Chapter 1
Entering lockdown was the only way for the government to suppress Covid-19 and bring the number of cases down to a manageable level. Now, easing lockdown requires new measures that attempt to contain the virus and enable people to return to some level of social and economic activity.
There is no such thing as a risk-free easing measure. If we want to avoid the impact of Covid-19 entirely, then we need to stay in lockdown until a vaccine is found. This would lead to unprecedented economic damage and would severely harm the population’s health and wellbeing in other ways.
Easing is a fine balancing act. The government must identify and choose easing measures that bring the maximum gains for the lowest increases in infection rates. Most discussion has considered this through the lens of the R rate, circling in on measures that will keep the average number of infections down per infected person. This paper looks at the ‘k’ rate – the dispersion of infections – and explores the “super-spreader phenomenon” where different people can spread more or less of the virus than others.
By focusing on who, how and where transmissions occur, we can begin to identify high leverage easing measures where there is the lowest likelihood of producing “super-spreading events” (SSEs) – moments that cause a spike in infections. We identify schools as a good example of this, owed to emerging data on transmission rates among children and the impact of Covid-19 on young people. Focusing on schools is especially important because the cost of closure is so high – for education outcomes, for social inequality and for the impact on parents’ time and ability to work. Based on what we know, reopening schools should be one of the highest priorities while allowing large events should be one of the lowest.
Such a policy focus must be underpinned by containment infrastructure. This includes mass testing and tracing, as set out in our previous paper The Architecture of Containment: Getting to Gold. Tracing is particularly important as it will mitigate the undoubtedly low but real chance of children becoming infected and spreading to their families out of school. We welcome the contribution from the independent SAGE group in this regard.[_]
Ultimately, this paper presents the super-spreading phenomenon as a potential blueprint for informing and guiding policy decisions around easing lockdown. It should continue to be extensively researched and understood.
Chapter 2
Deciding on what measures should be implemented to ease lockdown has mainly focused on reducing the average number of transmissions per infected person: the R rate.
The k rate of Covid-19 represents the dispersion factor of a virus and shows who and how many people are transmitting it rather than the average rate of transmission (the R rate). It has the potential to better inform containment strategy by identifying targeted measures to focus on heavy transmitters or “super-spreaders”.
These will be more effective than measures applied evenly across the entire population – for example when it comes to closing all schools.
The impact of Covid-19 appears much less severe on children, and emerging evidence suggests that children are less likely to transmit the disease.
School safety is the responsibility of adults and the government should work to minimise their chances of getting the disease or passing it on.
When it comes to reopening schools, the challenge facing government is three-fold:Protect and prevent adults in the learning environment from getting and transmitting Covid-19.Protect and prevent children in the learning environment from getting and transmitting Covid-19.Contain any transmissions and prevent them spreading beyond schools.
Based on emerging evidence on the interrelation between age and transmission, we propose a focus on adults in the learning environment.
This leads to three practical recommendations:Reopen schools for under-10s.Equip teachers and adults in school environments with PPE and ensure they practice social distancing in and outside of lessons.[_]Create rapid response strategies to respond to school outbreaks, including testing and localised school closures where necessary. This requires testing and tracing infrastructure to be in place, preventing transmissions beyond the school gates.[_]
This approach is supported by emerging evidence. In countries that have reopened their schools, we have not seen significant increases in new cases.
We must remain cautious: Where localised outbreaks occur, it makes sense for swift action to be taken in order to temporarily close schools.
We recommend further scientific investigation into the super-spreader phenomenon to enable decision-makers to better understand what types of people are super-spreaders, to predict SSEs, and to develop optimal strategies to respond to such events and, as a result, prioritise translating those into policy.
This will provide governments with an opportunity to move from expensive blanket restrictions to a more targeted containment strategy. This includes enumerating SSEs and establishing regular review of the basic characteristics of these events, then working with key influencers (e.g. faith leaders in the case of religious gatherings) to develop a plan, which can be adjusted as we learn more.
Such an approach could be similarly effective in containing Covid-19, while enabling many economic activities to restart.
Chapter 3
Why do some Covid-19 patients infect many others, whereas most don’t spread the virus at all?[_]
To date, government responses to Covid-19 have primarily been informed by the reproduction number, R. Reducing this number through suppression of Covid-19 and keeping it below 1 by containing the spread of the disease has been the key objective. This has meant most countries, bar some notable examples, have entered lockdowns and slowly eased from these.
The R rate is an average of every infected person. For example, R=3 means that the average infected person would infect three others. As an average measure, R provides no information on the variation in infectiousness between individuals. In practice, many infected individuals will not pass on the virus to anyone, while others may pass it on to tens or even a hundred people. A 2005 paper[_] looking at transmission concluded that a minority of individuals are responsible for a majority of transmission and the distribution of spread differs between pathogens. Scientists describe this with a measure called the “dispersion factor”, or k.
Models that account for this variation differ sharply from average-based approaches. This could potentially explain why some apparently contagious diseases don’t cause as much damage as initially projected. Clearly, this is not the case with Covid-19, which has seen a high number of deaths. However, the study demonstrates that “individual-specific control measures outperform population-wide measures.” This could have profound impact on the decisions governments must make when it comes to easing lockdowns and managing the disease until vaccines and therapeutics are able to play a large part.
K is a number greater than 0, where higher values indicate that the virus is spread more evenly and lower values indicate more variance between who is infectious. Each disease differs in terms of how many and to what extent people spread it.
Figure 1 – Evidence for the variation in individual reproductive number
a) Transmission data from the SARS outbreak in Singapore in 2003.
b) Expected proportion of all transmission due to a given proportion of infectious cases, where cases are ranked by infectiousness.
c) Proportion of transmission expected from the most infectious 20 per cent of cases, for ten outbreak or surveillance data sets (triangles).
d) Reported superspreading events (SSEs; diamonds) relative to estimated reproductive number R (squares) for 12 directly transmitted infections.
Source: Lloyd-Smith, J., Schreiber, S., Kopp, P. et al. Superspreading and the effect of individual variation on disease emergence. Nature 438, 355–359 (2005). https://doi.org/10.1038/nature04153
What About Covid-19’s ‘k’ Rate?
Research is still ongoing for Covid-19 but there are signs that the transmission rate is very unevenly spread. A recent study suggested that there was “a high degree of individual-level variation in the transmission of Covid-19” and that “80% of secondary transmissions may have been caused by a small fraction of infectious individuals (~10%).”[_]
Figure 2 – Illustration of R rate of transmission (first graph) versus reality/k rate (below)
Source: Wie sich Covid-19 in Österreich ausbreitete - derStandard.at, 2020
This chimes closely with an Israeli study on the genomic sequencing of the virus[_], conducted by nine research institutes. The study tracked the rate of infection in Israel and concluded that 5 per cent of the infected population (“super-spreaders”) were responsible for 80 per cent of the spread of the virus in Israel.
According to the study, many of the patients tested were found to have genomes of the virus that were 100 per cent identical, leading to the conclusion that they originated from a small set of super spreaders.
The research team, led by Tel Aviv University scientists, aims to continue with the study in order to yield detailed information on the rate of infection. This should provide authorities with a means of targeting containment measures in case of a second wave, in order to target closures and other measures more effectively.
How Do We Know If Someone Is a Super-Spreader?
We will be unable to identify specific individuals who are super-spreaders until the underlying causes are fully understood. These causes are often difficult to measure, and they interact with each other. Part of the explanation is likely biological – some people may shed more virus or be immunocompromised, leading them to be infectious for longer. Part of the explanation is going to be behavioural – how often people wash their hands, how many other individuals they come into contact with (particularly indoors), and so on.[_]
To identify likely super-spreader demographics, we still need to better understand the biological and physiological factors that give rise to higher individual transmission rates. The data on this is emerging. We have referenced some of the available evidence on children, a key group in relation to the important topic of school reopening.
Do Children Spread the Virus?
The government is currently undertaking a mass antibody surveillance test, which will give insight into transmissions more generally, and the role children play. It is already widely known that children are much less likely than adults to die from coronavirus and usually are asymptomatic or present with mild symptoms (see section on “Child Outcomes” below).[_] This has led to concerns that asymptomatic cases in children are a significant vehicle of Covid-19 transmission.
Looking at the available evidence, we can see that, so far, there has been a lower number of cases, worldwide, of children compared to adults. This may owe, in large part, to the criteria for selecting who is tested. However, research across multiple countries has generally found that children who are tested are less likely to be found to be infected, suggesting that children are less likely to contract Coivd-19 compated to adults.[_] For example, an Iceland study found that children were about half as likely to test positive,[_] while in the Italian town of Vo, blanket testing found that despite a population infection rate of 2.6 per cent, no children under 10 had a positive test.[_]
The role of children in passing the disease to others is largely unknown, not least because when children are infected, they are generally asymptomatic. Since children appear to be significantly less likely to become infected than adults, this implies their role in spreading the disease is small. However, it is not currently well understood whether infected children are likely to spread the disease. A summary of what we know about child transmission is in the the section on “Child Outcomes” below. Overall, while the evidence is equivocal, and children may be involved in some transmission, it is unlikely that a child will be a super-spreader who infects many individuals.[_]
Child Outcomes
The latest ONS figures shows that deaths in children caused by Covid-19 are very rare.[_] In March and April, there were eight deaths recorded where the patient was under 20. Of these, seven were in the 15-19 age group, one was in the 10-14 age group, and no deaths were recorded for younger children. This is consistent with a global pattern of very low death rates for children.[_]
Children tend to have a mild illness. A large epidemiological study in China looked at 2,143 paediatric patients from Hubei with suspected or confirmed cases of Covid-19 and found only 5.6 percent of children had severe cases, and 0.6 per cent developed a critical condition, which suggests risks of hospitalisation are low, although there is some evidence that infants are at greater risk of developing severe cases.[_] A report from the US Centres for Disease Control and Prevention in April similarly found 2,572 confirmed cases involving under 18s (1.7 per cent of all confirmed cases), of which 5.7 per cent were known to have been hospitalised. By comparison, 10 per cent of those aged 18-64 were known to be hospitalised.[_]
There is currently very little clinical data on whether children with comorbidities are worse affected than other children. Recently, examples have emerged of children developing severe cases of a delayed inflammatory response.[_] The mechanisms are not fully understood, but the condition appears to be exceptionally rare.
How Infectious Are Infected Children?
While current evidence strongly suggests children are less likely to contract Covid-19 than adults, little is fully understood on how likely infected children are to transmit the virus. Some evidence suggests that children are less likely to pass their infection on to others:
A review of studies in China, Singapore, South Korea, Japan and Iran found that children were the primary case of a chain in household infection cluster in less than 10 per cent of cases.[_]
A case study of an infection cluster in the French Alps found that an infected child spread the virus to no other individuals despite being exposed to more than 100 other children.[_]
A study from the Netherlands found that transmission from children is very rare.[_]
An Australian study in New South Wales found that 18 infected individuals (nine students and nine staff) did not transmit the virus to anyone else in their school, despite coming into contact with 735 other students and 128 staff.[_]
The China/WHO joint commission could not recall any episodes during contact tracing where transmission occurred from a child to an adult.[_]
This evidence, however, is largely based on small numbers of cases. For the larger studies looking at household transmission, criticism has been raised that children with asymptomatic cases may not have been tested.[_]
A recent German study, by contrast, claimed to find that viral loads for infected children are not significantly different from adults.[_] However, the statistical methodology of this paper has been challenged, with some researchers arguing that the data suggests children do have lower viral loads than other groups.[_]
Chapter 4
Even if it is not possible – or sensible or ethical – to identify individual super-spreaders, the low k rate of Covid-19 suggests that simply preventing large gatherings of individuals where there is a high probability that at least one attendee is a super-spreader can substantially bring down infections. In other words, we can reduce infections by preventing super-spreading events.[_]
Is School a Super-Spreading Event?
The efficacy of school closures as a containment measure for Covid-19 depends on how likely SSEs are to occur within schools. Based on the size of the gathering – potentially hundreds or thousands of pupils – school closures make sense. The case for closure erodes if a very significant number of the gathering is highly unlikely to include one or more super-spreaders. A school operating with appropriate measures in place, including tracing, doesn’t possess the potential to host a super-spreading event.
Protecting Teachers
Given that children are at low risk of infection and their outcomes when infected are good, the main concern about opening schools is that teachers can spread it among each other, endangering their lives while preventing them from being able to care adequately for the children. The low risk of transmission from teacher to child and vice versa means there is also a threat of increased community spread. For schools to open safely, therefore, the focus should be on minimising transmission between teachers and ensuring adequate tracing infrastructure is in place to prevent any (albeit unlikely) outbreaks from reaching the community. Given that the risks to primary school children appear to be the lowest of any age group, and primary schools tend to have smaller class sizes than other schools, such schools are a good place to start.
Recommendations for Schools:
Teachers should wear masks and PPE where possible, and adhere to social distancing from other teachers and adults while at school to stop transmission.
Phase the re-opening of schools to focus on primary and early years first.
Monitor schools to rapidly detect and respond to localised outbreaks, ensuring positive engagement with local authorities to identify any inter-school patterns. There is a pre-requisite of adequate tracing infrastructure being in place for this.
Commit to meaningful and ongoing dialogue with schools and key stakeholders, including teacher associations and unions, setting out the science and mitigating measures.
Reopening Schools Should Be a Priority in Exiting Lockdown
A review article from the Lancet looked at the impact of school closures for previous coronavirus outbreaks, and found that during the SARS outbreak, school closures did not appear to help control the epidemic.[_]
It is difficult to measure the impact of school closures for the Covid-19 pandemic directly, because countries have tended to implement school closures and other social-distancing measures simultaneously, making the effects difficult to disentangle. Comparing across countries is also problematic, both because it is impossible to control for all the relevant differences between countries, and because – as we have discussed in previous papers – confirmed case and death numbers are not recorded consistently between countries.
Many European countries have begun reopening their schools. While there is a lag in the impact of a new measure on Covid-19 cases and deaths, and most countries only reopened schools in the past two weeks, the picture can cautiously be characterised as positive. In the 22 European countries that have reopened schools, none have seen a significant rise in overall infections.[_] A few countries began opening schools in April, such as Denmark and Norway, but in both cases, the overall trend of infections has followed a similar downward trend from their peaks.[_] Taiwan never closed its schools (although it extended a winter break by ten days in February to disinfect schools and enact policy changes)[_], and has had fewer than 500 confirmed cases and seven deaths.
In France, school reopening has attracted some backlash, after 70 confirmed cases were linked to schools and eight schools were closed.[_] While this might appear to be a cautionary tale about reopening too early, it is important to note that this pattern has not been the norm in other countries. When such cases arise, it is important to respond quickly, and localised school closures may be an appropriate response until the outbreak is contained. The French response is characteristic of an approach that flexes to the spread of the virus to curtail any potential super-spreading events.
As governments continue to ease lockdown restrictions, the order in which restrictions are lifted needs to be prioritised. Governments should aim to remove restrictions based on how costly they are to society to maintain, and how much they help to keep infection rates under control.
Reopening schools as a cornerstone of easing lockdown is therefore sound policy. It is likely that closures have limited benefit in controlling the virus. School closures result in children losing out on their education and the opportunity to engage in essential socialisation activities with their peers. They also create challenges in childcare for parents, who may be juggling many commitments with work and family. Children who rely on free school meals may struggle to access them, and confinement can expose children to greater risks of child abuse and neglect.[_]
The evidence on school closures is equivocal, but even if closing schools reduced transmission rates, it may have been possible to achieve the same benefit by adjusting the way schools are run. In light of these early indications, a focus on protecting teachers may achieve similar benefits to school closures without so many of the associated harms.
Chapter 5
Given the strong evidence that a small minority of individuals infected with Covid-19 are responsible for the majority of transmission, this suggests a path forward for governments looking to restart their economies while keeping new infections under control. A tactical, targeted approach focused on preventing and rapidly containing super-spreading events could allow many population-wide restrictions to be lifted.
In the case of schools, the evidence suggests that they are unlikely to be hotbeds of SSEs as long as teachers are protected and tracing infrastructure is in place. Going forward, it should be a priority for governments to improve their understanding of what types of individuals are most likely to be super-spreaders, forecast SSEs, and develop rapid response policies to contain SSEs when they occur. This will make it far easier to deploy policies that thread the needle between rapid economic recovery and avoiding a second wave of infections.