Examining changes in parent‐reported child and adolescent mental health throughout the UK's first COVID‐19 national lockdown

Introduction

The COVID-19 pandemic has caused substantial disruption to children and adolescents through potential threat of illness to themselves and others, school closures, exam disruption, restrictions to social interactions, and increased family pressures (Office for National Statistics, 2020). Pandemic-associated restrictions have meant that children and adolescents have often experienced prolonged periods of social isolation in addition to increased feelings of uncertainty and high levels of stress amongst the adults around them (Dalton, Rapa, & Stein, 2020).

The most robust evidence to date for the prevalence of mental health difficulties during the pandemic comes from the NHS Digital survey of children and adolescents' mental health in England (NHS Digital, 2020) that reported that the proportion of children and adolescents with a probable mental health disorder was 1 in 6 in July 2020 (after the end of the national lockdown but while many restrictions were still in place) compared to 1 in 9 in 2017. This deterioration may have been a continuation of the pattern of increasing mental health problems seen in previous surveys; although the finding that over 40% of adolescents reported that the pandemic had made their mental health worse highlights the potential contribution of the pandemic. However, notably, 27.2% of adolescents reported that their mental health had improved during lockdown (Newlove-Delgado et al., 2021), and self-report data from over 11,000 pupils aged between 6 and 18 showed that average well-being remained relatively stable between May and July 2020 (ImpactEd, 2021). Findings across other studies also indicate a mixed picture; while there is some evidence of UK adolescents reporting higher levels of worry and a decline in their mental wellbeing during lockdown (Children's Parliament, 2020), others suggested that some adolescents 'thrived' (Selwyn, 2020) and experienced improvements in their mental health during the first national lockdown (Widnall, Winstone, Mars, Haworth, & Kidger, 2020). For example, around 2,000 young people aged 8–17 years surveyed in June 2020 reported being less frequently stressed over the past month than a previous panel of around 1,850 young people from similar backgrounds, surveyed in March 2020 as the pandemic unfolded (Children's Commissioner, 2020). Notably, most studies to date have either involved retrospective reports or have compared children's adjustment between a pre-lockdown assessment and a single follow-up assessment during lockdown. There is a lack of empirical and longitudinal research directly examining how mental health symptoms have changed throughout the pandemic (Racine et al., 2020) and what might account for differences in children and adolescent's responses.

There has been particular concern about the mental health impact of the pandemic and associated lockdown restrictions on children who were already vulnerable prior to the pandemic, for example, children and adolescents in low income households (Gutman, Joshi, Parsonage, & Schoon, 2015), with pre-existing mental health problems (Jefsen, Rohde, Nørremark, & Østergaard, 2020), with SEN (Asbury, Fox, Deniz, Code, & Toseeb, 2021; ImpactEd, 2020) and/or ND (Nonweiler, Rattray, Baulcomb, Happé, & Absoud, 2020), with pre-existing chronic health conditions (Butler et al., 2018) and where parents experienced high levels of distress (Lawrence, Murayama, & Creswell, 2019). Particular contextual factors may also have created strain on families during the lockdown restrictions. Most notably, being in a single adult or a single child household (Rosen et al., 2020).

In addition to understanding the contextual factors that may increase the risk of a decline in children and adolescents' mental health throughout the pandemic, it is also critical to explore potentially modifiable factors that have previously been associated with resilience (Fritz, de Graaff, Caisley, van Harmelen, & Wilkinson, 2018). These include individual factors (e.g., cognitive factors and emotion regulation), community factors (e.g., social support), and family factors (Fritz et al., 2018). For example, having a good family climate is associated with a lower prevalence of mental health problems in adolescents (Klasen et al., 2015). Immediate family support has also been shown to weaken the relationship between childhood adversity and the development of emotional symptoms (e.g., depression; Shahar & Henrich, 2016). Given the particular circumstances of the pandemic, when lockdown typically restricted children and young people to being in their family home, we focus here on family factors.

Present study

In the present study, we aimed to explore the trajectories of change in children and adolescents' mental health (as reported by their parents/carers) during the UK's national lockdown in response to the COVID-19 pandemic. From 23 March until the end June/beginning of July, schools, workplaces, and all non-essential shops were forced to close, and the public were encouraged to stay at home.

Specifically, we explored the following questions:

How did children and young people's mental health change through the first 4 months of the pandemic?
Is change in children and young people's mental health over time predicted by family contextual and resilience factors?
Can we identify children and young people with different sub-types of trajectories of change in mental health symptoms through this stage of the pandemic?
Do family contextual and resilience factors predict the probability of children and young people having these different trajectories?

Method

Design

The 'COVID-19: Supporting Parents, Adolescents and Children during Epidemics' (Co-SPACE) study is an online longitudinal survey composed of a convenience sample of UK parents and carers of children and adolescents aged between 4 and 16 years. The research protocols for the overall Co-SPACE study and supporting material for this specific project are available via the Open Science Framework (https://osf.io/8zx2y/; https://osf.io/c2v4d/).

Eligibility

Parents and carers of children and adolescents aged between 4 and 16 years who lived in the UK were eligible to take part.

Procedure

Participants were invited to report on their child in an online Qualtrics (www.qualtrics.com/uk) survey from 30 March 2020. Parents of multi-child families were asked to identify one 'index' child who they would report on each time. Following completion of the baseline survey, participants were invited back monthly for a follow-up survey. Informed consent was obtained from the parents/carers. Ethical approval for the study was provided by the University of Oxford Medical Sciences Division Ethics Committee (reference R69060).

Participants

Participants were eligible parents and carers (aged over 18). The current paper focuses on a sub-sample of 3,046 out of a total of 5,191^1. participants who completed their baseline survey between 30 March and 29 April 2020 and then at least one follow-up survey between the following dates: 30 April and 31 May (n = 2,584); June 1 and June 30 (n = 1,825); and July 1 and July 31 (n = 1,671).^2. Only those who completed the Strengths and Difficulties Questionnaire (SDQ; Goodman, 1997, 2001) and provided full data for the predictor variables (collected at baseline) were included in the analysis (April n = 2,988; May n = 2,533; June n = 1,792 and July n = 1,645). Demographic and other baseline information for participants and their children can be found in Table 1. Notably, respondents were predominantly female parents, with higher levels of education, of White ethnicities, and from relatively affluent backgrounds, and while we recruited parents of children aged 4–16 years, the mean age of children was within the primary school age range (around 9 years).

Table 1. Demographic and baseline information for participants included in the analyses

	April	May	June	July
	2,988	2,533	1,792	1,645
Parent/carer gender
Male	170	139	105	110
Female	2,818	2,394	1,687	1,535
Parent/carer ethnicity
White British	2,880	2,448	1,728	1,589
Other	108	85	64	56
Parent/carer education
School/vocational qualification	435	373	242	203
Undergraduate degree	1,199	1,007	722	652
Post-graduate degree	1,335	1,137	822	783
No qualifications	19	16	6	7
Child mean age (SD)	9.18 (3.42)	9.15 (3.44)	9.25 (3.40)	9.12 (3.37)
Child gender
Male	1,557	1,319	932	861
Female	1,431	1,214	861	784
Child ethnicity
White British	2,776	2,362	1,662	1,533
Other	212	171	130	112
Household income (p.a.)
<£16,000	148	130	90	71
£16–2,900	318	283	182	163
£30–59,000	888	747	509	478
£60–89,000	682	574	419	390
£90–119,000	375	312	230	209
>£120,000	371	308	238	210
Prefer not to say	206	179	124	124
Family composition
Single adult household	385	340	229	231
Multiple adult household	2,603	2,193	1,563	1,451
Child Chronic Health
No chronic health	2,648	2,248	1,581	1,449
Chronic health condition	340	285	211	196
Presence of siblings
No siblings	737	626	476	435
Siblings	2,251	1,907	1,316	1,210
Mean Depression Anxiety and Stress Scale (DASS; 9 items) (SD)	5.20 (4.47)	–	–	–
Mean family warmth (SD)	2.71 (0.54)	–	–	–
Mean family conflict (SD)	0.86 (0.65)	–	–	–

SEND/ND, Special educational needs/neurodevelopmental disorders.

Measures

Details and the coding of each measure can be found in the Supporting Information (see Appendix S1).

Child and adolescent mental health

The parent-report version of the SDQ was administered. In the current paper, we focus on the three SDQ sub-scales that measure mental health symptoms: emotional symptoms, conduct problems, and hyperactivity/inattention.

Parent/carer demographic information (baseline survey only)

Parents/carers were asked to report on their own and their child's age, gender, and ethnicity (see Table 1). We also obtained measures of household income, single adult and single child status, child chronic health, and child SEN and/or ND.

Symptoms of psychological distress in parents and carers

A self-report measure comprising a subset of nine items (McElroy et al., 2020) from the Depression Anxiety Stress Scales (DASS-21; Lovibond & Lovibond, 1995) was administered.

Family support

Family warmth and family conflict were assessed.

Analysis

Data were organized using R (R Core Team, 2018; v.3.6.2), and analyses were conducted in MPlus (v.8.4; Muthén & Muthén, 2000) and R. Prior to examining the research questions, associations between predictors and baseline SDQ scores were examined to test for potential differences between children at baseline. The first two questions were addressed by specifying latent growth curve models to investigate the change in SDQ scores over time. Linear and non-linear growth was tested, and time was coded as: 0 (April), 1 (May), 2 (June), and 3 (July). The three SDQ sub-scales were modelled separately as the dependent variables. In each model, the intercept (representing the baseline data) and the slope (representing the change over time) of the dependent variables were modelled. Missing data were addressed using full information maximum likelihood estimation (FIML), which uses all information available from all respondents, thus being less prone to biases than a complete case analysis with listwise deletion where the loss of information is larger and would lead to greater biases in estimates. To determine a good statistical fit, we accepted models that had Comparative Fit Measure values >.90 (Kline, 2016) and Root Mean Square Error of Approximation <.08 (Browne & Cudeck, 1993). These models analysed whether the change in the SDQ sub-scales was predicted by baseline measures of parent/carer psychological distress, family warmth and conflict, child age, gender, ethnicity, chronic health, and SEN/ND as well as total household income (per annum), presence of siblings, and single adult household status.

The third question was assessed using latent growth mixture modelling to identify child-specific trajectories on outcome measures (the three SDQ sub-scales). We ran models with an increasing number of trajectories until non-convergence was reached. Due to negative residual variances and correlations greater than 1 between the latent variables, the slope was constrained to 0 in all models. Model fit was evaluated using Bayesian Information Criterion, the Akaike Information Criteria, entropy index, and the Lo, Mendell, and Rubin (LMR; 2001) statistic (see Table S1). To address the fourth question, class membership was regressed on to the covariates using a multinomial logistic regression (mlogit package in R; Croissant, 2020) for each SDQ sub-scale separately. As all entropy values were <0.80, class probability weights were included in the regression models to account for the lower neatness of classification. The obtained trajectories were compared to the reference group (defined as the largest group) that were expected to be the low symptom groups over time. However, in the multinomial logistic regressions, <£16 k was used as the reference group. Results using the most frequent category (£30–59 k) as the reference group are reported in Appendix S2.

Results

Question 1. How did children and adolescents' mental health change through the first 4 months of the pandemic?

Question 2. Is change in children and adolescents' mental health over time predicted by family contextual and resilience factors?

The latent growth curve analyses explored the changes in children and adolescents' mental health over the first four months of the pandemic (see Figure 1). Estimating quadratic growth, compared to linear growth, significantly improved the fit for hyperactivity/inattention and emotional symptoms but not for conduct problems. Therefore, we included linear and quadratic growth for hyperactivity/inattention and emotional symptoms but included only a linear growth for conduct problems (see Table 2 for model fit indices).

Change in estimated means for the three SDQ sub-scales between April and July

Table 2. Predictors of baseline and change over time in children and adolescents' for Conduct Problems, Hyperactivity/inattention, and Emotional Symptoms

...

Comments

	Conduct problems		Hyperactivity/inattention		Emotional symptoms
	b (SE)	95% CI	b (SE)	95% CI	b (SE)	95% CI
Intercept
DASS	.14 (0.02)***	0.1, 0.17	.16 (0.02)***	0.12, 0.19	.26 (0.02)***	0.22, 0.30
Child ethnicity	.002 (0.02)	−0.03, 0.03	−.01 (0.02)	−0.04, 0.03	−.03 (0.02)	−0.07, 0.003
Child gender	−.06 (0.02)***	−0.09, −0.03	−.17 (0.02)***	−0.2, −0.13	.13 (0.02)***	0.09, 0.16
SEN/ND	.3 (0.02)***	0.26, 0.34	.42 (0.02)***	0.38, 0.46	.3 (0.02)***	0.26, 0.35
Single adult household	−.003 (0.02)	−0.04, 0.03	.01 (0.02)	−0.03, 0.05	.01 (0.02)	−0.04, 0.05
Presence of siblings	.1 (0.02)***	0.07, 0.13	−.02 (0.02)	−0.05, 0.01	.05 (0.02)*	0.02, 0.09
Family warmth	−.2 (0.02)***	−0.24, −0.17	−.1 (0.02)***	−0.14, −0.07	−.03 (0.02)	−0.07, 0.01
Family conflict	.45 (0.02)***	0.42, 0.49	.22 (0.02)***

Immunization