Show Hide image

How Sage sees the risks of leaving lockdown

Publicly available minutes and analyses tell a story of caution in the face of inadequate evidence – and trepidation about what comes next.

It is a lot harder to exit lockdown than it was to enter it. It took one week in March to move from hand-washing to a new world of isolation, quarantine and furlough. And that is how we stayed until 11 May, when steps began to be taken back from the new world to the old. People are out and about more, seeing friends and family, cautiously returning to work and even more cautiously getting some children back to school. But having been frightened into a remarkable degree of compliance, many still worry about the risk of infection. The first message – Stay Home – was a clear instruction. Its replacement – Stay Alert – is a vague instruction that asks individuals to navigate a changing set of rules in an environment of unknowable risk.

All this, the government insists, is led by science, but science is being questioned. From one direction it is being criticised for having led the country too slowly into lockdown, losing vital days waiting for better data and arguing for a graduated approach. From another direction come complaints that scientists panicked, and rushed the country into unnecessarily severe restrictions that took away personal freedoms and devastated the economy.

The target for both sets of critics is Sage, the Special Advisory Group for Emergencies. The body responsible for acquiring and channelling scientific advice has grown during the pandemic into a large and complex operation. Meetings are routinely attended by over 50 people. Much of the detailed work is now assigned to sub-groups and sub-sub groups. Those involved include not only high-profile modellers but experts from a range of disciplines, including biomedical, behavioural and environmental science. Two key sub-groups are SPI-M and SPI-B. SPI stands for "Scientific Pandemic Influenza Group", M refers to Modelling and B to Behaviours.

For example, Professor Peter Horby – the co-investigator responsible for Oxford University’s breakthrough treatment using dexamethasone – chairs Nervtag (New and Emerging Respiratory Virus Threats Advisory Group). This has a separate existence to Sage but feeds into its deliberations and has an overlapping membership. This was the group that considered such questions as the significance of the loss of taste and smell as symptoms, the mechanisms by which people could become infected, levels of immunity and trials of treatments.

Fortunately, Sage has been assiduous in publishing both its own minutes and the analyses that have shaped its advice. So far, most attention has been paid to the early assessments of the danger of the novel coronavirus and how the various interventions for containing and then suppressing it were evaluated. Now that we have Sage’s minutes up to mid-May, we can also analyse the first weeks of the group’s advice on when and how to get out of lockdown.

Those looking for confident recommendations on the way forward will be disappointed. The minutes are full of uncertainty and caution, frustrated until late April by the limited numbers of tests and the lack of evidence on the patterns of infection. The situation improved in late April, but until that time Sage was unsure not only about the nature of the coronavirus but also about what was going on in the country.

The behavioural models created before lockdown raised concerns that the public would not comply with strict measures of social distancing. The seriousness with which they were taken was therefore one of the most striking early observations. On 1 April, the behavioural science group (SPI-B) reported the results of a YouGov survey undertaken at the end of March. Just 13 per cent of the population were going to their place of work, 84 per cent had stopped seeing members of their family who did not live with them, and 91 per cent had stopped seeing friends. The immediate impact of lockdown had been to push R, the reproduction rate, to below 1. Meanwhile, the evidence coming directly from the NHS on admissions, critical care and fatalities, confirmed that the disease could affect anyone, other than possibly young children, but the worst hit were the elderly, men, and those with co-morbidities, notably heart and lung disease, as well as diabetes. Over time obesity took on greater salience while the disproportionate impact on ethnic minorities became more pronounced.

With social distancing in place and the patient numbers growing, the priorities for Sage changed. Prior to lockdown the focus had been on identifying the optimum moment to flatten the curve of the first wave, taking into account that a second wave was virtually inevitable. Once the decision had been taken to suppress the curve, the timing of any future surge of cases became a matter for policy. On the one hand the new regime, with its economic and social costs, could not be maintained indefinitely. On the other, lifting measures too early could see the return of exponential growth. The question became when and how restrictions could be eased safely. Fears about the combination of a second wave in the winter, combined with the arrival of seasonal flu, would have to wait.

This was reflected in the modelling. The fundamental problem was that, because too few tests were being carried out, scientists did not know exactly how the coronavirus was spreading. On 1 April, the modellers (SPI-M) observed that they still did not know what proportion of the population had been infected, nor did they fully understand transmission.

Because social distancing measures had been introduced quickly, it was hard to disentangle the effects of these two unknowns. The peak of infections and deaths was yet to be reached. It was hard to be sure what the R number really was, which made prediction difficult. If it was 0.6, the weekly death toll was predicted to fall to 500 by the end of April; if it was just over 1, the weekly toll was forecast to remain over 1,000 until the start of September.

On Friday 10 April, the day after the UK hit, as anticipated, its peak of infections and deaths, the Cabinet Office set Sage some weekend homework. The agenda was clear. Was R less than 1? Could any of the measures “be amended, eased or lifted immediately with a high degree of confidence that doing so would have a negligible effect on R”? (They gave as examples communal space in parks or garden centres.) There were questions about possible alternative means to suppress the virus with “lower wider economic and/or societal costs”, getting people to go to their place of work if they could not work at home, getting more children into schools, and the possibility of smarter, “non-pharmaceutical interventions”.

Both the modellers (SPI-M) and behavioural scientists (SPI-B) provided answers. The only good news from SPI-M was that the transmission of the virus had slowed and might be in decline. In the community, R was likely to be 0.8-0.9. The bad news was that that there was significant “nosocomial” transmission: people were catching the virus in hospitals. Individuals who were admitted to hospital virus-free accounted for up to almost a quarter (22 per cent) of those who were then hospitalised with Covid-19, and 5-11 per cent of the most recent deaths – possibly more. They noted the implications for care homes, but at the time they had little data. In these environments, estimates for R were not possible.

Modelling predicted that any changes to policy that increased transmission of the virus by as little as 10 per cent in the community would cause a return to exponential growth. Any amendments to social distancing would have to be made gradually, so that their individual impacts could be assessed. Social distancing remained the key to keeping R down.

SPI-B took a similar view. If there was to be any relaxation, outdoor spaces would be a good place to start, as they were both safer and contributed to public health by promoting exercise. They warned, however, against any suggestion that there were no longer risks in engaging “in non-essential work, even if this involves poor social distancing in the workplace or when travelling to it”. Their advice was that there should be no deviation from the core message that “even brief, socially distanced encounters outside the home are so risky that they must be avoided”.

Sage passed all this back to the Cabinet Office with a stark conclusion: “There are no amendments to social distance measures that Sage can say, with a high degree of confidence, would have a negligible impact on the reproduction number. Sage does not recommend changing any measures now.”


Inadequate testing continued to hamper understanding, especially given the evidence of persistent Infections in hospitals and care homes, then suffering from shortages of personal protective equipment (PPE). SPI-M noted on 20 April that a testing strategy “should be targeted at those of highest risk”, and was needed “as rapidly as possible”.

A week later, the group expressed its concern that “not enough is being done to protect those who are known to be at high risk of death if infected with Covid-19”.

The frustration with the failure to address the high numbers in care homes can be seen in the Nervtag minutes, where questions were asked about staff moving between care homes and hospitals. Its minutes of 24 April report a study over the Easter weekend that “found  that  75 per cent  of  the  residents  carried  the  virus  and  only  25-33 per cent were  symptomatic.  Approximately  45 per cent  of  the  healthcare  workers  were  also  carrying  the  virus, with 25-33 per cent symptomatic.”

Early in May, Sage was still urging more extensive testing of care home residents and staff, and improved systems for managing the disease in those settings. The Sage minutes of 21 May (the latest available) observed that “hospital or care home cases now represent a higher proportion of total cases, possibly a majority”. The most recent statistics suggest that a third of all care homes in England have experienced an outbreak.

Testing to establish past levels of infection was even further behind. Large-scale serology surveys (antibody tests) were needed but, despite early promise, the available tests were unreliable.

On 16 April, Public Health England acknowledged that it lacked the capacity to run large-scale community surveys, so the Office of National Statistics took on the task as part of its regular household surveys. Meanwhile, international evidence gave no reason to suppose that at this stage a large proportion of the population had already been infected. Because this was a new disease, there was little evidence on whether and for how long infection conferred immunity. All this cast doubt on the idea of “immunity passports”, equivalent to a certificate of vaccination, that at one point was canvassed as enabling individuals to travel widely.

By the start of May it had become possible to consider some relaxation of the measures, which the government planned to announce on 11 May. A week earlier, SPI-M assessed the potential effects of the plan. There were four distinct phases to consider: after 11 May, further steps would be taken on 1 June, 1 July and 15 August. Each would see a progressive increase in work and leisure contacts and children returning to school. Work contacts would go up by about 10 per cent at each stage, leisure contacts would rise to 75 per cent by phase four and, after a cautious start, all children would be back at school for September.

As if to emphasise that the findings were not just the result of one group of modellers, four teams (Imperial, London School of Hygiene and Tropical Medicine, Bristol and Warwick) produced separate analyses. They varied because of different assumptions about R across the country and the susceptibility of children to infection (another matter on which the evidence remained uncertain) but the conclusions were largely the same. Phases one and two (the points we have now reached) were not considered too problematic. Phase three would be challenging and phase four risked a return to exponential growth, especially if businesses involving “close, sustained contact with many people”, such as hairdressers, were included.

SPI-B emphasised the importance of clear explanations of “why and how the selected activities are safe to resume”. The message should be that any relaxation did not mean that the risk from the coronavirus was over or that it was now safe to abandon other measures. Any evidence that infections were on the rise again would trigger a return to restrictions.

In the face of this caution from its scientists, the government was left urging the public to be patient. Could anything be done, other than contact tracing, to make it easier to cope with the challenges of phases three and four? One popular answer was the use of face masks. This was a matter for Nervtag, which had long held the view, in line with the World Health Organisation, that these offered no obvious benefits to the general population. Nor could there be any spare supplies so long as front-line staff were struggling with shortages. Masks would not protect individuals from infection. At most they could prevent an infected individual from passing the virus to others, but those who already had symptoms should be at home.

Two factors shifted the debate on masks. The first was that mask-wearing was common practice in many countries with low rates of infections. The US had begun encouraging people to wear masks when they left home. Was this going to be another example of the UK going against the international consensus? In addition, there was evidence that individuals who were about to show symptoms (presymptomatic) or were infected but without symptoms (asymptomatic) could unwittingly shed the virus. After some back-and-forth with Nervtag, on 21 April Sage concluded that: “On balance, there is enough evidence to support recommendation of community use of cloth face masks for short periods in enclosed spaces where social distancing is not possible.” It would be a bad idea if they were worn for long periods and they would not be of great value outdoors, unless in very crowded situations. There could be no suggestion that they were a substitute for keeping distance and washing hands. Because the virus could attach itself to the masks, they would need to be taken off with care. So while people might wear face masks because it made them feel safer, they would not get much benefit – but if they were already infected it might reduce the risk they posed to others.

Another idea was that of “bubbles”. These would allow people in small, non-overlapping groups of households to come into contact with one another so that they effectively formed one large household. The obvious advantage of bubbles was that they enabled people to reunite with family and friends from whom they had been separated and provided additional support to vulnerable people and easing loneliness. The obvious downside was that infection transmits most readily within families. A large bubble would put more people at risk, including anyone who was supposed to be shielding. This would especially be the case if one person was likely to be exposed to Covid-19, for example as a health or care worker, and certainly if people tried to be part of more than one bubble. If one member of a bubble became infected then all would have to isolate. On 14 May Sage advised against bubbling in the short term until the effects of the other relaxation measures could be assessed. The advice was that should “bubbling” be allowed, as it was this month, then it should be introduced in a staged way.

Through this period the importance of an effective test, trace and isolate (TTI) system was emphasised. On 27 April SPI-M assessed that it would “require around 80 per cent of non-household contacts of symptomatic cases to be traced and isolated rapidly, ie within two days of symptom onset for the index case”.

At the end of April, SPI-B observed that there was likely to be a better response by contacts if the index case had been tested, and was not just showing symptoms. This required tests to be completed within 24 hours. This was the position adopted by Sage. On 6 May, the group saw problems with using a phone-based app as opposed to manual tracing. The apps were not being well used in other countries, there would be problems with vulnerable people lacking smartphones or not using them properly, and a lack of trust meant that individuals would be less likely to self-isolate for two weeks solely from a phone message. One advantage of manual tracing was that it could tell testers more about the environment in which an outbreak had occurred. A report from the Royal Society on 18 May considered the phone-based app as no more than an adjunct; in this respect, the problems with the development of the app might not be as damaging as some have suggested.

The Royal Society also estimated that in combination with other interventions, manual tracing might reduce the number of new infections by 5-15 per cent. The most important question, however, was what might happen without it.

On 19 May, looking forward to more lockdown measures being eased on 1 June, Sage considered models of the “reasonable worst-case scenario” for the future incidence of Covid-19. As before, each of the four modelling groups produced an assessment, and as before, they came out with similar results. The worst that could happen after 1 June was that R would go up to 1.7 for four weeks until the lockdown had been largely reimposed, when it would be back to 0.7. This pessimistic scenario would require “multiple system failures, such as a failure to quickly identify increasing incidence and failure to respond in a timely manner”. At the request of the Cabinet Office, Sage looked at the numbers again two days later. It emphasised that this was exactly what the TTI system was supposed to prevent, but also that less dire scenarios, with R moving to 1.2, would be harder to detect because of a more gradual increase in hospitalisations and deaths.


By and large, the government has stuck with Sage’s caution. This worries those who want the government to do more to revive the economy, hence the suggestions that if the science poses a barrier then it should be ignored. Yet in Sweden, the poster country for those opposed to a stringent lockdown, persistent rises in infections and deaths have led to second thoughts about the wisdom of its more relaxed approach. Those American states that abandoned lockdown early are now seeing new spikes. Even more cautious countries, which have managed the pandemic effectively, are still announcing localised returns to stringency.

At the same time, the health effects of sticking with lockdown would also be severe. Something has to give. The critics have set up as a test case the two-metre social distancing rule. This has been in place from the start, but strict adherence makes any revival impossible for the hospitality and entertainment sectors, among others.

Sage reviewed the issue on 28 April and concluded that the “rule” was appropriate, although less so if the contact was fleeting. But Sir Patrick Vallance, the chief scientific adviser, has recently suggested that it really is not fixed as a rule. The science nonetheless remains awkward. A paper from 4 June by the Environmental and Modelling Group reports one study that puts the risk of two-metre face-to-face contact as around ten times lower than the risk at one metre. Another analysis suggests that it might be closer to half. Risks can be further reduced by making contact shorter in duration, not face-to-face, with face masks, and by practising hand hygiene.

With the exception of this paper, the material available does not cover the past four weeks, and so it does not cover the consequences of the revelations about Dominic Cummings’ trip to Durham. There are no assessments as yet of the confused government communications that have accompanied the move out of lockdown, or the problems faced in getting children back to school.

Without government papers, we only have one side of the story. Yet the trend of policy is clear enough. It is towards “smarter”, less comprehensive measures. The aim will be to try to mitigate the infectiousness of Covid-19,  while finding ways around measures such as the two-metre rule. There will also be a focus on preparing to deal with individual outbreaks, rather than reimposing restrictions across the country (this will be the role of the Joint Biosecurity Centre).

All this is about the management of risk, both for individuals and for society. The question for health is whether people continue to adhere to the guidance still in place. The question for the economy is whether people are ready to take full advantage of reduced restrictions. It is not just the scientists that are holding them back. As we have seen with schools, as individuals are told what they are now allowed to do, they will still need to be persuaded that they can do so safely.

Lawrence Freedman is emeritus professor of war studies at King’s College London