INTRODUCTION

In March 2020, the first UK COVID-19 national lockdown commenced and schools were closed except to children whose parents were essential workers (also known as ‘critical’ or key workers), or those who were considered ‘vulnerable’, such as those with mental health needs or living in care (Cabinet Office, 2020; Department for Education, 2020). All other pupils were provided with varying degrees of educational support while at home. From 1 June, pupils in some year groups were also invited back to school, such as those approaching key national examinations. This implies that those offered places in school could be more at risk of mental health difficulties than those who remained at home for one or more reasons; either because they met the criteria for ‘vulnerability’, because their parents were performing essential roles outside the family home or because they needed to prepare for national examinations in the forthcoming academic year.

There was a concern that partial school closures may reinforce or exacerbate pre-existing inequalities by disproportionally affecting young people who are already at increased risk of poor mental health (Armitage & Nellums, 2020; Van Lancker & Parolin, 2020; Viner et al., 2020). School closures might have some unique benefits to those pupils who frequently experience behavioural difficulties or peer victimisation at school, but it might be detrimental for many others with mental health problems due to loss of support (Courtney et al., 2020; Golberstein et al., 2020; Lee, 2020; YoungMinds, 2020). Cumulative risk factors within homes, such as domestic abuse, limited physical space, economic challenges and single parenthood, may contribute to increased adversity, potentially leading to both immediate and long-term consequences for mental health (Clemens et al., 2020; Cluver et al., 2020; Courtney et al., 2020; Crawley et al., 2020; Gilbert et al., 2009; Usher et al., 2020).

To inform practice and policy for future periods of school lockdown and long-term consequences, this study investigates which young people were most at risk of negative impacts during school closures on their mental health and wellbeing (Golberstein et al., 2020; Holmes et al., 2020; Lee, 2020; O'Connor et al., 2020) and how increased risk relates to whether or not pupils were getting access to in-school educational provision. We used a large cross-sectional survey of school pupils during June–July 2020 and compared established factors and lockdown-specific demographic and situational factors on three different outcomes: clinical depression, clinical anxiety and pupils' self-reported deterioration of their mental wellbeing.

METHODS

RESULTS

Risk factors for mental health difficulties

Depression and anxiety, and self-reported deterioration to wellbeing, were each modelled against the demographic and situational indicators. Adjusted ORs and 95% confidence intervals were calculated (see Table 2), where all indicators were accounted for together in each model.

TABLE 2. Logistic regression of mental health and self-reported worse wellbeing with adjusted odds ratios for all risk and contextual factors

Outcome	Predictor	Adjusted odds ratio	95% Confidence interval	p-Value
Depression (RCADS depression subscale t-score ≥ 70) (n = 8798)	In-school provision	1.16	1.00, 1.35	.0640
	Upcoming exams (year 10 or 12)	1.46	1.26, 1.69	<.0001
	Not living with both parents	1.20	1.03, 1.40	.0294
	Free school meals	1.03	0.81, 1.32	.7820
	Ever experienced food poverty	3.36	2.79, 4.04	<.0001
	Past access to mental health support	3.91	3.42, 4.47	<.0001
	Female gender	3.62	3.04, 4.32	<.0001
	Essential worker parents	1.34	1.18, 1.54	<.0001
Anxiety (RCADS anxiety subscale t-score ≥ 70) (n = 8798)	In-school provision	1.20	1.01, 1.42	.0495
	Upcoming exams (year 10 or 12)	1.49	1.24, 1.77	<.0001
	Not living with both parents	0.98	0.82, 1.17	.8151
	Free school meals	1.04	0.79, 1.37	.8151
	Ever experienced food poverty	3.03	2.48, 3.70	<.0001
	Past access to mental health support	3.83	3.29, 4.47	<.0001
	Female gender	2.55	2.10, 3.10	<.0001
	Essential worker parents	1.25	1.07, 1.46	.0063
Wellbeing got worse (n = 9309)	In-school provision	1.10	0.98, 1.22	.0977
	Upcoming exams (year 10 or 12)	1.54	1.37, 1.73	<.0001
	Not living with both parents	1.31	1.17, 1.47	<.0001
	Free school meals	0.84	0.70, 1.00	.0515
	Ever experienced food poverty	1.82	1.57, 2.11	<.0001
	Past access to mental health support	1.77	1.60, 1.96	<.0001
	Female gender	1.88	1.70, 2.08	<.0001
	Essential worker parents	1.25	1.14, 1.37	<.0001

• Abbreviation: RCADS, Revised Children's Anxiety and Depression Scales.

Depression and anxiety

Risk of depression was found to be higher for females (adjusted OR = 3.62, p < .0001), pupils who had previously accessed mental health support (adjusted OR = 3.91, p < .0001) and those who had ever experienced food poverty (adjusted OR = 3.36, p < .0001). Increased risk of depression was also found for pupils whose parents were likely essential workers (adjusted OR = 1.34, p < .0001), pupils who were approaching national examinations (adjusted OR = 1.46, p < .0001) and a small increased risk for pupils not living with both parents (adjusted OR = 1.20, p = .0294) that was not significant in the sensitivity analysis that included non-response weights (OR = 1.18, p = .0565).

There was no association between self-reported eligibility for free school meals and depression. The group allocated to in-school educational provision did not show clear evidence of elevated risk of depression in this adjusted model (adjusted OR = 1.16, p = .0640). For the model predicting anxiety, the risk factors demonstrated a similar pattern to that seen for depression (Figure 1). Pupils not living with both parents were not at elevated risk of anxiety. Pupils accessing in-school provision had a small but significant increased risk of anxiety (adjusted OR = 1.20, p = .0495), but this was not significant in the weighted analysis (OR = 1.20, p = .0514).

Self-reported deterioration to mental wellbeing

Factors associated with increased risk followed a similar pattern for self-reported deterioration of mental wellbeing during lockdown to the pattern for depression and anxiety (Figure 1). For factors especially relevant to the school lockdown, namely having key worker parents or preparing for key examinations in 2021, ORs for reporting wellbeing to be worse during lockdown were similar in magnitude to the clinical outcomes. For the deprivation and vulnerability predictors, ORs were smaller than those for the validated clinical measures.

DISCUSSION

This study provides insight into the mental health and wellbeing of different groups during the partial school closures of the first UK COVID-19 school lockdown in 2020 and how this relates to which pupils took up in-school places. The results show that those who were accessing on-site school were more likely to have depression or anxiety and were more likely to report a deterioration in their wellbeing. However, the poorer outcomes for the group in-school were accounted for by pre-existing vulnerability (e.g. experiencing food poverty and previous access to mental health support), in line with the practice during the first lockdown of trying to prioritise in-school places to the pupils who most needed them. Increased risk for poor mental health was not only associated with the established indicators of vulnerability and deprivation but also associated with contextual factors especially relevant during the lockdown measures – pupils approaching national examinations and parents performing essential roles.

Pupils who were accessing on-site school were hypothesised to represent a higher risk group due to some of the criteria used to allocate school places (especially vulnerability) and schools' knowledge of which pupils met those criteria. As predicted, the group accessing on-site educational provision had both higher scores on pre-existing risk and vulnerability factors and poorer mental health outcomes than their peers at home. However, with adjustment for each of the individual indicators, effects of accessing in-school provision on mental health outcomes were not robust and any differences between the groups should be interpreted with caution. Besides the national criteria for vulnerable children and children of essential workers, many schools used other information, at their discretion, to allocate places to the most vulnerable (Husband, D., personal communication, 7 September 2020), which might have included implicit or explicit knowledge of pupils' current mental health. Further studies could aim to clarify how different adaptations of in-school provision might have impacted mental health either positively or negatively.

Factors associated with the highest odds of a deterioration of mental wellbeing, as well as depression and anxiety, were established risk factors, confirming predictions that more vulnerable young people would be at greater risk of mental health problems during the partial school closures (Armitage & Nellums, 2020; Van Lancker & Parolin, 2020). We found that pupils who had previously accessed mental health support had almost four times the odds of reaching clinical thresholds for depression and anxiety and were also more likely to report a further deterioration of their wellbeing. This is consistent with the finding by YoungMinds that 80% of the 2036 adolescents surveyed with pre-existing mental health needs felt that their mental health had become worse during the pandemic (YoungMinds, 2020). This deterioration may be related to the reduced school provision during lockdown, missing many of the usual school rituals and structures, including interactions with peers and school staff, structured and unstructured daytime activities and systems of support (Courtney et al., 2020; Golberstein et al., 2020; Lee, 2020; YoungMinds, 2020). Notably, female pupils were especially at increased risk of depression, consistent with other studies (Altemus et al., 2014). Self-reported food poverty was higher than self-reported eligibility for free school meals, and pupils reporting food poverty had more than three times the odds of having depression or anxiety. In contrast, self-reported access to free school meals was not reliably related to any of the mental health outcomes and did not differ according to educational provision, possibly due to pupils not being fully aware of which category they might fall into.

In addition to the established risk factors of vulnerability, deprivation and gender, a slightly increased risk was identified on all three mental health outcomes for factors especially relevant to lockdown. The first of these is the increased risk for pupils in school years 10 and 12, who were invited back to school from June to help them prepare for key examinations in 2021. These pupils were hypothesised to represent a group at increased risk during lockdown because they might experience longer term impacts of school closures on their educational and vocational outcomes, a hypothesis which is in line with responses to some smaller surveys suggesting that young people's mental health was impacted during lockdown by concerns about academic performance (NHS Digital, 2020; YoungMinds, 2020). These concerns might relate to reduced instruction time for key components of the curriculum, which could be more detrimental for pupils approaching national examinations. Pupils whose parents were leaving the house to go to work most days during lockdown were also more likely to report their wellbeing to be worse, potentially reflecting concerns about their parents or increased stress and reduced support at home (Clemens et al., 2020). The fact that pupils of essential workers were at increased risk even when controlling for other key characteristics, including in-school provision, supports the allocation of school places to these pupils during lockdowns and has implications for targeting support.

CONCLUSION

Our findings highlight how young people already at increased risk of mental health difficulties due to existing deprivation and vulnerabilities were more likely to perceive their mental wellbeing to have deteriorated during lockdown, amplifying existing inequalities. In addition to the predicted vulnerabilities, two new risk groups were identified as being affected by the pandemic: adolescents approaching key examinations in the next academic year and those whose parents were likely performing essential roles during lockdown. This study raises awareness of the needs of those children who are most likely eligible for in-school places during periods of national lockdown. Compared to their counterparts who are accessing educational provision at home, those meeting the criteria for in-school places will likely need good access to a range of educational and social activities to alleviate some of the additional pressures experienced during national lockdown periods as well as access to pastoral support.

CONFLICT OF INTEREST STATEMENT

The authors have declared that they have no competing or potential conflicts of interest.

ETHICAL STATEMENT

Ethical approval was granted by the University of Oxford Medical Sciences Division Research Ethics Committee (Reference R62366/RE0010 and RE0011). Participating schools sent a detailed information sheet to parents of school pupils under 16 years of age, with contact details for the research team and instructions on how to opt-out by contacting the school. All pupils were asked to watch a 3-min video about the survey content and how the responses are stored and used. Pupils provided active assent to participate in the research (for pupils under 16) or informed consent (for those over 16) before starting the survey. It was made clear to pupils both before and within the survey that the research was voluntary and that they could skip any questions that they did not want to answer.

AUTHOR CONTRIBUTIONS

Karen Mansfield and Mina Fazel conceptualised the study, acquired funding, designed and carried out the investigation and coordinated the research planning and project administration. Karen Mansfield, Danielle Newby, Christoph Jindra and Galit Geulayov performed data curation. Karen Mansfield, Danielle Newby, Christoph Jindra, Nemanja Vaci and John Gallacher developed the methodology and formal analysis models. Danielle Newby developed the analysis code and ran the models, including validation using a sensitivity analysis, with support from Karen Mansfield and Galit Geulayov. Karen Mansfield, Emma Soneson, John Gallacher and Mina Fazel prepared the original draft of the manuscript and contributed to the general presentation. Galit Geulayov, Danielle Newby, Christoph Jindra and Nemanja Vaci critically reviewed and edited the draft. All authors checked and approved the final version of the manuscript.