Remember me
The global COVID-19 pandemic was an extraordinary public health crisis, involving unprecedented public health measures such as social distancing and school and business closures. On March 23, 2020, the UK government announced a national lockdown and urged the public to stay at home to reduce the spread of COVID-19. Lockdown restrictions were relaxed during the summer (May to August 2020), allowing for some easing of social measures, such as permitting 2 households to meet while adhering to social distancing guidelines. The lockdown was reintroduced in winter 2020 to mitigate the increasing transmission and protect the health care system from being overwhelmed. However, such measures may put people at risk of depression, anxiety, stress, and helplessness [,].
Adolescence is a life stage when individuals are susceptible to the onset of mental illness, of which depression and anxiety are most common. Therefore, adolescents may have been more susceptible to the mental health impacts of the public health crisis than adults. A United Kingdom–based survey found that 80% of the respondents believed that the pandemic had worsened their mental health, and 67% of the respondents believed that the pandemic would have a long-term negative effect on their mental health []. Adolescence is a critical period when peer relationships become more influential than family relationships on their life []. Social isolation from peers and loneliness are related to depressive symptoms, suicidal ideation, and anxiety []. In addition, school closures were associated with fear, restlessness, and sadness []. However, whether the impacts of the pandemic on mental health were acutely related to public health measures (eg, mental health status improves as restrictions ease) or enduring (eg, persistently poor mental health status during the pandemic) remains unclear. Understanding these impacts remains a policy priority to protect and improve young people’s mental health [].
Use of mobile phones and other wireless devices is ubiquitous among children and adolescents. In the United Kingdom, 97% of the adolescents aged 12 to 15 years and 100% of the adolescents aged 16 to 17 years own a mobile phone, and most of them use a mobile phone, tablet, or laptop to go to the web []. Our recent systematic review found suggestive but limited evidence indicating associations between greater technology use and poorer mental health in children and adolescents []. More high-quality longitudinal studies with detailed information on digital technology use are needed, particularly in the context of the COVID-19 pandemic when the digital environment is increasingly important for learning, connection, and social support to offset the negative impacts of school closures and social isolation on mental health.
Previous cross-sectional studies found a high prevalence of depression and anxiety in young people during the pandemic [-]. Female gender, older age, and lower socioeconomic status (SES; eg, financial strain and living in rural areas) were associated with more mental health problems []. However, it is difficult to assess the impacts of the pandemic without information on the prepandemic symptom levels for comparison. In addition, other factors related to mental health including increased digital technology use (particularly the use of social network sites [SNSs]) and more sleep problems were reported during the pandemic [,]. Given that the associations between such behaviors and mental health are complex and often bidirectional, longitudinal analysis is necessary to unravel this interrelationship. A few longitudinal studies in adolescents have investigated depression and anxiety before and during the pandemic with mixed results [-]. One study reported an increase in depression and anxiety []. One study found an increase in depression, but anxiety remained stable both before and during the pandemic []. Another survey in school Year 9 students (the third year of secondary school; aged 13 years) found a decrease in anxiety during the UK lockdown []. However, older adolescents in critical examination years (General Certificate of Secondary Education [at the end of compulsory secondary education, aged approximately 16 years] and advanced level [for university admissions, aged approximately 18 years] examinations) may experience more stress owing to missing school time, as the potential impact on future university and employment opportunities is more immediate. However, none of these studies examined the longitudinal associations between digital technology use, sleep, and mental health.
ObjectivesThis study aims to investigate the impact of the COVID-19 pandemic on adolescents’ mental health in a large representative (in terms of gender, ethnicity, and SES) longitudinal adolescent cohort across Greater London (the Study of Cognition, Adolescents and Mobile Phones [SCAMP]), for whom there was detailed information on mental and physical health and digital behaviors such as mobile phone use, SNS use, and video gaming, both before and during the pandemic. Thus, we were in a unique position to assess and tease out potential relationships between depression and anxiety, sociodemographic characteristics, digital technology use, sleep, and COVID-19 infection status and ascertain impacts in this important age group specifically related to the pandemic and public health measures. This will help to identify the groups at greatest risk who may benefit from targeted support.
The SCAMP is a prospective adolescent cohort study that was originally set up to investigate cognitive and behavioral outcomes affected by the use of mobile phones and other wireless technologies that emit radio-frequency electromagnetic fields. The details of this study have been reported previously []. Between November 2014 and July 2016, baseline data were collected from 6581 participants in Year 7 (aged 11 to 12 years) from 39 secondary schools (26 state and 13 independent) in and around Greater London, the United Kingdom. Participants in all SCAMP schools completed a computer-based assessment using the Psytools software (Delosis Ltd) under examination-like conditions in school. The assessment included a questionnaire on their digital technology behaviors (eg, smartphone use, SNS engagement, and video gaming); a battery of cognitive tests; and physical and mental health, lifestyle, and behavior scales. Data collected from 4978 participants at 31 schools between November 2016 and July 2018 (T1) when they were in Year 9 or 10 (aged 13 to 15 years) were included in this study. Participants who had depression (n=3292) and anxiety measures (n=3350) were defined as the T1 cross-sectional sample.
All SCAMP participants were invited to complete another assessment between July 2020 and June 2021 (T2) when they were aged 16 to 18 years, comprising cognitive tests and questionnaires on mental health, digital technology behaviors, home and outdoor environment, lifestyle, and access to public health information. Participants completed the assessment at home from July to September 2020. The assessment from September 2020 onward was conducted in school, supervised by the SCAMP team onsite or remotely via internet depending on school policies. Data collected from 1328 adolescents at T2 were included in this study. Participants who had depression and anxiety measures (n=968) were defined as the T2 cross-sectional sample.
A subset of the participants had longitudinal depression (n=421) and anxiety (n=425) data, defined as the longitudinal sample. In the longitudinal sample, we excluded participants with clinically significant symptoms at T1 when assessing the longitudinal associations between the exposure variables and new incident depression and anxiety. The remaining participants were defined as the sample for incident case analysis (n=364 for incident depression analysis and n=367 for incident anxiety analysis). The sample for incident case analysis had 80% power to detect odds ratio (OR) of 1.82 and 1.84 when assessing the associations between digital technology use (eg, total mobile phone use) in tertile categories and incident depression and anxiety, respectively. shows the structure of the SCAMP cohort data relevant to this study.
Figure 1. Structure of the Study of Cognition, Adolescents and Mobile Phones (SCAMP) cohort data relevant to this study. GAD-7: 7-item Generalized Anxiety Disorder; PHQ-8: 8-item Patient Health Questionnaire; T1: November 2016 to July 2018; T2: July 2020 to June 2021. Measures (Self-Reported Questionnaire Data)Sociodemographic CharacteristicsSociodemographic information including age, gender, ethnicity, and parental occupation was collected. The occupation of both parents was assessed at T1 only. We used the Office for National Statistics classification of occupation, categorizing it as managerial and professional, intermediate, and routine and manual []. The participants were assigned the higher parental occupation level. We also assessed the parental job situation since the COVID-19 lockdown at T2. We dichotomized the parental job situation as change during the pandemic and no change. School type (independent or state) was also included in the analyses. Missing data on ethnicity (262/4978, 5.26%), parental occupation (691/4978, 13.88%), and parental job situation (328/1328, 24.7%) were assigned a “missing” category for each variable rather than being excluded from the analyses.
Digital Technology Use (Total Mobile Phone Use, SNS Use, and Video Gaming)At T1, participants reported the daily duration of phone calls and internet use (eg, web browsing, WhatsApp, Facebook, YouTube, and any other web-based apps) on mobile phones separately for weekdays and weekends. At T1, total mobile phone use was calculated as the combined daily duration of phone calls and internet use on mobile phones, with the average taken over both weekdays and weekends. At T2, participants reported an average daily duration of total mobile phone use for any purpose in the previous week.
At T1, participants reported the daily duration of SNS (eg, Facebook, Instagram, and Twitter) use on mobile phones and other devices separately for weekdays and weekends. At T1, SNS use was calculated as the combined daily duration of SNS use on mobile phones and other devices, with the average taken over both weekdays and weekends. At T2, participants reported the average time spent on specific SNS platforms (regardless of device) per day, including Facebook, Instagram, Twitter, TikTok, Snapchat, and YouTube in the last week. SNS use at T2 was calculated as the combined duration of SNS use on different platforms.
At T1, participants reported the daily duration of playing video games on any device at T1 separately for weekdays and weekends. Video gaming at T1 was the average weekday and weekend daily duration. At T2, participants reported an average daily duration of playing video games on any device in the last week.
Categorical responses were provided for questions on total mobile phone use, SNS use, and video gaming. To enable calculation and combination, we took the midpoints of category intervals, except in the highest category where we used the lowest value. For instance, the category “11-30 minutes per day” was converted to 20.5 minutes per day or 0.34 hours per day, whereas “more than 5 hours per day” was converted to 5 hours per day. Details for response category intervals for each question are shown in Table S1 in . Participants were categorized into 3 groups based on the tertiles of daily duration of total mobile phone use, SNS use, and video gaming at each time point.
SleepAt T1, participants provided information on when they usually got into bed, how long it took them to fall asleep, and what time they usually woke up separately for weekdays and weekends. The sleep duration was derived from these responses. The details have been reported elsewhere []. At T2, participants reported an average duration of sleep each night in the past 4 weeks. At T1 and T2, a sleep duration of <7 hours and >10 hours per day was defined as insufficient sleep and oversleep, respectively [].
COVID-19 Infection StatusAt T2, participants were asked if they had had a confirmed (by a positive test) or suspected COVID-19 infection (by a physician or themselves, but not tested).
Mental Health Outcomes—Depression and AnxietyDepression was assessed using the 8-item Patient Health Questionnaire (PHQ-8). A summary score of ≥10 is considered indicative of clinically significant depression []. Cronbach α for the PHQ-8 score was .86 in the T1 cross-sectional sample and .88 in the T2 cross-sectional sample, indicating good internal reliability. To our knowledge, no previous study has examined the validity of the PHQ-8 in adolescents, but the PHQ-9 has demonstrated good validity in adolescents []. Anxiety was assessed using the 7-item Generalized Anxiety Disorder (GAD-7) scale. A summary score of ≥10 is considered indicative of clinically significant anxiety []. Cronbach α for the GAD-7 score was .89 in the T1 cross-sectional sample and .91 in the T2 cross-sectional sample, indicating good internal reliability. Evidence from current literature has indicated good validity and reliability of the GAD-7 in adolescents [].
Statistical AnalysesWe tested the significance of the difference between the T1 and T2 time-varying factors in participants with data at both time points. We performed multivariable logistic regression in T1 and T2 cross-sectional samples to assess the associations of sociodemographic characteristics, digital technology use, and sleep with clinically significant depression and anxiety at ages 13 to 15 years and 16 to 18 years. In the T2 cross-sectional sample, we also assessed whether mental health differed by public health measures (eg, lockdown, school closures, and school reopening), using time as a proxy. We categorized it as summer holiday (July to August 2020), school opening (September to December 2020 and March to June 2021), and school closures (January to February 2021). To assess whether the associations between COVID-19 infection and mental health were attributed to residual confounding bias, we used depression and anxiety at T1 as the negative control dependent variables. We expected no significant association between COVID-19 infection status at T2 and depression or anxiety at T1.
Multivariable logistic regression was performed to assess longitudinal associations of sociodemographic characteristics, digital technology use, and sleep with new incident depression and anxiety at T2 in the sample for incident case analysis. Multiple logistic regression was used to yield ORs and 95% CIs. All variables were mutually adjusted in the models to examine the associations between sociodemographic factors and depression and anxiety. Depression and anxiety associated with total mobile phone use, SNS use, video gaming, sleep, and COVID-19 infection status were analyzed separately for each exposure variable (to avoid potential collider bias) by adjusting for all sociodemographic variables.
We conducted sensitivity analyses to examine cross-sectional associations between sociodemographic variables, digital technology use, sleep, COVID-19 infection status, and mental health outcomes at T1 and T2 in the longitudinal sample only. These analyses aimed to ascertain whether the associations observed in the longitudinal sample were substantially different from those observed in T1 and T2 cross-sectional samples. We performed an additional sensitivity analysis to investigate both cross-sectional and longitudinal associations between SNS use on mobile phones and depression and anxiety. This was done to determine if the patterns of association differed from those observed between SNS use on any device and mental health.
We did not use survey weights in our analyses, as the SCAMP is a school-based study that did not use probability sampling in participant recruitment. All analyses in this study were complete case analyses. Multiple imputation is not appropriate in our study, as it is probable that our data are missing not at random (ie, missingness is related to unobserved data). Participation in the T2 assessment might have been self-selected, given that data collection was mostly undertaken in a remote setting []. A P value of <.05 was considered statistically significant. All analyses were performed using R (version 4.0.3; R Foundation for Statistical Computing) and STATA (version 16.0; StataCorp).
Ethical ConsiderationsThe North-West Haydock Research Ethics Committee approved the SCAMP study protocol and its subsequent amendments (#14/NW/0347). School head teachers consented to participate in the SCAMP. Participants received written information about the study in advance and were given the option to withdraw from the research at any time. This study was conducted in accordance with the Declaration of Helsinki.
shows that our sample was diverse in terms of sociodemographic characteristics. Table S2 in shows that the level of total mobile phone and SNS use significantly increased at T2 (P values for paired t test <.001). The prevalence of insufficient sleep also significantly increased at T2 (P value for 2-proportion z test <.001). The proportion of depression and anxiety significantly increased from 13.5% (57/421) at T1 to 33.3% (140/421) at T2 and from 13.6% (58/425) at T1 to 29.4% (125/425) at T2 (both P values for 2-proportion z test <.001), respectively.
Table 1. Sociodemographic characteristics, digital technology use, sleep, COVID-19 infection status, and depression and anxiety in the SCAMP cohort at T1a and T2b.CharacteristicsT1 (n=4978)T2 (n=1328)Age (y), median (IQR)14.02 (13.11-14.06)17.06 (17.01-17.1)Gender, n (%)aT1: November 2016 to July 2018.
bT2: July 2020 to June 2021.
cParental occupation was asked at T1 only; we assumed that parental occupation remained the same at T2.
dN/A: not applicable.
eSNS: social network site.
Results From Cross-Sectional Analysesand show analyses in T1 and T2 cross-sectional samples. shows that compared with male participants, female participants had higher odds of presenting depression (T1: OR 2.64, 95% CI 2.15-3.25; T2: OR 2.59, 95% CI 1.9-3.52) and anxiety (T1: OR 3.3, 95% CI 2.62-4.15; T2: OR 2.55, 95% CI 1.83-3.54) at both time points. At T1, older participants and those from state schools had higher odds of depression and anxiety than younger participants and those from independent schools. There were no associations between mental health measures and ethnicity or parental occupation at any time point. Depression and anxiety at T2 did not differ significantly between the summer holiday, school opening, and school closures. In sensitivity cross-sectional analyses restricted to the longitudinal sample, although there was some variation in OR values, the cross-sectional associations with gender persisted. However, the associations with age and school type at T1 were no longer statistically significant (Table S3 in ). It is worth noting that the smaller sample size in the longitudinal sample might have reduced the power to detect associations.
Table 2. The associations between sociodemographic factorsa and the presence of depression and anxiety in T1b and T2c cross-sectional samples.Outcome and exposureT1, ORd (95% CI)T2, OR (95% CI)DepressionaAll sociodemographic variables were mutually adjusted at T1 and T2.
bT1: November 2016 to July 2018.
cT2: July 2020 to June 2021.
dOR: odds ratio.
eN/A: not applicable.
Table 3. The associationsa between digital technology use, sleep, COVID-19 infection status, and the presence of depression and anxiety in T1b and T2c cross-sectional samples.Outcome and exposureT1, ORd (95% CI)T2, OR (95% CI)DepressionaMental health measures in relation to each exposure were analyzed separately by adjusting for the confounders as follows: T1 analysis: adjusted for age, gender, ethnicity, parental occupation, and school type; T2 analysis: adjusted for age, gender, ethnicity, parental occupation, parental job situation, school type, and time of data collection.
bT1: November 2016 to July 2018.
cT2: July 2020 to June 2021.
dOR: odds ratio.
eAt T1, P for trend <.001; at T2, P for trend <.001.
fSNS: social network site.
gAt T1, P for trend <.001; at T2, P for trend =.01.
hAt T1, P for trend <.001; at T2, P for trend =.14.
iCOVID-19 infection status reflects whether a participant has ever had a confirmed or suspected COVID-19 infection at any time before T2. Depression and anxiety at T1 were negative control outcomes when assessing the association between the COVID-19 infection status and mental health.
jAt T1, P for trend <.001; at T2, P for trend =.03.
kAt T1, P for trend <.001; at T2, P for trend =.001.
lAt T1, P for trend =.01; at T2, P for trend =.69.
shows that the total use of mobile phones and SNSs was associated with depression and anxiety at both time points with a dose-response relationship (all P values for trend <.05). Video gaming was associated with depression and anxiety at T1 with a dose-response relationship (both P values for trend <.05). Insufficient sleep was associated with depression (T1: OR 2.73, 95% CI 2.22-3.34; T2: OR 3.6, 95% CI 2.52-5.14) and anxiety (T1: OR 2.34, 95% CI 1.89-2.9; T2: OR 3.08, 95% CI 2.12-4.46) at both time points. Oversleep was associated with depression at both time points, with a stronger association at T2 (T1: OR 1.6, 95% CI 1.02-2.49; T2: OR 10.79, 95% CI 4.12-28.29). Suspected COVID-19 infection was associated with depression and anxiety at T2. This was not the case for confirmed COVID-19 infection; however, the low numbers of participants in this category (n<40) limit the sensitivity of this analysis. However, no significant associations between suspected or confirmed COVID-19 infection and depression or anxiety were observed at T1, as expected. In general, these cross-sectional associations were similar in the longitudinal sample, although some associations were no longer significant (Table S4 in ).
Results From Longitudinal AnalysesAfter excluding participants with preexisting depression or anxiety at T1, female participants had higher odds of new incident depression (OR 2.5, 95% CI 1.5-4.18) and anxiety (OR 2.11, 95% CI 1.23-3.61) at T2 than male participants (). Compared with White ethnicity, ethnic minority was not associated with greater new incident depression or anxiety at T2. Black ethnicity was associated with a lower incidence of anxiety at T2 (OR 0.33, 95% CI 0.09-1.22), although the association was marginally significant. Longitudinal associations between age and school type and the incidence of depression and anxiety were not marked.
Table 4. The longitudinal associations between sociodemographic factorsa and the incidence of depression and anxiety at T2b in the sample for incident case analysis.ExposureIncidence of depression at T2c, ORd (95% CI)Incidence of anxiety at T2e, OR (95% CI)Age at T1f (per 1-y increase)0.79 (0.47-1.33)0.81 (0.47-1.40)Age at T2 (per 1-y increase)0.82 (0.51-1.32)0.70 (0.42-1.15)GenderaAll sociodemographic variables were mutually adjusted.
bT2: July 2020 to June 2021.
cParticipants with clinically significant depression at T1 were excluded.
dOR: odds ratio.
eParticipants with clinically significant anxiety at T1 were excluded.
fT1: November 2016 to July 2018.
Compared with the participants who used mobile phones in the lowest tertile, those in the highest tertile had higher odds of new incident depression at T2 (OR 1.89, 95% CI 1.02-3.49; ). SNS use and video gaming at T1 were not associated with the development of depression or anxiety at T2. Moderate SNS users (ie, those in the 2nd tertile) had slightly lower odds of new incident depression at T2, although this association was not significant. The associations between SNS use on mobile phones and mental health outcomes were generally similar to those found between SNS use on any device and mental health outcomes (Table S5 in ). Insufficient sleep at T1 was also associated with new incident depression at T2 (OR 2.26, 95% CI 1.31-3.91).
Table 5. The longitudinal associationsa between digital technology use and sleep at T1b and the incidence of depression and anxiety at T2c in the sample for incident case analysis.Exposure at T1Incidence of depression at T2d, ORe (95% CI)Incidence of anxiety at T2f, OR (95% CI)Total mobile phone usegaAdjusted for age at T1 and T2, gender, ethnicity, parental occupation, and school type at T1.
bT1: November 2016 to July 2018.
cT2: July 2020 to June 2021.
dParticipants with clinically significant depression at T1 were excluded.
eOR: odds ratio.
fParticipants with clinically significant anxiety at T1 were excluded.
gFor depression, P for trend =.05; for anxiety, P for trend =.26.
hSNS: social network site.
iFor depression, P for trend =.62; for anxiety, P for trend =.68.
jFor depression, P for trend =.37; for anxiety, P for trend =.11.
To our knowledge, this is the first study to investigate detailed digital technology use in relation to longitudinal changes in mental health from the prepandemic period in a large representative adolescent sample with high sociodemographic diversity. We are also the first to consider mental health associated with public health measures (eg, lockdown, school closures, and school reopening) during the pandemic. We identified clear increases in depression and anxiety symptoms reaching clinical thresholds in adolescents over the course of the COVID-19 pandemic compared with the prepandemic assessments in the same cohort. Female individuals were more likely to develop depression and anxiety during the pandemic than male individuals. Black adolescents were less likely to develop anxiety during the pandemic than White adolescents. Depression and anxiety levels did not differ significantly according to socioeconomic factors or public health measures. High mobile phone use and insufficient sleep in the prepandemic period were associated with new incident depression during the pandemic. SNS use and video gaming were associated with depression and anxiety cross-sectionally only. Suspected COVID-19 infection was associated with increased depression and anxiety.
Comparison With Prior WorkCompared with previous longitudinal studies examining the impacts of the pandemic on adolescent mental health [-,], our sample had greater ethnic and SES diversity, enabling us to examine the associations between ethnicity and SES and mental health. In addition, we investigated mental health related to various aspects of digital technology use and sleep both cross-sectionally and longitudinally to indicate potential risk factors. There were no marked variations in depression and anxiety based on public health measures, suggesting that mental health impacts are not contingent on immediate context but rather arise from broader and potentially more lasting societal and individual influences.
Our findings indicate a marked increase in the prevalence of depression and anxiety during the pandemic. This could be attributed to factors such as enforced stay-at-home measures and school closures, which hindered peer interactions for school-aged adolescents. Such interactions are crucial for maintaining good mental health []. However, our resu
Comments (0)