TeenCovidLife: a resource to understand the impact of the COVID-19 pandemic on adolescents in Scotland

TeenCovidLife is part of Generation Scotland’s CovidLife projects, a set of longitudinal observational studies designed to assess the psychosocial and health impacts of the COVID-19 pandemic. TeenCovidLife focused on how adolescents in Scotland were coping during the pandemic. As of September 2021, Generation Scotland had conducted three TeenCovidLife surveys. Participants from previous surveys were invited to participate in the next, meaning the age ranges shifted over time. TeenCovidLife Survey 1 consists of data from 5,543 young people age 12 to 17, collected from 22 May to 5 July 2020, during the first school closures period in Scotland. TeenCovidLife Survey 2 consists of data from 2,245 young people aged 12 to 18, collected from 18 August to 14 October 2020, when the initial lockdown measures were beginning to ease, and schools reopened in Scotland. TeenCovidLife Survey 3 consists of data from 597 young people age 12 to 19, collected from 12 May to 27 June 2021, a year after the first survey, after the schools returned following the second lockdown in 2021. A total of 316 participants took part in all three surveys. TeenCovidLife collected data on general health and well-being, as well as topics specific to COVID-19, such as adherence to COVID-19 health guidance, feelings about school closures, and the impact of exam cancellations. Limited work has examined the impact of the COVID-19 pandemic on young people. TeenCovidLife provides relevant and timely data to assess the impact of the pandemic on young people in Scotland. The dataset is available under authorised access from Generation Scotland; see the Generation Scotland website for more information.


Introduction
The coronavirus disease 2019 (COVID-19) pandemic has affected the lives of people of all ages across the world. In the UK there have now been two national lockdowns, in which schools and workplaces were closed and non-essential travel was stopped. Moreover, the general population has been asked to make ongoing changes to their lifestyle to minimise the risk of contracting and transmitting the disease. This upheaval to everyday life may have long-term socioeconomic and psychological effects, necessitating careful documentation and study 1 .
This paper describes the TeenCovidLife dataset, a dataset collected by Generation Scotland on the health and well-being of adolescents in Scotland. This dataset is available through authorised access in the UK and abroad for use in research. More details on the Generation Scotland access procedure can be found on the Generation Scotland website. Generation Scotland is a long-running family and populationbased health study. Since 2006, Generation Scotland has been gathering data and collaborating with researchers to produce high-quality health research across many fields 2 . Moreover, longitudinal population studies such as Generation Scotland are particularly well-positioned to study the COVID-19 pandemic 3 . This led to the formulation of the CovidLife project, studying the impact of COVID-19 on over 18,000 adults in the UK 4 .
Findings from the CovidLife and other longitudinal population studies revealed that young adults showed elevated risks of depression and anxiety during the pandemic 5 . A meta-analysis of the prevalence of depression and anxiety among young people throughout the pandemic indicates that prevalence has increased and remains high 6 . Likewise, Chinese adolescents during the early stages of the pandemic showed higher than usual levels of depression and anxiety 7 . This suggests young people's mental well-being may be negatively impacted by the pandemic. Despite this, there is still little cohort research capturing adolescent's direct experiences of the COVID-19 pandemic.
The TeenCovidLife project was designed to address this important gap in the research literature. In this series of three surveys, over 7,000 young people age 12 to 19, living in Scotland, completed questionnaires about their experiences, feelings, and well-being during the COVID-19 pandemic. This complements existing work such as the Co-SPACE stream of resources 8 by capturing the experiences of young people in Scotland in particular, using multiple measures to capture resilience and general well-being.
The first survey was conducted from May to July 2020, during the first pandemic-related school closures in Scotland. The second survey was conducted from August to October 2020, when lockdown measures were being relaxed and schools reopened. The third survey was conducted approximately one year after the first, from May to June 2021, when most schools had reopened after the second national lockdown. This paper is a data note, and as such is intended to describe the TeenCovidLife data, as well as how it was collected, in order to act as reference for future researchers. Analysis and interpretation of the data and its potential implications for health and policy is beyond the scope of the current paper.

Materials & Methods
Questionnaire Development. The TeenCovidLife questionnaires were developed using Qualtrics survey software, a survey development tool 9 , with versions dated May 2020, August 2020 and May 2021 for each wave of the survey. Data collection was limited to remote online assessments due to the COVID-19 restrictions. However, this also enabled quick data capture, allowing the sampling of psychological and health data at different stages of the pandemic. The online survey was suitable for completion across many devices, including desktop computers, tablets, and smartphones. The surveys were developed and tested by the Generation Scotland team at the University of Edinburgh, in collaboration with the Schools of Health and Wellbeing Improvement Research Network (SHINE) based at the University of Glasgow.
Given the sensitivity of some of the questions, as well as potential reservations about providing personal information in an online study, none of the questions in the surveys required an answer. Many sensitive questions also had a "prefer not to answer" option. If participants left a question unanswered, they were asked to confirm if they wanted to continue without answering. For data privacy reasons, after moving to the next page of questions, participants were not able to go back and amend their answers. As Qualtrics does not have password-protected accounts, his was to prevent other people in the same household from using the same device to view the participant's responses.
The questions included in all three TeenCovidLife surveys can be seen in the Extended data 10 . A copy of the Qualtrics survey (.qsf) for any survey can also be requested from the authors. Some questions only appeared in one survey. If a participant had taken part in previous surveys, not all questions were asked again as some items were judged as unlikely to have changed between surveys.

Measures
Several commonly used psychological measures were presented in all three surveys: • The Adolescent Sleep-Wake Scale (ASWS) 12 , a tenitem measure assessing sleep quality and disturbances in adolescents.
• Brief Resilience Scale (BRS) 13 , a short measure assessing trait resilience -the ability to "bounce back" from setbacks and distress.
• World Health Organisation Well-Being Index (WHO-5) 15 , a five-item measure assessing overall wellbeing.
Several subscales of the Social Emotional Health Survey (SEHS) 16 were also applied, assessing the level of social support from family members, friends, and school staff, as well as optimism and self-efficacy. See Table 1 for further details of measures used in all surveys, as well as the Extended data 10 for full questionnaires.
Sample TeenCovidLife Survey 1. Anyone who was aged 12 to 17 and resident in Scotland was able to take part in the study. As this was an online survey, internet access was required to participate. The questionnaire could be accessed using any device, including a tablet or smartphone. Data collection commenced on Friday 22 nd May 2020 and closed Sunday 5 th July 2020, during the first coronavirus-related school closure period in Scotland, which lasted from 23 rd March to 11 th August 2020. The recruitment period lasted a total of 44 days.
After participants began the survey, they had seven days to complete it. The final sample consisted of 5,543 young people age 12 to 17. Figure 1 shows the number of participants included in the final sample for Survey 1 by the day they began the survey.
TeenCovidLife survey 2. All participants in TeenCovidLife Survey 1 with a working email, who consented to re-contact were sent an email invite. This included any participants who may have turned 18 since the first questionnaire. As such, the potential age range for returning participants was 12 to 18. However, a separate Qualtrics survey was set up for any participants who had not taken part in TeenCovidLife Survey 1. These participants needed to be age 12 to 17 and living in Scotland. The 'New' and 'Repeat' versions of the surveys only differed in that some items (such as sex) were not asked again to previous participants. The full questionnaires can be seen in the Extended data 10 . As before, internet access was required to take part.
Data collection took place from Tuesday 18 th August 2020 to Saturday 10 th October 2020, closely following the initial re-opening of schools on the 11 th of August. The recruitment period lasted 54 days. After participants began the survey, they had 14 days to complete it.
The final sample consisted of 2,232 young people age 12 to 18. Of this sample, 761 had taken part in TeenCovidLife Survey 1. See Figure 2 for the numbers included in the final sample for Survey 2 by the day they began the survey, as well as when reminder emails were sent to previous participants.

TeenCovidLife Survey 3.
All participants who took part in either previous TeenCovidLife surveys and gave permission for re-contact along with a working email address were invited to take part. Some returning participants may have turned 18 or 19 since the first survey. Consequently, the returning sample ranged between ages 12 to 19.
As before, young people age 12 to 17 living in Scotland who had not taken part in a previous survey were also able to take part. As in Survey 2, two Qualtrics surveys were created for new and repeat participants. These only differed in that some items, such as sex, were not asked again to repeat participants. Both questionnaires for Survey 3 can be seen in the Extended data 10 .
Data collection began Tuesday 12 th May 2021 and ran until Sunday 27 th June 2021. Data collection took place when students were returning to school after another period of school closures. The end-date for data collection was chosen as this was when most schools in Scotland closed for the summer holidays. The recruitment period lasted a total of 46 days. After participants began the survey, they had 14 days to complete it.
The final sample consisted of 597 young people age 12 to 19 years old. Of the full Survey 3 sample, 316 had taken part in both previous surveys. Figure 3 shows the number of participants Notes.
-in Source indicates that question was formulated in-house.
* indicates that question was asked to previous participants again in TCL2 and TCL3. 1 Not all domains of the SEHSS were repeated. For participants who took part in a previous survey, they were only asked about Optimism and School Support at TCL2 and TCL3. Only mentioned subscales of the SEHSS was used; this is not the full questionnaire.
2 Not all domains of the Good Childhood Index were used in TCL2 and TCL3. 'Satisfaction about Home Life' was not included in TCL2+3. 3 Question about education was asked to all, question about exam only asked to participants who were expecting to sit an exam in 2020  by the date they began the survey, as well as when reminder emails were sent to previous participants.

Recruitment
Similar recruitment methods were used for all three surveys.
Generation Scotland Generation Scotland is a family health study of approximately 24,000 adults living in Scotland aged 18 to 99 years at recruitment from 2006 to 2011 2 . Participants who had children age 12 to 17 and for whom a working email address was known were sent an email prompting them to invite their children to take part in TeenCovidLife. Postal invitations were also sent to participants who had children in the appropriate age range, but for whom no email address was known.
SHINE network SHINE is a network of over 500 schools that aims to bring together schools, policymakers, and academic researchers to conduct schools-based health and well-being research, and to support health improvement planning and implementation. Of the 514 schools in the SHINE network, 138 were secondary schools. The SHINE network helped promote the study to member schools, particularly for Survey 1.
Before the launch of Survey 1, the SHINE network announced the TeenCovidLife survey as the headline item in its May 2020 newsletter to all existing SHINE school members. The benefits of participation were outlined to schools, including TeenCovidLife's incorporation of measures from the SHINE mental health survey, additional support from the SHINE team in promoting the study in school, and the offer of a school-level report. Additionally, one of the SHINE schools recorded a promotional video encouraging participation in the TeenCovidLife survey. This video was featured on the SHINE website and Twitter account.
Surveys 2 and 3 were likewise promoted to SHINE schools via the monthly newsletter and Twitter. However, school-level reports were not offered for these subsequent surveys.
General public In addition to these recruitment routes, all three TeenCovidLife surveys were open to anyone age 12 to 17 living in Scotland. Both mainstream media and social media were used to advertise the study to the general public and encourage participation, as well as University of Edinburgh outreach programmes. Paid social media campaigns were run on Twitter, Instagram, and Facebook, and the surveys were also promoted through public engagement talks hosted by members of Generation Scotland.

Previous participants
In Surveys 2 and 3, participants who had taken part in a previous survey and provided a valid email address were re-contacted and invited to take part. Re-invited participants were sent a personalised link that gave them access to the survey and linked new responses to those from previous surveys.
Procedure A link to the study was included in emails and postal study invitations. A link to the study was also shared on social media and the Generation Scotland website. On arriving at the TeenCovidLife landing page, participants first read the volunteer information sheet. Participants also answered two questions to check they had read and understood the information sheet. Participants could not proceed to the main consent form until they answered both of these questions correctly. Next, participants completed the online consent form. Participants also gave consent to future re-contact from Generation Scotland.
The consent form highlighted that they were not obliged to take part in future studies if they were re-contacted. Consent and information sheet text for each survey are available in the Extended data 10 .

Results
Full demographic details for each survey, as well as comparison to population estimates, can be seen in Table 2.

TeenCovidLife Survey 1
The data cleaning process is presented in Figure 4. A total of 10,263 participants accessed the survey during the recruitment stage. After data cleaning, 5,543 participants were included in the final sample. Respondents were retained as participants if they had a) completed and agreed to the consent form; b) progressed past the first page of the questionnaire, which contained only basic demographic information; and c) answered at least one of the questions. Two members of the research team conducted the data cleaning separately. Final records were compared, and any inconsistencies were investigated and resolved until both researchers had identified the same records for inclusion.
The time to complete the survey varied, as participants could save their data so far and complete the survey later. The median time taken to complete the survey was 21 minutes, with an interquartile range of 15 minutes.
The sample was predominantly female (63.2%; 3,505), and there were slightly more participants in the 12 -14 age group (55.5%; 3,074) than the 15 -17 age group (43.6%; 2,415). Figure 5 shows the number of participants by age and sex.       Figure 6 for exclusions at each stage of the data cleaning.
Over a third (34.2%; 768) of the final sample had taken part in TeenCovidLife Survey 2, with an overall follow-up rate of 13.9% from Survey 1. As some older participants had birthdays between the first and second surveys, TeenCovidLife Survey 2 also includes data from 18-year-old participants.
As in Survey 1, participants could save their responses and return to the study later, meaning the time taken to complete the survey was highly variable. The median time taken to complete the survey was 18 minutes, with an interquartile range of 15 minutes.
As in TeenCovidLife Survey 1, the sample was majority female (62.7%), and there were slightly more participants in the 12 -14 age group (55.2%) than the 15 -18 age group (43.8%). Figure 7 shows the sex ratio by age. final dataset. Data were cleaned in the same manner as in previous surveys. Figure 8 summarises the exclusions at each stage of the data cleaning.
The majority of participants (93.6%; 559) had taken part in at least one previous TeenCovidLife Survey. Figure 9 shows how many Survey 3 participants had taken part in previous TeenCovidLife surveys. Over half (52.9%; 316) had taken part in both Survey 1 and Survey 2., 30.2% (180) had taken part in only Survey 1, and 10.6% (63) had taken part in only Survey 2.
As in previous surveys, participants could save their responses and return to the study at a later date, meaning the time taken to complete the survey was highly variable. The median time taken to complete the survey was 11 minutes, with an interquartile range of seven minutes.
As in the previous surveys, the sample was majority female (72.7%). As some returning participants may have turned 19 since the first survey, the sample included participants up to age 19. The majority of participants were in the 15 -19 age group (75.2%). Figure 10 shows the number of participants in each age band by sex.
Over half of the participants (55.4%, 331) were from urban areas, with 13.2% (79) from rural areas. As in previous surveys, the majority of the sample was white (86.8%, 518), and almost half (45.7%, 273) were from schools with 10% or fewer pupils living in deprived areas.
Participants were from 146 different schools across 31 local authority areas in Scotland. As in Survey 1, the most frequent local authority area was the Scottish Borders, accounting for 16.1% (96) of the sample. Table 5 shows summary statistics for the commonly used psychological measures included in the study. As the majority of participants took part in previous surveys, only measures that had been used again in Survey 3 are included here.

Full participation subsample
A subsample of 316 participants took part in all three surveys, indicating a 5.7% complete follow-up rate from Survey 1. Table 6 shows the demographic details of this sample. The subsample was heavily skewed towards female participants, with only 21.2% being male. The majority of participants were white (94.0%; 297), and over half went to school in urban areas (59.8%; 189). The mean age at Survey 1 was 14.94 (SD = 1.48). At Survey 3, this was 15.92 (SD = 1.50).

Strengths and limitations Strengths
The core strengths of this dataset are that it is timely, rich, and longitudinal -few other cohorts have assessed the impact of the COVID-19 pandemic in such a large sample of adolescents. The COVID-19 pandemic has led to disruptions to long-term education, such as through school closures and cancelled Almost half of the participants (46.1%, 1,029) were from urban areas, with 28.4% (635) from rural areas. As in Survey 1, the majority of the sample was white (82.8%, 1,847), and almost half (46.2%, 1,032) were from schools with 10% or fewer pupils living in deprived areas.
Participants were from 166 different schools across Scotland over 31 local authority areas. School data was not available for 8.3% (186) participants. The most frequent local authority area was Falkirk, accounting for 24.5% (551) of the sample. This may relate to a SHINE school in the Falkirk area that showed a high response rate. Table 4 shows summary statistics for the commonly used psychological measures included in the study. Other summary statistics can be seen in the TeenCovidLife Survey 2 General Report 34 , as well as in the Exams Mini Report 35 . Both are available for free download on the Generation Scotland website.

TeenCovidLife Survey 3
A total of 641 participants accessed the survey during the recruitment stage. Of these, 597 participants completed the survey with a high enough rate of completion to be included in the

Limitations
The study was restricted to those with internet use, due to the need to adhere to COVID-19 mitigation measures. As such, those from rural communities or lower socioeconomic backgrounds with less stable internet access may be under-represented.
The sample was also self-selected, meaning more altruistic or conscientious young people may have been more likely to take part.
As can be seen in the demographics, the dataset is not representative of the general adolescent population, with female participants being over-represented. Additionally, over 80% of the participants in all three surveys were white. While 2011 census data indicates Scotland's population is 96.0% white 25 , making this relatively expected, the low number of ethnic minority participants limits the analyses that can be conducted on ethnicity.
There also seemed to be an over-representation of young people with caring responsibilities. Between 12 to 21% of TeenCovidLife participants cared for a member of their household, while 2011 Scottish census data suggests only 3% of young people age 4 to 24 identify as carers 37 . This may reflect a misunderstanding of the survey item, particularly if young people were taking more responsibility for younger siblings during the lockdowns, or some bias in the recruitment process. It is also possible that those with caring responsibilities were more likely to be interested in the project.
Furthermore, there were relatively low follow-up rates. Only 5.7% who participated in Survey 1 also took part in Survey 2 and 3. However, although at early stages participants were informed that they may be contacted for future surveys, this was not a defined goal of the study from the outset. Survey 1's recruitment was also considerably larger as schools were mostly closed, with young people's studies relatively disrupted or often suspended. As such, participants had more time to take part. Moreover, the pandemic was an even more salient topic at this early stage. By comparison, Survey 3 was conducted when schools were open and most adolescents were in a very busy school assessment and exam period, and the pandemic had been on-going for over a year, meaning it was relatively less salient.
The impact of the SHINE network's active promotion of TeenCovidLife Survey 1 may have also significantly contributed to the difference in uptake. The SHINE Network Manager, a former Deputy Headteacher, was able to advise schools accordingly to promote a whole school approach to data collection, appropriate during remote learning. While the poor retention remains a limitation, the sub-sample of participants involved at all three waves (n = 316) may still be useful for analysis.
Finally, most participants did not fully answer every single question. Due to ethical reasons, participants were permitted to skip questions that they were uncomfortable with or did not wish to answer. Consequently, there is missing data and exams. These disruptions may have long-term effects on health, well-being and success of young people, warranting study for years to come.
TeenCovidLife was designed in cooperation with the schoolsbased health behavioural research study SHINE. SHINE also forms part of the wider Health Behaviours in School-Aged Children study 11,36 . TeenCovidLife uses many of the same measures and questions as in both HBSC studies, as well as SHINE projects such as the SHINE networks pupil mental health and wellbeing survey. This harmonisation facilitates cross-cohort comparisons. Moreover, SHINE's expertise ensured TeenCovidLife asked questions relevant and meaningful to young people.
Finally, the surveys were implemented at key time pointsduring the first period of school closures in the UK, when schools were beginning to open again and lockdown measures were easing, and finally a year after the first lockdown, following the second national lockdown. This allows researchers to assess the impact of school closures, as well as the long-term effects of the pandemic on young people over time.

Ethical considerations
The TeenCovidLife study was reviewed and given a favourable opinion by the East of Scotland Research Ethics Committee (Reference: 20/ES/0021 AM03).

Conclusions
The data obtained through the TeenCovidLife project aimed to capture the impact of the COVID-19 pandemic on adolescents in Scotland. Three datasets were collected at three key time points for young people, assessing the emotional impact of both the pandemic and the national lockdowns on health, well-being, and education. A subsample of 316 participants took part in all three waves of data collection, allowing for analysis of change over time. This dataset is a valuable resource for researchers, and is available through the established data access procedure from Generation Scotland.

Data availability
Underlying data Non-identifiable data from the TeenCovidLife surveys are available to researchers in the UK and internationally through authorised access. Researchers who wish to use the TeenCovidLife data can apply for access using the standard Generation Scotland application process. More information about the process can be found on the Generation Scotland website (www.generationscotland.org).

Open Peer Review
sample obtained. The authors could include a column in the combined table with this information (e.g., % male in this age group) from population studies. This would be very helpful for users of the dataset to understand the representativeness of the sample (rather than some references to aspects of representativeness in the text).
Did the authors use any methods for dealing with missing data on individual items in the psychometric measures used in TeenCovidLife? For example, mean imputation? Please detail.

5.
Minor point: are the authors confident that the sample was based on living in Scotland? Was there a question pertaining to this? Given the questionnaires were online, it is possible that participants from outside Scotland completed the measures.

6.
helpful for users of the dataset to understand the representativeness of the sample (rather than some references to aspects of representativeness in the text). Response: Thank you for this useful feedback. We have collapsed all the tables for each survey timepoint into one table, see Table 2, and included population estimates where data is reasonably available.