Panel D plots the same data as panel C, however with age at assessment on
the horizontal axis and lines showing individual changes in depression symptoms. This shows a relatively sharp increase in depression symptoms from
before to during lockdown, compared with the relatively weak effect of age on depression symptoms (shown in the black, dashed line) estimated
from the mixed linear model. In panel D, child-­reported mental health measures are reduced by ~1.3 to aid visualisation, as the model estimated that
children reported higher depression compared with parents on by this amount on average. RCADS, Revised Child Anxiety and Depression Scale; SDQ,
Strengths and Difficulties Questionnaire.

lockdown and childhood mental health should be incorporated increase across 2 weeks of the COVID-19 pandemic, but depres-
in the decision-­making process of policy makers. When children sive symptoms specifically increased and anxiety symptoms
return to school, their well-­being, socialisation and enjoyment decreased.33 34 Studies that only measure mental health using
are paramount. Additional resources and training will likely be broad, brief mental health measures may fail to detect more
required to equip school teachers in how to support children specific effects.
with low mood and to increase their awareness of referral path-
ways for professional support. Study limitations
The small sample size of the current study is a limiting factor,
Future directions which reduces the statistical power and precision of estimates.
Future work should follow children over longer time periods Therefore, the lack of a statistically significant effect on SDQ
to assess long-­term effects. First, because there is potential for scores or RCADS anxiety scale, or moderation effects, should
‘sleeper effects’ (effects that emerge sometime after an initial be interpreted with caution. The current study does not have
adversity, often in a different phase of development), and second, the statistical power to detect small but clinically meaningful
because we need to test whether children’s mood rebounds when changes. Second, because this is a convenience sample collected
school resumes.30 Larger cohorts with greater statistical power within our main cohort, the proportion of responders is rela-
are needed to address whether the epidemic has had dispropor- tively small compared with the size of the overall cohort. This is
tionate effects on particular children and households. Of partic- perhaps to be expected given the timing of our survey and the
ular concern are children with existing mental health and other context of the pandemic. However, the baseline characteristics
needs. Initial reports have highlighted challenges during school were only very weakly associated with which families responded
closures facing children with autism and attention deficit hyper- to our invitation to take part. As we only sampled a small region
activity disorder.31 32 of the UK, caution should be applied in generalising the results
Finally, our analysis of individual questionnaire items suggests to different populations. Third, the mixture of reporters is a
that particular symptoms may be differentially affected by the limitation for the study, as well as the lack of child-­reported
lockdown. Larger epidemiological studies could further explore measures during lockdown. Children and adults report mental
this potential differential association between lockdown on health symptoms differently. This is why reporter is directly
mental health. One study in 80 Dutch students reported prelim- incorporated within the model, and the effects are subsequently
inary evidence that global mental health problems did not replicated in a subsample with just longitudinal caregiver report.
Bignardi G, et al. Arch Dis Child 2020;0:1–7. doi:10.1136/archdischild-2020-320372 5
 Original research

Arch Dis Child: first published as 10.1136/archdischild-2020-320372 on 9 December 2020. Downloaded from http://adc.bmj.com/ on March 21, 2021 by guest. Protected by copyright.
Changes in Item Scores

SDQ Emotion Prob.

Many fears, easily scared -0.17 [-0.46, 0.11]

Nervous or clingy in new situations, easily loses confidence -0.09 [-0.38, 0.19]

Often unhappy, down-hearted or tearful -0.07 [-0.35, 0.22]

Many worries, often seems worried -0.39 [-0.67, -0.10]

Often complains of headaches, stomach-aches or sickness 0.02 [-0.26, 0.30]

RCADS Anxiety

My child thinks about death 0.09 [-0.19, 0.37]

My child worries that something awful will happen to someone in the family 0.28 [-0.01, 0.56]

My child worries about what is going to happen 0.10 [-0.18, 0.39]

My child worries that something bad will happen to him/her 0.07 [-0.22, 0.35]

My child worries that bad things will happen to him/her 0.05 [-0.23, 0.33]

RCADS Depression

My child has problems with his/her appetite 0.04 [-0.24, 0.33]

My child feels like he/she doesn't want to move 0.73 [ 0.44, 1.01]

My child is tired a lot 0.48 [ 0.19, 0.76]

My child feels sad or empty 0.48 [ 0.19, 0.76]

Nothing is much fun for my child anymore 0.63 [ 0.35, 0.92]

-1.0 -0.5 0.0 0.5 1.0

Mean Standardised Change Score

Figure 3  Mean standardised change score and CIs for each mental health question, comparing during to before lockdown, using solely caregiver
reported mental health at both time points. Positive numbers indicate worsening of symptom during lockdown. Results support the interpretation that
RCADS depression items have increased during lockdown, with more minor changes in other symptoms. RCADS, Revised Child Anxiety and Depression
Scale; SDQ, Strengths and Difficulties Questionnaire.

CONCLUSIONS purpose, provided the original work is properly cited, a link to the licence is given,
We report longitudinal evidence for the negative association and indication of whether changes were made. See: https://​creativecommons.​org/​
licenses/​by/​4.​0/.
between UK lockdown measures and child mental health. Specif-
ically, we observed a statistically significant increase in ratings ORCID iDs
of depression, with a medium-­to-­large effect size. Our findings Giacomo Bignardi http://​orcid.​org/​0000-​0002-​1153-​0838
emphasise the need to incorporate the potential impact of lock- Edwin S Dalmaijer http://​orcid.​org/​0000-​0003-​3241-​0760
Alexander L Anwyl-­Irvine http://​orcid.​org/​0000-​0002-​3792-​7745
down on child mental health in planning the ongoing response Tess A Smith http://​orcid.​org/​0000-​0003-​3671-​8551
to the global pandemic and the recovery from it. Roma Siugzdaite http://​orcid.​org/​0000-​0002-​4063-​1128
Stepheni Uh http://​orcid.​org/​0000-​0002-​6541-​3039
Twitter Giacomo Bignardi @BignardiG, Edwin S Dalmaijer @esdalmaijer, Alexander Duncan E Astle http://​orcid.​org/​0000-​0002-​7042-​5392
L Anwyl-­Irvine @AlexanderIrvine, Roma Siugzdaite @RSiugzdaite and Duncan E Astle
@DuncanAstle
Contributors  GB conceptualised and designed the study, collected the data, REFERENCES
drafted the initial manuscript, carried out the statistical analysis, and reviewed and 1 Brooks SK, Webster RK, Smith LE, et al. The psychological impact of quarantine
revised the manuscript. DEA conceptualised and designed the study, drafted the and how to reduce it: rapid review of the evidence. Lancet 2020;395:912–20.
initial manuscript, supervised the study, obtained funding, and reviewed and revised 2 Lee J. Mental health effects of school closures during COVID-19. Lancet Child Adolesc
the manuscript. ESD conceptualised and designed the study, collected the data, Health 2020;4:421.
supervised the study, and reviewed and revised the manuscript. RS, ALA-­I, TAS and 3 Orben A, Tomova L, Blakemore S-­J. The effects of social deprivation on
SU conceptualised and designed the study, collected the data, and reviewed and adolescent development and mental health. Lancet Child Adolesc Health
revised the manuscript. 2020;4:634–40.
4 Kwong ASF, Pearson RM, Adams MJ, et al. Mental health during the COVID-19
Funding  This study was supported by grant TWCF0159 from the Templeton World
pandemic in two longitudinal UK population cohorts. medRxiv.
Charity Foundation to DA and by the UK Medical Research Council (MC-­A0606-­
5 Pierce M, Hope H, Ford T, et al. Mental health before and during the COVID-19
5PQ41). SU was supported by the Gates Cambridge Trust.
pandemic: a longitudinal probability sample survey of the UK population. Lancet
Competing interests  None declared. Psychiatry 2020;7:883-892.
Patient consent for publication  Not required. 6 Li HY, Cao H, Leung DYP, et al. The psychological impacts of a COVID-19 outbreak
on college students in China: a longitudinal study. Int J Environ Res Public Health
Ethics approval  The study protocol was approved by the University of Cambridge 2020;17:3933.
Psychology Research Ethics Committee (PRE.2017.102). 7 Daly M, Sutin A, Robinson E. Longitudinal changes in mental health and the
Provenance and peer review  Not commissioned; externally peer reviewed. COVID-19 pandemic: evidence from the UK household longitudinal study 2020.
8 Ettman CK, Abdalla SM, Cohen GH, et al. Prevalence of depression symptoms
Data availability statement  Data are available on reasonable request. Ninety-­six in US adults before and during the COVID-19 pandemic. JAMA Netw Open
per cent of the participants gave consent for their data to be shared on the condition 2020;3:e2019686.
that researchers applying for access have the appropriate ethical permission. 9 van der Velden PG, Contino C, Das M, et al. Anxiety and depression symptoms, and
Requests for data access should be sent to the corresponding author. Analysis code lack of emotional support among the general population before and during the
and R output are available from https://​osf.​io/​ajy57. COVID-19 pandemic. A prospective national study on prevalence and risk factors. J
Open access  This is an open access article distributed in accordance with the Affect Disord 2020;277:540–8.
Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits 10 Widnall E, Winstone L, Mars B, et al. Young People’s Mental Health during the
others to copy, redistribute, remix, transform and build upon this work for any COVID-19 Pandemic 2020.

6 Bignardi G, et al. Arch Dis Child 2020;0:1–7. doi:10.1136/archdischild-2020-320372

 Original research

Arch Dis Child: first published as 10.1136/archdischild-2020-320372 on 9 December 2020. Downloaded from http://adc.bmj.com/ on March 21, 2021 by guest. Protected by copyright.
11 Loades ME, Chatburn E, Higson-­Sweeney N, et al. Rapid systematic review: the impact 23 Rochon J. Sample size calculations for Two-­Group Repeated-­Measures experiments.
of social isolation and loneliness on the mental health of children and adolescents in Biometrics 1991;47:1383.
the context of COVID-19. J Am Acad Child Adolesc Psychiatry 2020;59:1218–39. 24 Leon AC, Heo M. Sample sizes required to detect interactions between two binary
12 Taquet M, Quoidbach J, Fried EI, et al. Mood homeostasis before and during the fixed-­effects in a mixed-­effects linear regression model. Comput Stat Data Anal
coronavirus disease 2019 (COVID-19) Lockdown among students in the Netherlands. 2009;53:603–8.
JAMA Psychiatry 2020. doi:10.1001/jamapsychiatry.2020.2389. [Epub ahead of print: 25 Sellers R, Warne N, Pickles A, et al. Cross-­Cohort change in adolescent outcomes for
29 Jul 2020]. children with mental health problems. J Child Psychol Psychiatry 2019;60:813–21.
13 Witt A, Ordóñez A, Martin A, et al. Child and adolescent mental health service 26 Crenna-­Jennings W, Hutchinson J. Access to Child and Adolescent Mental Health
provision and research during the Covid-19 pandemic: challenges, opportunities, and Services in 2019 2020.
a call for submissions. Child Adolesc Psychiatry Ment Health 2020;14:19. 27 Ghandour RM, Sherman LJ, Vladutiu CJ, et al. Prevalence and treatment of depression,
14 Ford T. Prof Tamsin Ford: Challenges for children’s services as Lockdown eases. anxiety, and conduct problems in US children. J Pediatr 2019;206:256–67.
Reachwell, 2020. Available: https://​reachwell.​org/​2020/​07/​16/​prof-​tamsin-f​ ord-​ 28 Newlove-­Delgado T, Moore D, Ukoumunne OC, et al. Mental health related contact
challenges-f​ or-​childrens-​services-​as-l​ ockdown-​eases/ [Accessed 21 Jul 2020]. with education professionals in the British child and adolescent mental health survey
15 Bignardi G, Dalmaijer E, Anwyl-­Irvine A, et al. Collecting Big Data with Small Screens: 2004. J MH Training, Ed and Practice 2015;10:159–69.
Group-­Tests of Children’s Cognition with Touchscreen Tablets Are Reliable and Valid. 29 Achterberg M, Dobbelaar S, Boer OD, et al. Home lockdown: Bloom or boom?
Behaviour Research Methods 2020. In Press. perceived stress as mediator for longitudinal effects of the COVID-19 lockdown on
16 Goodman R. The strengths and difficulties questionnaire: a research note. J Child wellbeing of parents and children 2020.
Psychol Psychiatry 1997;38:581–6. 30 Wade M, Prime H, Browne DT. Why we need longitudinal mental health research
17 Muris P, Meesters C, Schouten E. Questionnaire of anxiety and depression children. with children and youth during (and after) the COVID-19 pandemic. Psychiatry Res
Clin Psychol Psychother 2002;9:430–42. 2020;290:113143.
18 Smith T, Noble M, Noble S, et al. The English indices of deprivation 2015. London Dep 31 Bobo E, Lin L, Acquaviva E, et al. Comment les enfants et adolescents avec le trouble
Communities Local Gov 2015. déficit d’attention/hyperactivité (TDAH) vivent-­ils le confinement durant la pandémie
19 Taylor C. The reliability of free school meal eligibility as a measure of socio-­economic COVID-19 ? Encephale 2020;46:S85–92.
disadvantage: evidence from the millennium cohort study in Wales. British Journal of 32 Colizzi M, Sironi E, Antonini F, et al. Psychosocial and behavioral impact of COVID-19
Educational Studies 2018;66:29–51. in autism spectrum disorder: an online parent survey. Brain Sci 2020;10:341.
20 Pinheiro J, Bates D, DebRoy S, et al. Package ’nlme. Linear nonlinear Mix Eff Model 33 Eiko F. Preprint: COVID-19 pandemic and student mental health, 2020. Available:
version 2017;3. https://​eiko-​fried.c​ om/​preprint-​covid-​19-p​ andemic-​and-​student-​mental-​health/
21 Team RC. R: a language and environment for statistical computing 2013. [Accessed 22 Jul 2020].
22 Lakens D. Calculating and reporting effect sizes to facilitate cumulative science: a 34 Fried E, Papanikolaou F, Epskamp S. Mental Health and Social Contact During the
practical primer for t-­tests and ANOVAs. Front Psychol 2013;4:863. COVID-19. Pandemic: An Ecological Momentary Assessment Study, 2020.

Bignardi G, et al. Arch Dis Child 2020;0:1–7. doi:10.1136/archdischild-2020-320372 7

Steep rise in depressive symptoms among 7-12 year olds during UK-wide lockdown

Effect size medium to large, the findings indicate; implications for future school closures

The prevalence of depressive symptoms rose substantially among young children during the UK-
wide lockdown in response to the first wave of the coronavirus pandemic, finds research focusing
on one region of England and published online in Archives of Disease in Childhood.

The effect size was medium to large, indicate the findings, which have implications for future
partial or complete school closures, suggest the researchers.

Social distancing and school closures during the lengthy lockdown across the UK earlier this year
are thought to have badly affected children’s mental health.

But there has been little hard evidence to substantiate this, largely because it is hard to find good
baseline data, for the same children, collected before the first lockdown on March 23.

The researchers were able to get round this by drawing on data on children living in the East of
England who were part of the Resilience in Education and Development (RED) study.

During lockdown,168 of their parents--equivalent to 29% of the overall sample--completed online

validated mental health assessments to rate their children’s emotional wellbeing, anxiety and
depression.

These ratings were then compared with baseline data collected around 18 months earlier, which
included a mixture of parent-, teacher-, and child-rated mental health measures.

Compared with the initial baseline assessment, there were no significant changes in levels of
anxiety or emotional wellbeing during lockdown.

But a significant increase of 0.74 in depressive symptoms was observed, the effect size of which
was medium to large. Put simply, this means that, on average, there was around a 70% chance
that depressive symptoms worsened during lockdown in any child.

These findings held true even after accounting for potentially influential factors, including age,
gender, and socioeconomic status, although larger numbers of children are needed to confirm
this, say the researchers.

To find out whether changes during lockdown might have been driven by specific items within the
depression scale, the researchers looked at changes in responses to each individual question in
the mental health scales.

This revealed significant increases for 4 out of 5 of the depression questions during lockdown in
respect of lethargy, struggling to enjoy activities, and feelings of sadness or emptiness.

This is an observational study; the numbers are relatively small; and from one area of England
only, so may not be more widely applicable, caution the researchers.

But they point out: “The backdrop is that children’s mental health appears to be worsening across
successive cohorts, and even before lockdown, the resources for Child and Adolescent Mental
Health Services were stretched thin.

“The current findings suggest that lockdown measures will likely exacerbate this, specifically with
an increase in childhood depression symptoms, something previously relatively uncommon in
children of this age.”
They add: “Our findings emphasise the need to incorporate the potential impact of lockdown on
child mental health in planning the ongoing response to the global pandemic and the recovery
from it.”

Dr Karen Street is a paediatric consultant and mental health lead at the Royal College of
Paediatrics and Child Health (RCPCH), which co-owns the journal with BMJ.

She commented: “This study reports what many paediatricians have observed--while children
rarely become ill with COVID-19, they have been significantly affected by the measures taken to
reduce transmission of the virus.

She added: “The RCPCH welcomes the government’s ongoing commitment to keep schools
open. It would also be good to see extra-curricular activities and opportunities for children and
young people return as soon as possible.

“While we hope that for many children a return to normality will see a 'rebound' in their emotional
wellbeing, we also know that the socioeconomic impact of lockdown for many families will be
ongoing for many years, and that this will have secondary negative effects on the mental health
of children.

“Persisting mental health problems in childhood and adolescence are associated with poor
outcomes for educational attainment, employment, and long term physical and mental health so
it is vital there is sufficient investment in health, education and the voluntary sector to support
children's mental health as we recover from the pandemic.”