AbstractPurpose

This study examines the mental health impact of the COVID-19 pandemic on adolescents in the United Kingdom as well as social, demographic, and economic variations in the impact.

Methods

Nationally representative longitudinal panel data from the Understanding Society COVID-19 survey were analyzed. The analytical sample comprises 886 adolescents aged 10–16 years surveyed both before and during the pandemic. The Strengths and Difficulties Questionnaire was used to measure adolescents' mental health.

Results

The results from person fixed-effects regression models show that adolescents with better-than-median mental health before the pandemic have experienced an increase in their emotional problems, conduct problems, hyperactivity, and peer relationship problems, but a decrease in their prosocial tendency during the pandemic. In contrast, adolescents with worse-than-median mental health before the pandemic have experienced opposite changes in each Strengths and Difficulties Questionnaire domain. Further results from lagged dependent variable regression models show that compared with girls, boys have experienced a smaller increase in emotional problems but a greater decrease in prosocial tendency. The negative mental health impact is particularly prominent among adolescents in one-parent, one-child, and low-income households. Adult household members' COVID-19 symptoms and illness have undermined adolescents' peer relationships.

Conclusions

The results reveal the pandemic's diverse impacts on adolescent mental health, which vary with adolescents' prepandemic mental health and sociodemographic backgrounds. The findings underline the need for tailored mental health support for adolescents and targeted measures to mitigate inequalities in the mental health impact of the pandemic.

KeywordsImplications and Contribution

This study provides the first nationally representative evidence regarding the impact of the COVID-19 pandemic on adolescent mental health in the United Kingdom. It reveals considerable disparities in the impact across social, demographic, and economic groups.

The outbreak of the COVID-19 pandemic has had wide-ranging impacts on individual well-being. A growing body of research has shown that the pandemic and its associated lockdown, school closure, and social distancing measures have considerably undermined adults' mental health [1Pfefferbaum B. North C.S. Mental health and the COVID-19 pandemic., 2Prevalence and predictors of general psychiatric disorders and loneliness during COVID-19 in the United Kingdom., 3Bu F. Steptoe A. Mak H.W. Fancourt D. Time-use and mental health during the COVID-19 pandemic: A panel analysis of 55,204 adults followed across 11 weeks of lockdown in the UK.]. However, less attention has been paid to the mental health impact of the pandemic on adolescents [[4]Guessoum S.B. Lachal J. Radjack R. et al.Adolescent psychiatric disorders during the COVID-19 pandemic and lockdown.,[5]Youth mental health before and after the control of the coronavirus disease 2019: A nationally representative cohort study of Chinese college students.]. Although adolescents are less likely than adults to become severely ill owing to COVID-19 [[6]Naja M. Wedderburn L. Ciurtin C. COVID-19 infection in children and adolescents.], the pandemic may impair adolescents' mental health through stress spillover from other family members, contextual and policy changes, and the disruption of everyday life routines [[4]Guessoum S.B. Lachal J. Radjack R. et al.Adolescent psychiatric disorders during the COVID-19 pandemic and lockdown.,7Qi M. Zhou S.J. Guo Z.C. et al.The effect of social support on mental health in Chinese adolescents during the outbreak of COVID-19., 8Oosterhoff B. Palmer C.A. Wilson J. Shook N. Adolescents’ motivations to engage in social distancing during the COVID-19 pandemic: Associations with mental and social health., 9Mental health effects of school closures during COVID-19.]. Therefore, our research is motivated by a need to address the relative scarcity of research examining adolescent mental health during the pandemic.In the nascent body of research exploring the impact of the COVID-19 pandemic on adolescent mental health [[4]Guessoum S.B. Lachal J. Radjack R. et al.Adolescent psychiatric disorders during the COVID-19 pandemic and lockdown.,[5]Youth mental health before and after the control of the coronavirus disease 2019: A nationally representative cohort study of Chinese college students.,[7]Qi M. Zhou S.J. Guo Z.C. et al.The effect of social support on mental health in Chinese adolescents during the outbreak of COVID-19.,[9]Mental health effects of school closures during COVID-19.,[10]Ragavan M.I. Culyba A.J. Muhammad F.L. Miller E. Supporting adolescents and young adults exposed to or experiencing violence during the COVID-19 pandemic.], a few notable limitations exist. First, existing research tends to draw on cross-sectional data collected during the pandemic [[7]Qi M. Zhou S.J. Guo Z.C. et al.The effect of social support on mental health in Chinese adolescents during the outbreak of COVID-19.,[11]Chinese adolescent research under COVID-19.,[12]Rogers A.A. Ha T. Ockey S. Adolescents’ perceived socio-emotional impact of COVID-19 and implications for mental health: Results from a US-based mixed-methods study.]. While researchers have reported clear evidence of a negative association between adolescents' COVID-19 exposure and their mental well-being [[7]Qi M. Zhou S.J. Guo Z.C. et al.The effect of social support on mental health in Chinese adolescents during the outbreak of COVID-19.,[12]Rogers A.A. Ha T. Ockey S. Adolescents’ perceived socio-emotional impact of COVID-19 and implications for mental health: Results from a US-based mixed-methods study.], analysis of cross-sectional data collected during the pandemic is insufficient in capturing the impact of the pandemic. Second, understandably owing to limited opportunities for face-to-face contact during the pandemic, existing research is mostly based on online surveys of convenience samples of adolescents [[4]Guessoum S.B. Lachal J. Radjack R. et al.Adolescent psychiatric disorders during the COVID-19 pandemic and lockdown.,[7]Qi M. Zhou S.J. Guo Z.C. et al.The effect of social support on mental health in Chinese adolescents during the outbreak of COVID-19.,[11]Chinese adolescent research under COVID-19.,[13]Youth mental health before and after the control of the coronavirus disease 2019: A nationally representative cohort study of Chinese college students.]. However, online surveys may underrepresent adolescents with limited access to the Internet. Because adolescents with limited digital connections are more likely to experience social isolation and a lack of Internet access is often associated with socioeconomic deprivation [[14]Ellis W.E. Dumas T.M. Forbes L.M. Physically isolated but socially connected: Psychological adjustment and stress among adolescents during the initial COVID-19 crisis.], existing analyses based on online survey data may have underrepresented adolescents who are particularly vulnerable to the pandemic. Third, despite growing evidence for the overall impact of the COVID-19 pandemic on adolescent mental health [7Qi M. Zhou S.J. Guo Z.C. et al.The effect of social support on mental health in Chinese adolescents during the outbreak of COVID-19., 8Oosterhoff B. Palmer C.A. Wilson J. Shook N. Adolescents’ motivations to engage in social distancing during the COVID-19 pandemic: Associations with mental and social health., 9Mental health effects of school closures during COVID-19., 10Ragavan M.I. Culyba A.J. Muhammad F.L. Miller E. Supporting adolescents and young adults exposed to or experiencing violence during the COVID-19 pandemic., 11Chinese adolescent research under COVID-19.,[13]Youth mental health before and after the control of the coronavirus disease 2019: A nationally representative cohort study of Chinese college students.], we still know relatively little about how the impact may vary across different groups of adolescents characterized by distinct demographic attributes, socioeconomic backgrounds, and household members' experiences of COVID-19.

This study addresses the aforementioned gaps by providing the first nationally representative evidence of the pandemic's impact on adolescent mental health. To do so, we analyze new, high-quality nationwide data collected from adolescents aged 10–16 years in the United Kingdom (UK) using a postal paper questionnaire. The survey collected data from the same group of adolescents both before the outbreak of the COVID-19 pandemic in the UK and during the pandemic in July 2020. The two-wave panel allows us to estimate the mental health impact of the pandemic. We also examine how the impact varies with the adolescents' age, gender, ethnicity, family circumstances, parents' socioeconomic status, and adult household members' experiences of COVID-19 symptoms and illness.

Methods Data and sampleWe analyze data from the youth panel of the Understanding Society (USOC) COVID-19 survey and the preceding main USOC survey (for more information, see www.understandingsociety.ac.uk/topic/covid-19). Initiated in 2009, the USOC is a nationally representative longitudinal panel survey [[15]Understanding Society: Design overview.]. The survey contains a youth panel, which surveyed adolescents aged between 10 and 16 years across the UK both before COVID-19 (before March 2020, with more than 99.0% of the adolescents surveyed before 2020) in the preceding main survey and during the COVID-19 pandemic (in July 2020) in the USOC COVID-19 survey. While existing research on adolescents' mental health during COVID-19 often relies on parents' reports [[16]Racine N. Cooke J.L. Eirich R. et al.Child and adolescent mental illness during COVID-19: A rapid review.], the USOC data provide us with a unique opportunity to examine adolescents' own reports. Particularly, we combine data from the main USOC survey and the COVID-19 survey to trace continuity and changes in adolescents' mental health before and during the pandemic. The USOC survey has also collected data from adults, including parents, living in the same household as the adolescents. The main USOC survey collects data from face-to-face interviews, complemented by mixed-mode techniques [[15]Understanding Society: Design overview.]. In collecting information on adolescent mental health, a self-completion paper questionnaire was used in both the main and COVID-19 youth surveys, which ensures consistency in data collection methods and that the sample represents adolescents with differential access to the internet. A survey weight provided by the USOC team is used in our analysis where appropriate to adjust for complex survey design and potential nonresponse bias.To construct our analytical sample, we limited the sample to 901 adolescents who were surveyed both before and during the COVID-19 pandemic. In preparing the data set, we used as many sources as possible (e.g., from the adult and household panels) for constructing time-invariant variables to minimize the rate of missing values. Then, we list-wise deleted 15 adolescents with missing values on the variables used. Our final analytical sample contains 886 adolescents aged 10–16 years in July 2020 (M = 13.3, SD = 1.38) who were surveyed both before and during COVID-19—that is, each adolescent was observed twice (N = 1,772 person-wave observations). The characteristics of the analytical sample are presented in Table 1.

Table 1Sample characteristics

Min. = minimum; Max. = maximum; N = 886 adolescents who were observed twice—before and during the COVID-19 pandemic; Prop. = proportion; Ref. = reference category; SD = standard deviation; SDQ = Strengths and Difficulties Questionnaire.

Dummy variables are indicated by a minimum value of 0 and a maximum value of 1. For prosocial tendency, a higher score indicates a better state of mental health, which is opposite to the other four SDQ subscales. Weighted statistics with unweighted sample sizes.

The secondary data analyzed here were collected under a protocol compliant with the Helsinki Declaration on human subjects testing, with full ethical approval at the institution of the original data collector (i.e., the University of Essex), and with informed consent from all participants (www.understandingsociety.ac.uk/documentation/mainstage/user-guides/main-survey-user-guide/ethics). The anonymized data set was then released through the UK Data Archive. Because we conducted only analysis of a fully anonymized secondary data set and we had agreed to the relevant conditions of confidentiality and terms of use set out by the UK Data Archive, further ethical approval from our own institutions was not required. Measures Adolescent mental healthThe USOC measured adolescent mental health using the Strengths and Difficulties Questionnaire (SDQ)—a standard set of measures that are widely used to capture adolescents' self-reported mental health [[17]Goodman R. Ford T. Simmons H. et al.Using the Strengths and Difficulties Questionnaire (SDQ) to screen for child psychiatric disorders in a community sample.]. The SDQ contains 25 items covering five subscales (five items for each subscale, see www.sdqinfo.org for item wordings): emotional problems, conduct problems, hyperactivity, peer relationship problems, and prosocial tendency. The response to each item was recorded using three options—not true (0), somewhat true (1), and certainly true (2)—denoting whether an adolescent had experienced a given mental health issue. The scores for each subscale, ranging from 0 to 10, were calculated by summing up the scores for its constituent items. A higher score indicates a worse state of mental health for the first four subscales, whereas a higher score indicates better mental health for the prosocial tendency subscale. Our further analysis shows that each SDQ subscale has good reliability and internal consistency, with a Cronbach's alpha of .6 or greater and an eigenvalue from factor analysis greater than 1. Alt though the subscales can be summed up to generate a total SDQ score, we analyze the separate subscales to provide a nuanced understanding of the impact of the pandemic on adolescent mental health. Age

Adolescents' age at the time of the COVID-19 survey (in July 2020) is measured using a continuous variable.

 Gender

We use a dummy variable to distinguish between girls (=0; 52%) and boys (=1; 48%). The USOC surveys did not ask about adolescents' nonbinary gender identity.

 Ethnic minorityOwing to notable ethnic inequalities in COVID-19 infection and mortality rates and in the cascading socioeconomic impact of the pandemic in the UK [[18]Evidence for ethnic inequalities in mortality related to COVID-19 infections: Findings from an ecological analysis of England.,[19]Intersecting ethnic and native–migrant inequalities in the economic impact of the COVID-19 pandemic in the UK.], we use a dummy variable to distinguish between white British/Irish adolescents (=0; 84%) and ethnic minority adolescents (=1; 16%). Small cell sizes prevented us from disaggregating ethnic minority adolescents into more detailed subgroups. Coresidence with other children during COVID-19

The presence of other child(ren) may make a substantial difference to one's opportunity for peer interaction. Thus, we use a dummy variable to distinguish whether an adolescent lived with at least one other child younger than 18 years of age (62%) during the pandemic.

 One-parent householdWe use a dummy variable to distinguish whether an adolescent lived with one or both parent figures (including both biological parents and stepparents). We measure parents' coresidential status rather than their marital or relationship status because parents' physical presence or absence in the household is more salient in shaping adolescents' wellbeing [[20]Brown S.M. Doom J.R. Lechuga-Peña S. et al.Stress and parenting during the global COVID-19 pandemic.,[21]The role of parent-adolescent relationships in the development of (pre) hypertension in young adulthood in the US.]. Parents' prepandemic income level

As the impact of the COVID-19 pandemic and adolescents' resilience may vary with their socioeconomic positions, we measure parents' income before the pandemic and divide the sample into three equal groups to distinguish adolescents from low-, middle-, and high-income families.

 COVID-19 symptoms and illness in the householdAlthough adolescents are unlikely to exhibit (severe) COVID-19 symptoms [[6]Naja M. Wedderburn L. Ciurtin C. COVID-19 infection in children and adolescents.], the presence of COVID-19 symptoms or illness in the household not only limits adolescents' socializing opportunities owing to state-mandated self-isolation, it can also place a mental strain on the adolescents. Based on the monthly COVID-19 adult surveys (April–July 2020), we create a dummy variable to capture whether any adult household member who responded to the COVID-19 surveys currently have or had ever reported COVID-19 symptoms or illness between April and July 2020 (13%). Pandemic “treatment” variable

We use a dummy variable to distinguish the observations before and during the pandemic for each adolescent.

In our supplementary analysis, we also tested additional covariates, including rural-urban residence, household size, household income loss during COVID-19, adult household members' essential/key worker status, and parental education. These variables were not associated with changes in adolescent mental health and including them neither affected the results for other variables nor improved the overall model fit. Therefore, they are not included in the analysis presented in this article.

 Statistical analysisWe conduct the analysis in two steps. First, we run person fixed-effects models to estimate changes in the adolescents' mental health before and during the pandemic. As the distributions of the SDQ subscale scores were within a range sufficient to assume a normal distribution, we used a linear specification for the models. A fixed-effects regression model effectively captures within-person change in mental health by controlling for each adolescent's time-invariant attributes that are both observed and unobserved [[22]Fixed effects regression models.]. In the models, we use the pandemic “treatment” variable as the sole predictor. As adolescents with different mental health states before the pandemic may be susceptible to the pandemic in distinct ways, we also fit separate person fixed-effects models for adolescents falling above and below the median of the prepandemic distribution of each SDQ subscale.Then, we fit lagged dependent variable models to estimate how the mental health impact of the pandemic varies with the covariates [[23]Two-wave panel analysis: Comparing statistical methods for studying the effects of transitions.]. In the models, the adolescents' SDQ subscale scores during the pandemic are regressed on the covariates, while controlling for adolescents' prepandemic SDQ subscale scores. The lagged dependent variable approach explicitly models changes in mental health by conditioning the adolescents' mental health before the pandemic; thus, the coefficients for the covariates in the models denote how the predictors relate to changes in SDQ scores [[23]Two-wave panel analysis: Comparing statistical methods for studying the effects of transitions.]. While it is less straightforward to model multiple time-invariant covariates in fixed-effects models [[22]Fixed effects regression models.], our robustness checks using a fixed-effects approach yielded findings that are consistent with those reported in the following text. As it is possible for multiple adolescents to live in the same household, we estimate robust standard errors to account for sample clustering at the household level [[24]Robust standard errors for panel regressions with cross-sectional dependence.].Results The impact of the COVID-19 pandemic on adolescent mental healthFigure 1 presents the histograms of the adolescents' SDQ subscale scores before and during the pandemic. The results show a decrease in the proportion of adolescents with a low level of emotional problems and peer relationship problems (i.e., scores 0–1) and those with a high level of prosocial tendency (i.e., score 9–10), suggesting worsened adolescent mental health during the pandemic. However, during the pandemic, the proportion of adolescents exhibiting little or no conduct problems has increased, possibly because more time spent at home and under parental supervision [[20]Brown S.M. Doom J.R. Lechuga-Peña S. et al.Stress and parenting during the global COVID-19 pandemic.] prevented adolescents from enacting problem behaviors such as fighting with or bullying other children.

Figure 1Histograms of adolescents' SDQ scores before and during the COVID-19 pandemic. Note: SDQ = Strengths and Difficulties Questionnaire. N = 886 adolescents who were observed twice—before and during the COVID-19 pandemic. For prosocial tendency, a higher score indicates a better state of mental health, which is opposite to the other four SDQ subscales. Unweighted results, as probability weights cannot be applied in histograms.

Figure 2 depicts the impact of the COVID-19 pandemic on adolescent mental health and how the impact varies with the adolescents' prepandemic mental health, estimated using person fixed-effects models. In the full sample, although changes in some SDQ subscales are statistically significant at the 5% level, the effect sizes seem small across the board. Specifically, compared with before the pandemic, UK adolescents have experienced an increase in emotional problems (B = .23; 95% CI: .09, .38; p = .002) and peer relationship problems (B = .27; 95% CI: .15, .40; pB = –.13, 95% CI: –.26, –.01; p = .030). Meanwhile, the prevalence of conduct problems has decreased during the pandemic (B = –.18, 95% CI: –.30, –.07; p = .002). However, no change has been found in the level of hyperactivity(B = –.01; 95% CI: –.16, .14; p = .847).

Figure 2The impact of the COVID-19 pandemic on adolescents' SDQ scores, estimated using person fixed-effects models. Note: SDQ = Strengths and Difficulties Questionnaire. Color bands indicate 95% confidence intervals. Better-than-median mental health refers to the lower 50% of SDQ scores for the first four subscales and the higher 50% of scores for the prosocial tendency subscale, and worse-than-median mental health refers to the higher 50% of SDQ scores for the first four subscales and the lower 50% of scores for the prosocial tendency subscale. For prosocial tendency, a higher score indicates a better state of mental health, which is opposite to the other four SDQ subscales. Effects estimated using person fixed-effects models, controlling for within-person time-invariant traits. N = 886 adolescents who were observed twice—before and during the COVID-19 pandemic. Weighted statistics with unweighted sample size.

After separating the sample based on the adolescents' prepandemic mental health, we find that the small effect sizes in the overall sample are owing to heterogeneous trends in different subgroups. On the one hand, adolescents with better-than-median mental health before the pandemic have experienced a notable increase in emotional problems (B = 1.05; 95% CI: .88, 1.22; p < .001), conduct problems (B = .28; 95% CI: .16, .39; p < .001), hyperactivity (B = .64; 95% CI: .46, .82; p < .001), and peer relationship problems (B = .94; 95% CI: .79, 1.08; p < .001), and they have also become less prosocial (B = –.89; 95% CI: –1.03, –.75; p < .001). On the other hand, opposite trends are noted among adolescents with worse-than-median mental health before the pandemic—that is, a marked decrease in emotional problems (B = –.49; 95% CI: –.70, –.28; p < .001), conduct problems (B = –.96; 95% CI: –1.19, –.73; p < .001), hyperactivity (B = –.86; 95% CI: –1.09, –.63; p < .001), and peer relationship problems (B = –.31; 95% CI: –.50, –.13; p = .001) and an increase in prosocial tendency (B = .42; 95% CI: .26, .59; p < .001).

 Sociodemographic variations in the mental health impactsTable 2 shows how changes in the adolescents' mental health from before to during the pandemic vary with their sociodemographic backgrounds and household experiences of COVID-19. We report results that are statistically significant at the 10% level given the relatively small sample size. Compared with their younger counterparts, older adolescents have experienced a greater decrease in conduct problems and a greater increase in prosocial tendency, but they have shown a greater increase in hyperactivity during the pandemic. Boys have experienced a smaller increase in emotional problems but a greater decrease in prosocial tendency than girls. Ethnic minority adolescents have exhibited a slightly smaller increase in peer relationship problems during the pandemic, compared with their white ethnic majority counterparts.

Table 2Lagged dependent variable regression models estimating changes in SDQ scores before and during the COVID-19 pandemic

CI = confidence intervals; Dependent variable = SDQ scores during the COVID-19 pandemic; N = 886 adolescents who were observed twice—before and during the COVID-19 pandemic; Ref. = reference category; SDQ = Strengths and Difficulties Questionnaire.

For prosocial tendency, a higher score indicates a better state of mental health, which is opposite to the other four SDQ subscales. Weighted statistics with unweighted sample size.

+ p < .10, ∗p < .05, ∗∗p < .01, ∗∗∗p < .001.

The results also highlight the important role played by household composition in shaping the impact of the pandemic on adolescent mental health. The presence of other children in the household is found to protect the adolescents from the adverse impact of the pandemic on their emotional and social well-being. As shown in Table 2, the increase in emotional and peer relationship problems is greater among adolescents who did not live with any other child than those living with at least one other child during the pandemic. In addition, compared with adolescents living with both parents, adolescents living in one-parent households have experienced a greater increase in peer relationship problems.

The pandemic's impact on adolescent mental health also varies with parents' socioeconomic positions. Compared with adolescents whose parents fall in the bottom tertile of the prepandemic income distribution, adolescents with high-earning parents have experienced a greater decrease in conduct problems and a smaller increase in hyperactivity and peer relationship problems during the pandemic. Thus, adolescents from low-income families have experienced a greater decline in mental health than their better-off counterparts.

Finally, experiences of COVID-19 symptoms or illness in the household are found to heighten adolescents' risk of peer relationship problems. Compared with adolescents whose household members have not had any COVID-19 symptoms or illness, adolescents living with adult household members who currently have or ever had COVID-19 symptoms or illness have experienced a greater increase in their peer relationship problems during the pandemic.

DiscussionThe COVID-19 pandemic has affected adolescents' mental health in the UK and elsewhere [5Youth mental health before and after the control of the coronavirus disease 2019: A nationally representative cohort study of Chinese college students.,