Long COVID-the physical and mental health of children and non-hospitalised young people 3 months after SARS-CoV-2 infection ; a national matched cohort study ( The CLoCk ) Study


 Introduction: We describe post-COVID symptomatology in a national sample of 11-17-year-old children and young people (CYP) with PCR-confirmed SARS-CoV-2 infection compared to test-negative controls.Methods and analysis: A cohort study of test-positive (n=3,065) and age-, sex- and geographically-matched test-negative CYP (n=3,739) completed detailed questionnaires 3 months post-test.Results: At PCR-testing, 35.4% of test-positives and 8.3% of test-negatives had any symptoms whilst 30.6% and 6.2%, respectively, had 3+ symptoms. At 3 months post-testing, 66.5% of test-positives and 53.3% of test-negatives had any symptoms, whilst 30.3% and 16.2%, respectively, had 3+ symptoms. Latent class analysis identified two classes, characterised by “few” or “multiple” symptoms. This latter class was more frequent among test-positives, females, older CYP and those with worse pre-test physical and mental health.Discussion: Test-positive CYP had a similar symptom profile to test-negative CYP but with higher prevalence of single and, particularly, multiple symptoms at PCR-testing and 3 months later.


Introduction
Children and young people (CYP) are more likely to have a mild illness and less likely to be hospitalised following SARS-CoV-2 exposure compared to adults. More children 1 recover without sequelae compared to adults 2 . However, we know little about the diagnosis, prevalence, phenotype or duration of Long COVID (also called long haulers and post-acute COVID syndrome) in CYP. More than 200 different symptoms have been associated with Long COVID 3,4 , with an adult prevalence up to 80% 5 . Long COVID has been reported in those symptomatic or asymptomatic at time of SARS-CoV-2 testing, and in laboratorycon rmed or uncon rmed cases, with symptoms beginning at or after acute infection, which may be persistent, intermittent or relapsing in nature 3,5 . Adolescents may have a higher risk than younger children 1,6 but it is unclear whether the features associated with Long COVID are related to the viral infection or the effects of the pandemic, lockdown and school closures with consequent social isolation.
A literature review of Long Covid in CYP identi ed 17 relevant publications with sample size ≥10 (Supplementary Table 1). Existing studies were all observational, including cohort (n=9), matched cohort (n=2), and cross-sectional (n=6) designs. Clinical diagnosis of COVID-19 infection was con rmed by either a positive reverse transcription polymerase chain reaction (PCR) result or serological testing in 13 studies; the remaining studies included COVID-19 cases that were diagnosed clinically, were self-reported COVID-19, using rapid antigen tests or through an unspeci ed method of con rmation. These studies included 15,250 CYP up to 20 years with follow-up of 28 to 324 days. The most common persistent symptoms across studies were fatigue, insomnia, anosmia and headaches. The prevalence of Long COVID in CYP with laboratory-con rmed or suspected COVID-19 ranged from 1%-51%, with smaller studies reporting higher prevalence rates.
An online, nationally-representative survey of 313,216 people asking about self-reported Long COVID (unexplained symptoms persisting for more than four weeks after suspected COVID-19) estimated the UK population Long COVID prevalence to be 0.14% for 2-11 years, 0.50% for 12-16 years, and 1.51% for 17-24 years 6 .
The mental health of CYP have received less attention than the physical symptoms of COVID-19, despite the signi cant negative impact of the pandemic on their wellbeing 7 . Miller,et al. 8 found 10% of children with Long COVID reported unspeci ed persistent psychological/psychiatric symptoms. However, no signi cant differences between depression, anxiety, and perceived stress were found comparing CYP with Long COVID to COVID negative controls 9 . Similarly, a study of 1,560 CYP (median age 15 years) found a high rate of neurocognitive, pain and mood symptoms but no difference between seronegative and seropositive participants, highlighting the importance of including a negative control group in longitudinal studies of Long COVID 10 .
The CLoCk study is a national, longitudinal cohort study of CYP in England 11 with the primary aim of describing the clinical phenotype and prevalence of post-COVID physical symptoms and mental health problems among CYP with laboratory-con rmed SARS-CoV-2 infection compared to test-negative controls, to identify those most at risk and their illness trajectory. This paper presents the results of the study 3 months after PCR-testing in CYP across England. negative CYP were contacted. A letter was posted to all those selected, inviting them to take part in this study using an online link which provided them with details of the study, an option to consent online and complete a short recruitment questionnaire.

Design
We began contacting individuals from April 2021 onwards. In this paper, we focus on those who were tested in January-March 2021 because only they could report symptoms 3 months post-test with minimal recall bias of symptoms at time of testing. For this group, a total of 50,846 individuals (23,048 testpositives, 27,798 test-negatives) were invited to participate (Fig. 1).

Sample size calculations
The original study design was based on the calculation that 5,000 participants (2,500 test-positives, 2,500 test-negatives) would have 80% power to detect at least a 4% difference in symptom frequency at 5% signi cance, if test-negative participants had a 34% prevalence (based on available data at the time from the sKIDs study 12 ), accounting for attrition and possible lower baseline symptom prevalence. However, studying multiple symptoms and identifying risk factors for Long COVID requires a larger sample size.
For this reason, we amended our calculations to invite all available participants in England (except those tested in December 2020 due to funding constraints at present) 11 .

Data collection
Participants who were tested between January-March 2021 were contacted 3 months after testing.
Following online informed consent, the CYP self-completed an online (or paper) questionnaire about their physical and mental health at the time of the original test ("baseline") and at the time of completing the questionnaire; younger CYP and CYP with special educational needs or disability could request the help of their carer. The completed questionnaires were returned at a median time of 14.9 weeks after testing [25th ,75th centiles: 13.1, 18.9]. A total of 63 test-negative CYP reported having had a previous positive SARS-CoV-2 test and were excluded from analysis.

Measures
The rst questionnaire sent to CYP included demographic characteristics, elements of the International Severe Acute Respiratory and emerging Infection Consortium (ISARIC) Paediatric COVID-19 follow-up questionnaire 13 and the recent Mental Health of Children and Young people in England surveys (https://tinyurl.com/NHSWave1FU). The follow-up questionnaires were identical but did not include questions on demographic characteristics. They were designed together with ISARIC Paediatric Working Group to produce a harmonised data collection tool, to facilitate international comparisons regarding the risk factors and pro le of Long COVID in CYP.
The elements taken from the ISARIC Paediatric COVID-19 follow-up questionnaire 13 included questions about physical symptoms, particularly cough and fever (the main acute symptoms in non-hospitalised CYP) 14 and gastrointestinal symptoms which were commonly reported in seropositive CYP 15 . Other symptoms which might manifest later in Long COVID (e.g., tiredness, headaches, myalgia etc.) were also included.
We asked CYP to rate their general physical and mental health before their SARS-CoV-2 test, in two separate questions using a 5 category Likert scale; in analyses we recoded these variables into two categories (very poor/poor/ok versus good/very good). To measure mental health and wellbeing, the Strengths and Di culties Questionnaire (SDQ) 16 was summarised into the total di culties score that excluded the prosocial dimension, along with the short 7-item version of the Warwick Edinburgh Mental Wellbeing Scale (SWEMWBS) 17 . A higher SDQ total di culties score is indicative of more problems, whereas a higher SWEMWBS score indicates a higher level of mental well-being. Quality of life/functioning was measured via the EQ-5D-Y 18 and fatigue was measured by the 11-item Chalder Fatigue Questionnaire (CFQ) 19 .

Statistical methods
To assess the representativeness of our study participants we compared their demographic characteristics (sex, age, region of residence) to those of the target population. The participants' demographic characteristics, physical symptoms at "baseline", and physical symptoms, mental health status, well-being, quality of life/functioning, and fatigue 3-months post-test were compared by SARS-CoV-2 test status. We carried out comparisons separately by age-groups (11-15y vs. 16-17y) as the prevalence of Long COVID may vary by age (https://tinyurl.com/ONSPrevalence0721).
We used latent class analysis 20 to assess whether and how baseline and 3-month physical symptoms clustered among CYP, allowing for differential model parametrization by SARS-CoV-2 test status (while analysing the data jointly by test status but separately by time). The number of classes was selected by comparing the Bayesian Information Criteria. Predicted class membership was estimated and used to assign CYP to their most likely class; this classi cation was then used to describe the characteristics of the latent classes.
As this is mainly a descriptive study, we do not report p-values for comparisons by SARS-CoV-2 test status. We do report estimates of latent class prevalence by SARS-CoV-2 test status, as well as their ratio, with con dence intervals computed using the delta method 21 . To assess the impact of potential response bias, we reweighted all symptom frequencies according to the age, sex, region and SARS-CoV-2 test status of the responders.

Study representativeness
A total of 6,804 CYP who had been tested between January and March 2021 participated in the study by completing the 3-month questionnaire. The overall response rate was 13.4%, with a similar proportion of test-positives (13.3%) and test-negatives (13.5%) contributing (Table 1). More females and older CYP (16-17-year-olds) responded. Response rates also varied by region of England. Overall, there was little difference in demographic characteristics between test-positive and test-negative participants, re ecting the matched study design (Table 2).

Physical symptoms and pro le: baseline and 3-month post-test
At the time of testing, test-positive CYP had higher percentages of physical symptoms compared to testnegative CYP (Table 3); 35.4% of test-positives and 8.3% of test-negatives had any symptoms whilst 30.6% of test-positives and 6.2% of test-negatives had 3+ symptoms. The types of symptoms reported by test-positives and negatives were the same in the two age-groups: the most common symptoms among test-positives were sore throat, headache, tiredness and loss of smell while test-negatives had sore throat, headache, fever and persistent cough. The prevalence of these symptoms, however, varied by SARS-CoV-2 test result (e.g. 26.3% of positives compared to 4.8% of negatives reported headaches).
Three months after the SARS-CoV-2 test, the presence of physical symptoms was higher than at baseline in both groups; 66.5% of test-positives and 53.4% of test-negatives had any symptoms whilst 30.3% of test-positives and 16.2% of test-negatives had 3+ symptoms. The symptom pro le did not vary by age: for both 11-15y and 16-17y the most common symptoms among test-positives were tiredness, headache and shortness of breath and, among test-negatives, tiredness, headache and the unspeci ed category of "other". Again, the prevalence of tiredness and headache was consistently higher in the test positives, 39.0% and 23.2% versus 24.4% and 14.2% in negatives, respectively. Prevalence was higher for 16-17-year-olds; for example, 46.4% of test-positives reported being tired compared to 29.6% of test-negatives. When we reweighted the percentage of reported symptoms at baseline and at 3 months post-test, broadly similar patterns were observed to those reported above (Supplementary Table 2).
Mental health, well-being, quality of life/functioning and fatigue 3-month post-test There was no difference in the distribution of mental health scores (assessed by the SDQ total di culties scores) and well-being (assessed by SWEMBS) between test positives and negatives, overall or in either age-group. The SDQ median (25 th ,75 th centile) was 10 (6,15) for both test-positive and test-negative CYP aged 11-15y. For CYP aged 16-17y, the corresponding values were 11 (7,16) for test-positives and 12 No evidence of clustering of baseline symptoms was found for either test-positive or test-negative participants. There was, however, evidence of clustering in symptoms reported at 3 months, with two subgroups emerging for both test-positive and test-negative CYP ( Figure 2). In each, the largest subgroup (class 1) had very low prevalence of most symptoms, while the second subgroup (class 2) was characterised in both positives and negatives by multiple symptoms dominated by tiredness, headache, shortness of breath and dizziness. We refer to these classes as "few" and "multiple" symptoms classes. The estimated probability (risk) of being in the multiple symptom class (class 2) was 29.6% (95% con dence interval, 27.4%, 31.7%) for test-positives and 19.3% (17.7%, 21.0%) for test-negatives and the risk ratio of being in class 2 versus class 1 comparing test-positives to test-negatives was 1.53 (1.35, 1.70).
For both test-positive and test-negative CYP, those assigned to class 2 were more likely to be female, older, to have poorer baseline physical and mental health (relative to the overall percentages of 19% and 30%) and, at 3-months, to be more likely to have problems with mobility, self-care, usual activities and pain/discomfort. They also have higher SDQ total di culties and CFS scores, and lower SWEMBS scores (Table 4).

Discussion
Given that there is no universally agreed de nition of Long COVID, we elected not to start with an arbitrary de nition of a new condition but to seek the descriptions from almost 7,000 CYP of their physical and mental experiences in the months following a SARS-CoV-2 test. This is the largest study on symptoms post-COVID in children to date and, uniquely, uses child-reported symptoms, is con ned to PCR-proven SARS-CoV-2 status, has a PCR-negative control group, and uses standardised measures to assess mental health, well-being, and fatigue.
Several important ndings contribute to the current knowledge of Long COVID in CYP. First, three months after the SARS-CoV-2 test, the presence of physical symptoms was higher than at the time of testing. This nding emphasises the importance of having a comparison group to objectively interpret the ndings and derive prevalence estimates. Although 64.6% of test-positives reported no symptoms at time of testing (compared to 91.7% of test-negatives), they did not continue to remain asymptomatic, with only 33.5% of test-positives (and 46.7% of test-negatives) reporting no symptoms at 3 months. This nding warrants further exploration and could be due to self-selection into the study because they were experiencing ongoing symptoms, recall bias, external factors relating to the pandemic such as returning to school and exposure to other sources of infection, and the actual trajectory of the illness, although this wouldn't explain the high prevalence among test-negative CYP.
Second, symptoms reported at time of testing among test-positives were sore throat, headache, tiredness and loss of smell while test-negatives had sore throat, headache, fever and persistent cough. The symptom pro le does not distinguish test-negatives and test-positives. However, the two groups could be separated according to the number of symptoms at three months, when 30.3% of test-positives and 16.2% of test-negatives had 3 + symptoms, with tiredness and headache being common in both groups, but higher in the test-positives. Consideration of number of symptoms, rather than pro le, is particularly important given that 53.3% of the test-negatives had at least one symptom 3 months post-test. These gures should be seen in the context of published norms. For example, high levels of fatigue have been reported in the general adolescent population with an estimated incidence of 30% in CYP aged 11-15 over a 4-6-month pre-pandemic period 22 .
Third, our ndings showed that, for both test-positives and test-negatives, those assigned to the latent class with "multiple symptoms" at three months were more likely to be female, older and have poorer physical and mental health before COVID-19, suggesting that pre-existing physical and mental health di culties may in uence symptoms at three months. Unsurprisingly, those with multiple physical symptoms had poorer mental health, re ecting the close relationship between physical and mental health.
Fourth, whilst the prevalence of physical symptoms differed between test-positives and test-negatives, no differences were found in mental health, wellbeing and fatigue scores. Our ndings should be seen alongside the CYP literature. The commonest symptoms reported at three months in test-positives of tiredness, headache, shortness of breath, dizziness and anosmia are consistent with Molteni, et al. 1 where parents reported symptoms ≥ 28 days (fatigue 84%, headache 80% and anosmia 80%). They also reported that persistent symptom prevalence was higher in girls, teenagers and children with long-term conditions. The ONS survey 6 describes population prevalence rates of persisting 'symptoms more than 4 weeks after COVID-19' whereas the present study gives rates in a testpositive population; hence apparent differences may be attributable to differences in de nitions and methodology.
Taking the studies together, there is consistent evidence that some teenagers will have persisting symptoms after testing positive for SARS-CoV-2 and that mental and physical health symptoms are closely related. Avoiding false dichotomies between mind and body is likely to be helpful as, for example, stressed individuals may present with somatic symptoms or conversely persisting physical symptoms may be associated with depression and anxiety. Some individuals may develop somatic symptoms disorders 29 and the existing evidence for effective management of conditions such as pain, headache and fatigue 30 might be usefully evaluated in CYP presenting with persistent physical symptoms post-COVID. CYP with clinically impaired mental ill health should receive the appropriate evidence-based treatments whether or not they have physical symptoms. Family approaches and understanding of persistent symptoms is key 31 . Investigation of persisting symptoms may be needed or requested, with consideration of the potentially negative impact of protracted medical treatments or investigations if no abnormalities are found 29 .
This study has limitations. PCR-testing can result in some false negative and false positive results and we were unable to independently determine whether the test-negatives had previously had COVID unless they had been tested although this is likely to account for only a minority of cases. We could not recruit based on ethnicity as this was not recorded at time of test but ethnicity was very similar in test-positives and negatives (Table 2) and geographical region served as a proxy for socio-economic status; both these variables are thought to in uence COVID in adults and could be important in Long COVID 32 . As in any self-selected online study, we need to acknowledge our response rate of 13.5%. It is possible that there is a response bias for example, towards those continuing to experience symptoms at 3 months being more motivated to participate, resulting in an over-representation of symptom prevalence. It is also possible that recall bias in uenced the reporting of symptoms at the time of testing as well as physical and mental health prior to testing, in particular, if tested positive. However, we tried to minimize the impact of this bias by only considering CYP that reported on baseline ~ 3m later. We did not assess whether symptoms were continuous for the entire 3 months, or whether they waxed and waned. Finally, the experiences of the CYP in January, February and March were likely to be highly varied with regard to school closure. At the time of testing, schools were closed, while, at 3 months after testing, schools had reopened albeit with social distancing, repeated testing and restriction of activities. Schools can be a source of both stress and support, and the return to school may partly explain some of the ndings, in particular, the higher prevalence of symptoms at 3 months compared to baseline. The responders are largely representative of our target population though we have over-representation of girls and older CYP, with under-representation from North-West England and London. Inclusion of the comparator group was essential to place the ndings within the wider context of the pandemic.
These data also re ect symptomatology at a time when the Alpha variant was predominant in the UK. Whilst these ndings may change with different variants, the prospective nature of this study makes it uniquely placed to detect such changes across the pandemic waves.
In summary, post-COVID is different in CYP to adults and one should not extrapolate from the adult literature to decide policy and services for CYP. Our research demonstrates (1) the importance of having a control, test-negative group to interpret ndings and prevalence estimates, (2) that it is essential to consider multiple symptoms in any clinical phenotype of Long COVID, (3) that mental and physical health symptoms should both be considered, (4) PCR-proven SARS-CoV-2 positive CYP had a higher frequency of any symptoms and multiple symptoms three months post-test than test-negatives. These results would have been very di cult to interpret without a control group and laboratory-con rmed SARS-COV-2 status. Hence, prevalence estimates and de nitions of Long COVID should consider the presence of multiple symptoms, proven SARS-COV-2 status and include control groups. We will use these data and a Delphi consensus process to formulate a research de nition of long COVID in CYP and will follow up the CYP over time to understand the long-term course in SARS-CoV-2 positive CYP. This paper does not provide an evidence base for management of Long COVID or for any infective/immunological mechanisms underlying it. More research is needed. Elizabeth Whittaker, e.whittaker@imperial.ac.uk contributed to the work on infection and reviewed the manuscript. Shamez Ladhani shamez.ladhani@phe.gov.uk developed the study methodology, operationalised the regulatory and recruitment ideas for the study and revised the manuscript.
Olivia Swann and Elizabeth Whittaker designed the elements of the ISARIC Paediatric COVID-19 follow-up questionnaire which were incorporated into the online questionnaire used in this study to which all the CLoCk Consortium members contributed.
All members of the CLoCk Consortium made contributions to the conception or design of the work; were involved in drafting both the funding application and this manuscript; approved the version to be published; and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.
Additional Co-Applicants on the grant application and CLoCk Consortium members (alphabetical): Trudie Chalder is a member of the NICE committee for Long COVID. She has written self-help books on chronic fatigue and has conducted workshops on chronic fatigue and post infectious syndromes.
Data availability statement: All requests for data will be reviewed by the Children & young people with Long Covid (CLoCk) study team, to verify whether the request is subject to any intellectual property or con dentiality obligations. Requests for access to the participant-level data from this study can be submitted via email to Clock@phe.gov.uk with detailed proposals for approval. A signed data access agreement with the CLoCK team is required before accessing shared data. Code is not made available as we have not used custom code or algorithms central to our conclusions.    (10,14) 16 (12,20) 11 (11,14) 17 (13,21) Mean (SD) 11.6 (4.5) 16.4 (6.0) 11.8 (4.2) 17.0 (5.7) *Some/a lot of problems with mobility (e.g. walking about), self-care (e.g. washing/dressing) or doing usual activities (e.g. going to school); some/a lot of pain/discomfort or a bit/very worried, sad or unhappy.