Royal Society DELVE Initiative
Balancing the Risks of Pupils Returning to Schools
Jul 24, 2020
This paper has drawn on evidence available up to 10 July 2020. Further evidence on this topic is constantly published and DELVE will continue to develop this report as it is prepared for publication. This independent overview of the science has been provided in good faith by subject experts. DELVE and the Royal Society accept no legal liability for decisions made based on this evidence.
As we move into a phase of continuously reviewing decisions to keep schools open or shut, we need to recognise there are risks from having schools open and risks from having schools shut. In open schools, the risks to pupils themselves from COVID-19 are very low, though there are risks to school staff, parents/carers and the wider community. Closing
schools causes loss of learning and deterioration in children’s mental and physical health, for example. Keeping schools shut increases inequalities, in both children’s education achievement and their long-term prospects. The goal of “levelling up” therefore needs schools to be open. Keeping schools open is also the key to unlocking the rest of the economy, allowing parents to leave teaching to teachers and return to their own jobs. The evidence on the infection risk from school opening is limited, though to date it suggests that the risk from opening schools, relative to restarting many other activities, is not as high. The experience of most other countries which have already taken this step supports this. By contrast, the evidence on the negative impact of closing schools is considerable and robust.
The report sets out this case in detail. The aim is to provide an evidence framework for understanding the risks involved in re-opening and re-closing schools.
The key issues are the effect of schools closing and re-opening on:
risks from infection
risks from loss of skills and increases in inequality
risks to child and parent mental health
risks from parents not being able to return to work.
The report addresses each these in turn, describing first the current state of knowledge, and second, the steps that we can take to collect more data specific to the UK to inform future decisions.
Cite as: DELVE Initiative (2020), Balancing the Risks of Pupils Returning to Schools. DELVE Report No. 4. Published 24 July 2020. Available from http://rs-delve.github.io/reports/2020/07/24/balancing-the-risk-of-pupils-returning-to-schools.html .
Table of Contents
Footnotes and References
A. Keeping schools open should be the default policy. The Government should do everything feasible in order to not close schools. This means:
Suppressing the virus in the wider community to reduce the risk of transmission in schools once at full capacity, and to minimise future disruptions to learning; if local outbreaks occur, other facilities where the risk of transmission is high (such as pubs or gyms) and non-essential shops should be closed, before considering school closures;
Providing realistic guidance and substantial extra resources to ensure schools can minimise chains of transmission. This includes: parental guidance, translated into multiple languages, on when to keep their child at home applying the precautionary principle; rigorous hygiene rules; distancing and reduced mixing; extra teachers, PPE (one or two full PPE sets and provision of cloth face coverings for teachers, older children and those with underlying heath issues); management of staff rooms, regular testing and prioritisation for vaccines for teachers;
Implementing an effective monitoring regime that can cope with the likely case load in winter, including broad surveillance, linked to an effective, sufficiently scaled and rapid test-trace-isolate system, as well as systematic outbreak investigation. This is all essential to ensure that schools are re-opened as quickly as possible.
B. If local full or partial closures must occur, this should be based on clear scenarios with objective criteria:
There must be a set of clear, comprehensible, data-driven and public criteria defining each escalation; these could be presented in the form of a series of pre-defined local alert levels, and will depend on key parameters, such as the number of local cases;
There must be associated plans for each escalation. These need to be defined but might look like this:
Sporadic cases or clusters in the local community: provide alerts to local schools and families, test, trace and isolate, and where appropriate group isolation.
Sporadic cases in school: provide alerts, test, trace and isolate and group isolation (e.g. class/year);
Clear clusters in school across year groups: provide alerts, consider individual school closure with outbreak control response to minimise onward transmission and re-open school as quickly as possible;
Widespread local community transmission: local area lockdown;
Last resort: national scale closures.
There must be clear responsibility for decision making regarding the closing and re-opening of schools. The key agencies are numerous, including the schools themselves, Local Authorities and embedded public health teams with responsibility for infection prevention and control, Public Health England (PHE) and local Health Protection Teams, the Joint Biosecurity Centre and the Department for Education (DfE). The scope for slow and muddled decision-making is obvious. Before September, these agencies need to determine a coherent chain of command and responsibility to take effective local decisions. While PHE and DfE have set out basic, and separate, criteria for school closure, we are clear that a much more joined up and granular approach is necessary.
C. The Government should urgently initiate data collection:
A system, including surveillance studies, must be in place to provide decision-makers with the local and timely data they need to monitor neighbourhood and school infection rates and to respond accordingly. That is, the information structure we set out in A.3 must be integrated with the decision structure in B.3. For example, the minimal data collection from individuals being tested for Covid-19 needs to include whether the person being tested is connected with a school (as teacher or parent), and if so, which school. This information needs to be collated and made available to the relevant decision makers charged with monitoring the conditions for school closures;
There needs to be a programme of anonymous assessment of education achievement and pupil mental health across all age ranges in a sample of schools in mid-September, to gauge the extent and nature of the learning loss and impacts on student mental health. Tests should also be administered at the end of the 2020/21 academic year in this sample of schools, to assess any improvement during the year;
Given the weak evidence base on both the effectiveness of specific strategies in schools to reduce infection risk, as well as the adaptions that will be needed to mitigate learning and other losses from school closures, we need to encourage experimentation, better evaluation and good knowledge exchange mechanisms to share learning across the education system.
We further recommend a cost-benefit analysis of widening the eligibility for influenza vaccination to secondary-school children (or certain age-groups thereof).
D. The Government must establish effective, clear and unified communication with school leaders, teachers and parents to manage opening and closing of schools in response to local conditions. The communication from government to parents should be informed by behavioural science to make sure the messages are salient, clear and consistent.
School closure affects many outcomes, and these are quantified in different metrics. Weighing these against each other to reach an overall decision is crucial. As in other aspects of the COVID-19 response, this can be expressed as “lives versus livelihoods”. This is true for schools also: lives saved by closing schools, set against the lost skills, lower future earnings, and consequent greater poverty. We cannot resolve that trade-off here but given the need to design policies that do the least harm, it is essential that we balance any increased risk to life from schools being open with lives shortened and diminished from greater poverty due to school closures. Our assessment of the evidence suggests that keeping schools open should be the default position given the substantial risks from closures.
While there are many factors to consider when weighing the risks from closing and opening schools, the core issue is the interplay between two dynamic processes: infection and learning. Closing or re-opening schools changes the evolution of both. For infections, school closures may reduce the effective reproduction number. When the infection rate is high, this might potentially result in a large reduction in the number of cases, but when infection rates are low, the reduction in the number of cases may not justify the costs of school closure. The learning process is characterised by dynamic complementarity, meaning that “learning builds on learning”. Missing some school reduces the child’s ability to learn the following material. This means that the gap of lost learning grows at an increasing rate as more and more time is missed from shut schools. The relative benefits and costs of opening or closing schools thus depends upon how long schools have been shut, and the local infection rate.
There are no perfect data to address any of these questions. Much of the discussion below is based on imperfect and incomplete evidence, though it should be noted that the evidence on the learning and other losses from school closures is more robust than that for the impact of schools on infection rates. This means that some of our recommendations are cautious and cannot be expressed with as high a level of specificity as we would normally prefer. Some of our key recommendations are therefore specific ways in which more data can be collected to better inform future re-opening and re-closing decisions.
Schools and infection
1. What happened to infections in other comparable countries when they re-opened schools?
We are able to learn from the experiences of other countries, comparable to the UK, and track the impact on infection rates of opening schools. Comparable international data are shown in Figure 1, before and after school opening. This suggests that, in general, the opening up of schools has not resulted in notable rises in infection at a national level. In all the countries reviewed there have been individual school closures associated with COVID-19 infections in schools, except for in New Zealand.
We note a number of caveats in interpreting the data. First, in most countries other interventions were made at the same time, so that any trends we see may result from a combination of different actions. Second, in some countries the school opening date was recent and it may be too soon to see any effect in the data. Third, in most countries, schools partially opened and a complete return of all pupils may look different.
2. What protective measures did other comparable countries take in schools, and did they help?
As governments around the world introduced phased opening of schools, several risk mitigation interventions (e.g. enhanced hygiene measures, limited interactions, distancing measures, targeted communication strategies and the use of personal protective equipment) were employed to reduce transmission of the virus among pupils and staff.
The type and severity of interventions used vary greatly between countries. Additionally, it is impossible to assess the impact of individual mitigation measures because a combination of procedures was introduced in all settings, and school-level data on interventions and outcomes is not available.
Most countries employed physical distancing measures in schools until cases in the community had been suppressed to manageable numbers. In a few countries, at the time of this writing, distancing measures have been lifted to allow schools to operate at close to normal capacity, e.g. in New Zealand, Australia, France, Switzerland and it is critical that these countries be closely monitored. Furthermore, evidence on transmission of the virus in school settings and the link between school attendance rates and the number of new cases of COVID-19 in the UK should be examined. All these data sets will give us better insight into infection risk in the school setting.
Additionally, enhanced hygiene measures continue to be implemented in most countries, although in countries like New Zealand, this has been greatly relaxed due to successful suppression of the virus. It is recommended that explicit guidance and additional resources are provided to local authorities and schools to ensure that the highest level of hygiene is in place in schools until the winter season has passed, when it can then be reassessed. Consistency should enable these practices to become normative.
There is evidence from other settings on the use of masks. If physical distancing in schools is not possible, masks will play a role in preventing infection. Given the potentially lower infection risk from young children, the difficulty of getting young children to wear masks, and the likely negative impact of masks on their verbal comprehension and language development, the case for primary school children wearing masks is not clear. However, older children, all adults in schools, as well as parents bringing their children into schools, should wear masks. Masks should also be worn on public transport on the way to school in line with national guidance.
Given the limited evidence on the effectiveness of specific mitigations, schools need to try to minimise infection risk using a variety of methods and need to be resourced accordingly.
3. What is the evidence on the level of risk to children, the prevalence of infection among children, and estimates of the level of transmission in schools between children / from children to adults?
The risks to pupils themselves from Covid-19 infection are very low, with life-threatening complications for children being very rare indeed. However, there are risks to school staff, parents/carers and the wider community. The risks to teachers, parents and carers will vary depending on age and underlying health conditions, with some (e.g. grandparent carers) at higher risk than others. Those at higher risk will need to follow the national guidance in this respect.
Whilst early reports suggested that children were much less susceptible to infection than adults, the largest-scale sero-surveillance studies in Europe have so far found only somewhat lower infection-rates in children, versus adults: viz, infection-rates for children of between 60% and 80% of those in adults (and smaller differences were found in countries such as Sweden, where schools for under-16s remained open). The Office for National Statistics (ONS) Infection Survey pilot did not find significant differences in infection-rates in England, between different age-groups, but its current estimates on Nucleic Aciden Amplification Test (NAAT)-positivity-rates in the age-groups 2-9 and 10-19 are based on just eighteen NAAT-positive cases. The evidence-base on the likelihood of adults infecting children is limited, but between them, the highest-quality studies suggest that adults may be approximately half as likely to transmit infection to children in the same household, as they are to transmit infection to adults in the same household. There is also weak evidence that children are less likely than adults to transmit infection to others.
So far, there are very few data-sets available on school-specific transmission, but the balance of evidence so far suggests that schools play only a limited role in overall transmission. This is in marked contrast with influenza, where there is strong evidence that schools are an important driver of overall nationwide transmission. The international data that are available shows at most very limited transmission in schools where a child was the index case.
There is an important need for targeted data-collection, to determine more accurately the frequency of transmission from children to adults and from children to other children (both within schools, and outside), and to determine more accurately the frequency of transmission from staff to pupils in schools. We recommend that any outbreak in a UK school should be followed up by a thorough outbreak investigation, utilising contact-tracing and NAAT-testing of schoolchildren, school staff and household contacts of both. Phylogenetic analysis should be used where appropriate, to investigate probable chains of transmission.
4. Are there significant additional risks for BAME families?
The risk to BAME children, as for other children, is very low indeed. BAME adults are at higher risk from COVID-19 than other ethnic groups. The extent of the additional risk for BAME groups varies by age and pre-existing health conditions. The reasons for the additional risk are fivefold: they are more likely to work in occupations with higher exposure, to have worse living conditions, to suffer from co-morbidities, to access poorer health care and to face discrimination.
Hence some BAME parents, carers and school staff will indeed be at greater risk due to some of these causal factors and need to follow national guidance in respect of their particular risk factors. BAME teachers who do not have these underlying risk factors are likely to be less at risk than the average BAME adult. This is because they do not have a major risk factor faced by many BAME people, which is being in a low-income job.
5. How can we best improve our understanding of schools and infection for the UK?
The key infection question to be answered is whether re-opening schools raises the local infection rate. The international evidence (see 1 above) is cautiously optimistic, but we need data precisely tailored to the UK situation, including data on the prevalence of infection in schools, and on transmission within schools.
To answer this question, we need to assess the outcome from opening schools relative to a counterfactual: what would have happened to local infection rates if schools had remained closed? To do this requires testing of pupils not returning to school, as well as those who do. June and July 2020 offered that opportunity, and we hope that similar circumstances deriving from administrative/institutional decisions will do so also in September.
The COVID-19 Surveillance in KIDs (sKID) programme run by PHE, is a proof-of-concept study that aims to test staff and students in 100 schools this summer mini-half term. An analysis of the sKID data collection protocol suggests that it will not be able to conclusively demonstrate a difference in risk for attending school, even if the risk is double that of non-attendance. This is because the sample size is too low given the current COVID-19 prevalence in the UK, and because sKID plans to test only children attending school. A case-control study of children both in and not in school would allow us to compare how attending school affects the infection rate in children whilst taking into account variation in school facilitates, rates of contact between children outside of school, and regional infection rates.
Surveillance studies such as sKID do require huge sample sizes to be effective when infection rates are low, and so responsive testing and detailed case study analysis of test results (including of phylogenetic data) from schools which have experienced an infection, may be a more feasible approach to understanding transmission in schools.
Schools and learning loss
6. How much learning has been lost, and what is the impact on skill levels from closed schools?
For most pupils in the UK, around 12 weeks of face-to-face learning has been lost, about a third of a year. This is likely to have a very significant impact, greater for younger children, given evidence that investments in children’s learning tend to be complementary over time.
The key methodological challenge is that estimating a causal impact of school time on skills requires a setting with exogenous variation in the former. Consequently, the most robust evidence covers the effects of school closures because of teacher strikes, weather extremes, and other quasi-random settings.
The most robust studies suggest an impact of between 6%SD to 10%SD, for the learning loss. This is roughly equivalent to the difference between being taught by a highly effective teacher for half the year and an ineffective teacher for half the year, compared to being taught by an ineffective teacher all year.
Some of this will have been offset by learning at home and the provision of remote schooling, but this has only been partially effective for most families. Wide variation in the quantity and quality of remote schooling and home learning support between pupils and schools underlies much of the variation in learning loss over this period.
7. Why and how much does learning loss matter?
The evidence shows that schooling raises skills, and so while qualifications are in part simply a “signal” of ability, missed school means lower skills which have real implications for individual lives and for the economy as a whole.
First, a huge base of evidence shows that earnings depend on skills, and lower skills means lower earnings. For example, if a student misses a third of a school year, and each school year brings roughly a 10% return, earnings potential is likely to be permanently lower by around 3% a year. Students at the lower end of the ability distribution are more likely to have lost heavily from school closure and are more likely to be low earners, so will face a significantly higher risk of poverty. Evidence from the prevalence of teacher strikes in Argentina shows that exposure to average strike incidence in primary school reduces earnings in mid-life for women by 1.9% and for men by 3.2%.
At the national level, lower aggregate skills will reduce the growth rate. The magnitudes are not trivial: 13 cohorts of students have been affected by the lockdown, so from the mid-2030s for the 50 years following that, around a quarter of the entire workforce will have lower skills, with a consequently lower growth rate. The present value of such a fall in the growth rate is measured in billions not millions. For example, the study of teacher strikes in Argentina estimates the aggregate earnings loss in Argentina from lower skills of $2.34 billion per year.
One of the consequences of lost skills from closed schools is greater risk of poverty: earlier research from the ONS shows that “those with a low level of educational attainment are almost five times as likely to be in poverty now as those with a high level of education”. There is a huge literature linking poverty to ill health and early death, though isolating a causal relationship is always difficult. One example shows a 7.9 year gap in life expectancy for women and a 9.7 year gap for men between the most and least deprived deciles of areas poverty.
8. What is the likely impact on inequality?
The actual loss of learning in the current pandemic will vary by context, depending in part on what schools and families have been able to provide in the way of remote schooling. It is very likely that these gaps in provision have exacerbated existing inequalities between students from low and high socio-economic status families.
Our analysis of new achievement data from an online learning platform in the UK shows significant increases in the test score gap between high and low performers pre- and post-school closure.
The Education Endowment Foundation presents a meta-analysis of the existing literature. They calculated that school closures will widen the attainment gap between poor and non-poor families. Their meta-analysis suggests a median impact of 36% (the range of this estimate is between 11% and 75%) by September 2020.
Differences in parent/carer knowledge and ability to help their children, differences in parent/carer time available, differences in availability of quiet study space in the home, differences in the availability of IT kit and fast internet, and the differences in the provision of useful material by the school are all correlated with family income and contribute to growing inequality. This is supported by a number of high-quality studies undertaken since the beginning of remote schooling.
From a school perspective, a survey of teachers and Headteachers by NFER reinforced this view. Most teachers were covering less of the curriculum than normal, and pupils were doing much less work than normal. Teachers reported that pupils from poor families were less engaged with their work. Three-quarters of senior school leaders reported that their schools were offering ‘social or welfare’ support to vulnerable pupils, and half reported significant concerns for the safety and wellbeing of vulnerable pupils.
9. Specifically, what has happened to BAME children?
Higher levels of educational attainment are correlated with higher family income and higher parental education. Since BAME pupils on average come from households with lower levels of income, this in turn impacts on their education level. According to the ONS, children in Bangladeshi and Pakistani households were most likely to live in low income and deprived areas out of all ethnic groups. The relationship between deprivation and education is the key to understanding life chances. Children of these ethnic groups with lower levels of income are more likely to have been significantly badly affected by school closures.
GCSE scores shows wide variation within the BAME category. Students of Asian ethnicity, including but not only Indian and Chinese students, as well as Black African students, score above White British students, while Black Caribbean students are about level with White British ones.
There is also evidence to suggest a more positive attitude to learning and to school in general among some BAME students, particularly those who are the children of immigrants. This higher level of existing achievement plus more positive attitudes may to a degree help to offset the effect of higher poverty levels on the impact of school closures.
10. How can we best improve our understanding of learning loss in the UK?
We do not currently have direct measures of children’s learning loss as a result of the school closures. This is clearly a major data gap. The first step is therefore to collect the necessary test score information needed to estimate the scale of the learning loss from school lockdown.
Ideal study designs to properly gauge the learning loss from lockdown are not available. Almost all pupils were undertaking remote schooling, so there is no natural control group; and a simple longitudinal approach is not possible because very few schools are testing pupils as this report is written (early July 2020).
The first date of general testing will be the return to school in September. Because there is no national, officially sanctioned test for all age groups at that date in the UK, we recommend an additional test at the beginning of the academic year and one at the end (the latter to assess improvements during the year). The burden on schools now and more so in September will be large, so we also recommend that these tests be carried out on an anonymous sample of schools, rather than be universal.
For these tests to have any value, there has to be a comparator with the previous cohort, a “before” benchmark. Consequently, we recommend using tests that have a ‘synthetic’ before score – that is, normed tests that have a known and validated typical score in normal times.
The age-appropriate tests would be taken by pupils of all age groups in school in September 2020 in samples of schools; ideally, we would pick different schools per age group to sample a wider range of school effects, with some useful regional spread. The follow up test would be administered at the end of the academic year 2020/21. The exact sample size required depends on the test used but in broad terms it is estimated that around 150 schools per secondary school age group and around 600 schools per primary school age group would be required.
Schools and health
11. What are the other losses for the children from closed schools?
Evidence from the impact of school holidays suggests that children may be less active and have worse diets as a result of being out of school.
Previous research also suggests that being out of school and more isolated will impact negatively on children’s mental health. The impact is likely to be greater for children and young people with pre-existing mental health conditions and in more socio-economically deprived households. Social isolation and lack of contact with peers is likely to be particularly harmful for adolescents.
Evidence from the pandemic on the impact on children’s mental health is still relatively limited and of variable quality. However, it does suggest adolescents are particularly negatively affected, as are vulnerable children, those with particular conditions (e.g. autism) and children in care. This will have been exacerbated by more limited access to public services, particularly mental health services.
Establishing the causal impact of shutting and closing schools on children and young people’s health is challenging methodologically since lockdown has involved a range of other social restrictions. Proposals made in this report to measure the impact of school closures on learning loss (see 10 above) should be extended to measure the impact on child and adolescent mental health and wellbeing using the same methodology.
Schools and parental employment and earnings
12. What are the other losses for parents from closed schools?
In terms of wider impacts, it is clear that children being schooled at home has had a negative impact on parents’ ability to work, particularly women. This impact is well documented, though the long-term impact of these changes is unclear. What is clear is that the ending of the furlough scheme will significantly sharpen the need for many families to return their children to school so that they may return to their jobs. If schools return in September and no further lockdowns occur, it is conceivable that the impact of school closures on parents’ earnings and career trajectories will be limited (though clearly there will be an impact from any downturn in the economy). If, by contrast, schools do not return full time or we have repeated lockdowns, the impact on parents’, and particularly women’s, labour market trajectories is likely to be major.
This in turn affects household income and the greater burdens falling on low-income families will widen income inequality. It also adversely affects gender equality in the workplace and in the home. For low-income families, the loss of income and the degradation of future prospects may push them in to poverty (noting that one in five children already live in poverty). Currently around one in five parents in low income occupations have children and cannot work at home. This group will be particularly at risk of job losses if schools do not stay open.
13. How can we best improve our understanding of these other losses?
Open access data sets are coming on stream that can help us monitor the impact of the lockdown on children and young people’s physical and mental health, and indeed the employment circumstances of parents. Longitudinal follow ups are vital if we are to accumulate sufficient evidence on the impact of lockdown on children and young people’s health, as well as parental employment and household economic circumstances.
Commentary on the Government’s school opening plan
We note the recent (2/7/20) Department for Education (DfE) guidance regarding school opening in September 2020 1 . Its recognition of the significant costs of school closures, as well as the need to minimise the risks from opening schools, is to be welcomed. Our report supports the DfE position that all children need to return to face to face schooling wherever possible and that keeping schools open should be prioritised. School attendance must indeed be compulsory for the majority of children (i.e. those who are not at particular risk from COVID-19) and decisions to open and shut schools must be made on objective criteria with closures minimised. As we state above, what is urgently needed are clear and well communicated criteria to guide school closures and re-openings. This will also help to address teacher, parent and carer concerns about pupils returning to school.
The DfE guidance provides a lot of specific information for schools on how they should open and operate, as well as what will be required in preparation for a potential return to remote teaching during any future school closures. No doubt this guidance will change as the situation changes and more evidence emerges as to the effectiveness of different strategies to minimise infection transmission. Clear, realistic and timely communication with school leaders and teachers is vital. It is a major challenge to change ways of working in schools, particularly in a fast-moving environment with rapidly changing guidance. This needs to be recognised and guidance kept as simple and stable as possible. Further, some of the changes that are required of schools will need additional resource (whether that be in terms of additional staffing or pieces of PPE and other kit). This too needs to be recognised and an adequate level of resource provided. The challenge of simultaneously adapting schools to the safety mitigation procedures required, alongside also ensuring that adequate provision will be available in the event of subsequent school closures, should not be underestimated. The DfE guidance also proposes additional resource to support academic catch up for pupils who have fallen behind. This will be vitally important if we are to mitigate some of the increasing inequality in academic achievement that we are likely to see as a result of the pandemic. The DfE guidance that schools need to support children’s wellbeing when they return is also supported by our report. However, again additional resources will be required. We also recommend that the Government develops policies and provides additional resource to support those students who have experienced learning and other losses from school closures to ensure they are not disadvantaged over their lifetime. We need to mitigate the economic scarring that will result from educational deficits in particular. The Government also needs to consider long term options to support re-entry to educational opportunity later in life for those disadvantaged by COVID-19 now.
There is much we still do not know. Further surveillance and studies are needed to:
Determine the extent to which children of different ages transmit COVID-19, including undertaking school case studies to better understand transmission processes;
Understand the effectiveness of different interventions and strategies designed to reduce infection transmission in schools, including learning from strategies being adopted in other countries;
Measure the learning and other losses from school closures during the pandemic; identify successful school interventions to reverse the detrimental impact of school closures on learning loss and other outcomes;
To address (2) and (3) we need to encourage experimentation, better evaluation, and good knowledge exchange mechanisms to share learning across the education system.
As we move into a phase of continuously reviewing decisions to keep schools open or shut, we need to recognise there are risks from having schools open and risks from having schools shut. In open schools, the risks to pupils themselves from Covid-19 are very low, though there are risks to school staff, parents and the wider community. Closing schools causes loss of learning and deterioration in children’s mental and physical health, for example 2 . Keeping schools shut increases inequalities, in both children’s education achievement and their long-term prospects. The goal of “levelling up” therefore needs schools to be open. Keeping schools open is also the key to unlocking the rest of the economy, allowing parents to leave teaching to teachers and return to their own jobs. The evidence on the infection risk from school opening is limited, though to date it suggests that the risk from opening schools, relative to restarting many other activities, is not as high. The experience of most other countries which have already taken this step supports this. By contrast, the evidence on the negative impact of closing schools is considerable and robust.
The report sets out this case in detail. The aim is to provide an evidence framework for understanding the risks involved in re-opening and re-closing schools. The key issues are the effect of schools closing and re-opening on:
risks from infection
risks from loss of skills and increases in inequality
risks to child and parent mental health
risks from parents not being able to return to work.
The report addresses each of these in turn, describing first the current state of knowledge, and second, the steps that we can take to collect more data specific to the UK. The aims of this report are therefore two-fold. The first is to summarise existing evidence, including pre COVID-19 evidence, that can help us understand the likely risks of school opening/closing on infection rates, children’s learning and other outcomes being affected by the closure of schools. In the absence of better data, this can help inform policy decisions. As noted above the quality of the evidence available does vary enormously, depending on the specific issue being considered. We have not been able to undertake systematic reviews of the existing literature and we do our best to guide the reader as to the nature of the evidence and its robustness. The second aim is to document the data that we do have on the impact of this crisis in relation to schools and to articulate the kinds of data collection that are needed to plug gaps in our understanding. There are of course a number of important issues related to education that we have not been able to consider in this report. For instance, we need to ensure that those left behind now in terms of their education achievement are not disadvantaged forever, and to mitigate the sizeable economic scarring that will arise from educational deficits. The Government has already announced some strategies in this respect, more will be needed. In the longer term it will also need to consider options to support re-entry to educational opportunity later in life for those disadvantaged by COVID-19 now. These important long-term term issues merit further consideration but are not discussed here.
In the UK schools moved to remote teaching and learning in March, except for key worker children who were permitted to attend in person. Remote schooling provision has been highly variable in terms of quality, with some students struggling to access support. Further, remote schooling cannot easily substitute for the social contact with other adults and peers that children also get from attending school in person. Despite the best endeavours of schools, the consequences of children having minimal or poor quality schooling for such an extended period of time is likely to be that they miss out on a substantial amount of learning, have poorer health (particularly mental health) and for some children it will potentially put them at greater risk of neglect or abuse. Additionally, the toll on some parents from trying to home school their children is likely to be substantial, both in terms of parents’ wellbeing and their ability to do their paid work. If children receive remote schooling on this basis for an even more extended period of time, these negative impacts will be greater and longer lasting.
In England, some year groups have already returned (or will be returning) to school during the 2020 summer term. The Government has also stressed that schools will be open for all children in September 2020, though policies may vary across the countries of the U.K. However, it is likely that ongoing decisions will need to be made about closing and reopening schools during the academic year, depending on the prevalence of the virus. Decisions about schools opening, and indeed closing again in the event of infection spikes or a second wave, must be firmly guided by the relative risks of different courses of action. First, the risk to public health in terms of transmissions and the need to maintain an effective reproduction number (Re) below 1. Equally such decisions must also involve consideration of the risks of closure on learning loss and children’s future life chances, as well as risks for parents’ jobs and earnings.
While there are many factors to consider, the core issue is the interplay between two dynamic processes: infection and learning. Closing or re-opening schools changes the evolution of both. For infections, school closures may reduce the effective reproduction number. When the infection rate is high, this might potentially result in a large reduction in the number of cases, but when infection rates are low, the reduction in the number of cases may not justify the risks and costs of school closure. The learning process is characterised by dynamic complementarity, meaning that “learning builds on learning”. Missing some school reduces the child’s ability to learn the following material. This means that the gap of lost learning grows at an increasing rate as more and more time is missed from shut schools. The relative benefits and costs of opening or closing schools thus depends upon how long schools have been shut, and the local infection rate.
To help inform such difficult decisions, it is imperative that analysts have access to good data that can enable robust estimation of the increase in the infection risk from opening schools, as well as the negative impact on children’s learning and other outcomes from closing schools. Currently we have insufficient high-quality data on both these elements, an issue which needs to be urgently addressed. Further, decisions about opening and closing schools are likely to need to be made at a local level, in response to spikes of infection in some localities. Hence there is also a need for effective surveillance mechanisms and for data to be available at the local area level. The UK is also not alone in facing these decisions. Almost every other country also closed their schools. Likewise, most countries are currently drawing up plans to open their schools, and many have indeed opened schools already. We can learn from the experience of other countries and international collaboration and sharing of data is vital. Lastly, another focus of this report is on how schools can open safely with infection mitigations in place. Yet we have very limited robust evidence on the effectiveness of any such mitigations. Prospective studies should try to evaluate not only the infection risks from opening schools but also the impact of different mitigations: again we might learn much from considering what other countries are doing.
2. Opening Schools Safely
In this section, we summarise the guidance provided by intergovernmental organisations on schools during the COVID-19 pandemic, the emerging evidence on transmission of the virus, the approaches adopted by select countries to reopen schools and the data required to allow schools to manage school operations during this pandemic.
a. Intergovernmental Guidance
In March, the World Health Organization (WHO), the United Nations International Children’s Fund (UNICEF) and the International Federation of Red Cross and Red Crescent Societies (IFRC) provided detailed high-level guidance on the prevention and control of COVID-19 in schools 3 . In this report, a checklist of core non-pharmaceutical interventions (NPIs) such as, enhancing hygiene, improving airflow and providing clear communication on the risks of COVID-19 was provided. In addition, the WHO’s Public Health and Social Measures Annex on schools- published in May 4 -recommended that governments should adopt a risk-based approach to determine if, when and how schools readmit pupils. It also suggested that decision-makers should be informed on the latest evidence on transmission of SARS-CoV-2 among children and the severity in children, the local situation and epidemiology of COVID-19 where schools are located and the school setting and ability to maintain COVID-19 prevention and control measures. In this guide, a more detailed list of NPIs was provided, including case management, one metre distancing measures, limited class sizes, staggered start and end times, leveraging outdoor space, the use of “tele-schooling” and developing a policy to ensure masks or face coverings are worn in line with national or local guidance3,p.4. In June, UNICEF, the World Bank, the World Food Programme and UNHCR provided a framework to help governments choose the policies that are most suitable locally 5 . The United Nations Educational, Scientific, and Cultural Organisation (UNESCO), the World Bank and other international groups have also provided additional guidance to Ministries of Education across the globe.
b. Schools and infections
Evidence on COVID-19 prevalence among children, illness severity among children, transmission from adults to children and from children to adults, and transmission in schools.
While there are still gaps in the evidence-base, and better-quality studies are continually emerging, the available evidence (internationally) indicates that we can draw the following conclusions at this time.
There is good evidence that the prevalence of Covid-19 infection among children has generally been either similar to the prevalence among adults, or lower, in the UK and internationally. More data are required to judge whether prevalence in the UK is significantly lower among children than among adults, or whether prevalence is similar among children to among adults. We proceed to outline the evidence.
Firstly, we observe that the ONS Infection Survey pilot 6 (for England) has not found a statistically significant difference between rates of NAAT-positivity in children, versus in adults, with nine out of 3,117 children aged 2-11 testing positive (NAAT-positivity rate 0.29%; 95% CI: 0.13% to 0.55%), and nine out of 2,860 individuals aged 12-19 testing positive (NAAT-positivity rate 0.31%; 95% CI: 0.14% to 0.60%), between 26th April and 27th June. These should be compared to NAAT-positivity rates of 0.42% (95% CI: 0.31% to 0.56%) in the age-group 20-49, and 0.28% (95% CI: 0.20% to 0.39%) in the age-group 50-69. However, the confidence intervals for the different age-groups remain wide, due to the low rates of NAAT-positivity. Hence, more data are required for us to judge accurately whether prevalence in England is significantly lower among children than among adults, or whether prevalence is similar. The ONS data do, however, provide good evidence that the rate of NAAT-positivity in children in England, between 26th April and 27th June, was not much higher than that in adults.
We do not yet have reliable age-specific seroprevalence data for England. However, the very large Spanish ENE-COVID seroprevalence survey 7 found a statistically significant difference between seroprevalence levels in children versus in adults, with that in children being lower. This was a survey of 61,075 individuals between 27th April and 11th May; one of the highest-quality seroprevalence studies to date, worldwide. From this survey, seroprevalence in children in Spain aged 5-9 was estimated at 3.1% (95% CI: 2.2%-4.2%); in children aged 10-14 at 4.0% (95% CI: 3.1%-5.0%), and for those aged 15-19 at 3.7% (95% CI: 2.9%-4.8%). This compares with an overall population seroprevalence estimate of 5.0% (95% CI: 4.7% to 5.4%), for the Spanish population as a whole. The Spanish lockdown was particularly stringent for children, with under-14s not allowed to leave their homes between 14th March and 25th April; however, it was also stringent for adults - who, with the exception of key workers, were only allowed to leave their homes to buy essential supplies, or in the case of an emergency. Moreover, many COVID-19 infections in Spain will have been contracted before the lockdown was imposed (14th March), i.e. when most schools and workplaces were still open.
There are also relevant prevalence data from Italy. Following the first death from Covid-19 in Italy, in the municipality of Vò, more than 85% of the population of Vò was NAAT-tested in late February, and no positive cases were found among the 217 children aged 10 or under who were tested (NAAT-positivity rate 0.0%; 95% CI: 0.0% to 1.7%), despite 73 out of the 2,812 people tested (NAAT-positivity rate 2.6%; 95% CI: 2.0% to 3.3%), testing NAAT-positive 8 . Among the 250 individuals aged 11-20 who were tested, three tested positive (NAAT-positivity rate 1.2%; 95% CI: 0.25% to 3.5%). The start of the two-week lockdown in Vò coincided with the testing; schools had been open up until that point.
In some countries, most schools were kept open throughout the pandemic. Iceland has kept schools for under-16s open throughout the pandemic, though it closed upper secondary schools (for those aged 16 and above), from 13th March until 4th May. Population screening in Iceland 9 in early April found no SARS-CoV-2 NAAT-positive children under 10 years old, out of the 848 tested (NAAT-positivity rate 0.0%; 95% CI: 0.0% to 0.43%), as compared with 100 testing positive out of the 12,232 tested persons of aged 10 or older (NAAT-positivity rate 0.8%; 95% CI: 0.7% to 1.0%). Iceland, however, has had low prevalence across all age-groups.
Sweden has also kept most schools for under-16s open, though high schools (for those aged 16 and above), and universities, were advised to close and switch to distance learning, from 18th March, and local decisions could be taken to close schools for younger children thereafter, in the case of local outbreaks. According to an announcement 10 by the Swedish Public Health Agency, a medium-sized (n=1,104) seroprevalence survey in Sweden found that by the end of April, 4.9% of those aged 0-19 tested positive for SARS-CoV-2 antibodies, compared to 6.7% of those aged 20-64. The full results of this survey have yet to be published (even in preprint form).
Overall, we can say that the risk to children arising from Covid-19 infection is low. We can be essentially certain that the risk of death and of severe illness from Covid-19 infection in children is extremely low. Specifically, the infection fatality rate for children aged 5-14 is estimated at 14 per million (95% CI: 6.6 per million to 24 per million) 11 . This is lower than the infection fatality rate of seasonal influenza among children aged 5-17 during the 2018-19 ‘flu season in the United States, which was estimated by the CDC at 28 per million 12 , and is substantially lower than the overall population infection fatality rate of Covid-19, which is approximately 13 per thousand for the UK (95% CI: 11 per thousand to 15 per thousand) 13 . As of 23rd June, only five COVID-linked deaths of UK children under the age of 18 had been recorded, and we are only aware of one such death in which no pre-existing health condition was known. Severe illness is also much rarer in children than in adults. The percentage of symptomatic cases requiring hospitalisation is estimated 14 to be 0.1% among children aged 0-9 and 0.3% among those aged 10-19, compared to an overall population hospitalisation rate of 4.4%, for the UK. A recent SJD (Barcelona) study 15 , following 724 children living in the household of a confirmed Covid-19 case, found that more than 99% of the children in the study who subsequently tested positive, had only mild symptoms.
An extremely rare but severe Kawasaki-like multisystem inflammatory condition has been linked to SARS-CoV-2 infection in children. When linked to SARS-CoV-2 infection in under-18’s, this condition is known as ‘Paediatric Inflammatory Multisystem Syndrome Temporally associated with SARS-CoV-2’, or PIMS-TS. The Royal College of Paediatrics and Child Health estimated that, by early June, there had been approximately 200 cases of PIMS-TS in the UK among under-18’s 16 . The most comprehensive UK study of this syndrome to date (of which we are aware) is that of Davies et al 17 . This was a multicentre observational study of all children admitted to 19 participating Paediatric Intensive Care Units (PICUs) in the UK between 1st April and 10th May, who fulfilled the case-definition of PIMS-TS. There are a total of 23 PICUs in the UK, of which two were closed during the study period, having been converted into ICUs for adults due to Covid-19, so it can be expected that this study covered at least 90% of cases admitted to PICUs in the UK. There were 78 cases admitted to the 19 participating PICUs in this 40-day period; two of these children died. The median age of all these patients was 11 years. Only 22% tested NAAT-positive for SARS-CoV-2; the other 78% tested NAAT-negative; but 24/25 (96%) of NAAT-negative patients who also underwent a serological test, tested IgG serology-positive for SARS-CoV-2, indicating prior infection with SARS-CoV-2. These data represent an ICU-admission rate (for PIMS-TS) of approximately 1.1 children per million, per week, during the 40-day period of the study. While still being an extremely rare event, children of Black and Asian ethnicities were more likely to be affected; for children of black ethnicity, the ICU-admission rate per week was about 6.3 children per million per week, and for children of Asian ethnicity, the rate was 4.3 per million per week. For comparison, in 2018, there were 26 ICU admissions per week per million children, in the UK (all causes). 18
In view of the above, we can be essentially certain that Covid-19 infection typically produces milder symptoms in children, than in adults. This means that the risk of severe disease is less in children, but also that children may more often carry the disease without detection. The proportion of children who are truly asymptomatic (as opposed to pre-symptomatic or pauci-symptomatic, at a given time) remains unknown. Data from Italian emergency departments suggested that 21% of SARS-CoV-2 NAAT-positive children were asymptomatic at the time of testing, but there was no follow-up to determine whether these children later developed symptoms; some could therefore have been pre-symptomatic. Other studies have found comparable figures, but we are not aware of any such study where follow-up was performed.
The balance of the available evidence suggests that children may be less susceptible to infection than adults, given the same level of exposure. Four high-quality studies found odds-ratios for secondary attack ratios of children, versus adults of 0.16 (95% CI: 0.06-0.46) 19 ; 0.27 (95% CI: 0.13-0.55) 20 , 0.21 (95% CI: 0.11-0.41) 21 and 0.34 (95% CI: 0.24-0.49) 22 , though one high-quality study, of Bi et al 23 , found no significant difference, with an odds-ratio of 0.82 (95% CI: 0.48-1.43). The very high-quality study of Zhang et al,2222 which found an odds-ratio of 0.34, excluded spouses for this purpose, and also considered only primary cases who were quarantined after diagnosis in local government facilities.
The study of Q.-L. Jing et al 20 also contained ten primary cases under the age of 20. The overall secondary attack rate for these cases was estimated at only 5.2% for household contacts (95% CI: 2.4% to 9.7%), and at only 1.4% for non-household contacts (95% CI: 0.04% to 7.6%). For primary cases between the ages of 20 and 59 (of whom there were 145 in the study), the corresponding secondary attack rates were 14.8% (95% CI: 11.7% to 18.4%) for household contacts, and 2.2% (95% CI: 1.4% to 3.4%) for non-household contacts. This indicates that children may be less likely than adults to transmit COVID-19 infection, though studies with a larger number of children as primary cases are needed, to be more certain of this conclusion.
So far, there is limited evidence available for quantifying the extent to which schools contribute to overall nationwide transmission of Covid-19, partly as many countries closed schools early on in their epidemics. The balance of evidence available so far suggests that schools play only a limited role in overall transmission 24 . This is in contrast with influenza, where there is strong evidence that transmission among children in schools is an important driver of overall transmission 25 , 26 . We outline some of the evidence below.
One of the most extensive studies for which results are available was a contact-tracing study 27 across 15 schools in New South Wales, Australia, from 5th March to 3rd April. This study involved 18 initial cases (nine children and nine teachers) who had opportunities to infect others. It indicated very limited transmission in a school setting, with just two likely secondary cases arising, both being students (one in a primary school, where the likely primary case was a teacher; and one in a high school, where the two likely primary cases were students). In detail, the six initial cases in primary schools (five staff and one pupil, across five schools) were judged to have 168 close contacts between them; of these contacts, 137 were pupils and 31 were staff. Just one additional case was detected, though in some cases, contacts were only tested if they developed COVID-like symptoms, so only 32% (53 out of 168) of contacts were tested (indicating a risk of possible undercounting of asymptomatic secondary cases). In the primary school where the single additional case was found, there was just one initial case (a staff-member). In that school, 21 close contacts of the initial case underwent serological testing; no other additional cases were found. A review found that it was most likely (but not certain) that this additional case (a pupil) was infected in the school environment, i.e. it was likely to be a secondary case corresponding to transmission from the primary case in that school (who, as mentioned above, was a staff-member). The 12 initial cases in high schools (8 students and 4 staff), were judged to have 695 close contacts between them, of whom 598 were students and 12 were staff. Only one additional case was detected (a student). Nose/throat swabs were taken from one third (235 out of 695) of these close contacts, for NAAT-testing; all of these tested negative. In one high school, where there were two initial cases (both students), 75 close contacts underwent antibody testing one month after contact with one (or both) of the two initial cases. Just one of these contacts (a student) tested positive.
During February and March, the nationwide test-trace-isolate programme in Singapore detected (through NAAT-testing of contacts of confirmed cases in the wider community), three index cases of Covid-19 in schools, as described in the study of Yung et al 28 . One was a pre-school student, aged 5, another a secondary school student, aged 12, and the third was a pre-school teacher. Only the third index case (a teacher) is thought to have caused any secondary cases, and no secondary cases were detected among students. In the first two cases, the students in question were found to be NAAT-positive from contact-tracing following their exposures to adult household-members who were part of a community cluster. Both students attended their respective schools on the first day of their symptoms, before being subsequently diagnosed with Covid-19 and isolated in hospital. No secondary cases were detected in either of these two schools. All close contacts of the index cases (including all classmates) were placed under a 14-day quarantine and requested to monitor their symptoms; those who were not deemed close contacts were permitted to continue with classes, and asked to monitor themselves for possible symptoms; they were admitted for Covid-19 evaluation if they became unwell during the subsequent 14 days. Eight students from the secondary school developed COVID-like symptoms during the 14-day period and were NAAT-tested, but all tested negative. Similarly, 34 student contacts from the pre-school developed COVID-like symptoms during the 14-day period and were NAAT-tested; all tested negative. However, in the preschool setting where the primary case was a teacher, 16 other cases of Covid-19 infection were confirmed among adult staff-members in the preschool, with an additional 11 cases subsequently being discovered in their households. A total of 77 children from this preschool (73% of all students) underwent NAAT-testing; all tested negative. The remaining 27% who chose not to provide a swab, did not develop any symptoms while under close monitoring and quarantine.
Finally, the study of Heavey et al 29 , of all reported COVID-19 cases in Irish schools from 1st March to 13th March, found six index cases (three students and three teachers), but no secondary cases were detected in school settings, despite 1,155 school contacts of these six index cases being identified. The only detected (probable) secondary case was in an adult, not in a school setting. (We remark that only those contacts displaying any possible symptoms, including mild symptoms, were NAAT-tested, so some asymptomatic secondary cases might have gone undetected.) According to the report, the available epidemiological data for all six index cases indicated that they had not been infected with SARS-CoV-2 in the school setting.
We note that there have been some other studies and media reports concerning COVID-19 outbreaks in schools, for example, at a school in Oise (Northern France) 30 , at Gymnasia Rehavia middle and high school in Jerusalem (Israel) 31 , and in a school in Santiago (Chile) 32 . However, in the schools in Oise and Jerusalem, it is unclear whether the initial cases were pupils or staff; in the school in Santiago there is good evidence that the index-cases were parents or teachers and that transmission took place during a week of parent-teacher evenings; and in the schools in Oise and Santiago there appears to be no evidence that transmission took place within the schools (excepting in the parent-teacher evenings in the school in Santiago). We are not aware of any official outbreak report or academic study on the Gymnasia Rehavia (Jerusalem) middle and high school outbreak.
Particular risks from infection for BAME families and teachers
Risks from COVID-19 are higher for some groups than others, for example varying by age and underlying health conditions. Clearly, as in other professions, care will be needed when opening schools to consider the health risks, and necessary mitigations, for older teachers and those who have been sheltering with particular underlying health conditions. A key question when opening schools is also whether teachers and parents from BAME groups 33 may be at particular risk and hence whether this should be taken into consideration when making decisions about school opening/closing.
The issues Black, Asian, and Minority ethnic groups are facing in the U.K. mirror those seen among minorities in the United States. According to the latest Public Health England report, Black, Asian, and Minority Ethnic populations have been disproportionately impacted by COVID-19. Five possible factors contribute to the increased vulnerability of members of the BAME community: Occupation, Living Conditions, Comorbidities, Healthcare Bias/accessibility, and Discrimination.
BAME families are overrepresented in low wage jobs and overcrowded areas, making them more vulnerable to COVID-19. BAME households are often extended and multigenerational cohabiting families 34 . In England, Black people are nearly four times as likely as white people to have no access to outdoor space 35 . Higher proportions of BAME people live in deprived areas of London and more likely to have concerns over safety and security. Due to these living conditions, when one member of the family contracts COVID-19, is it highly likely the rest of the family will contract the virus. The most deprived areas of England and Wales had more than double the number of COVID-19 related deaths than the least deprived areas (58 deaths per 100,000) 36 .
Studies have linked comorbidities to poorer COVID-19 outcomes, specifically hypertension, and diabetes 37 , 38 . Previous studies have shown that Black and South Asian populations in the U.K. have three to five times the prevalence of type-two diabetes compared to the white people and are diagnosed at a younger age 39 . Comorbidities are not only a risk for BAME adults but also children from BAME back