• Following measures by the UK Government, a survey was conducted on the 18 of March to assess public attitudes. 77% of respondents were worried about an outbreak and while 93% reported taking protective measures, only 50% were avoiding social events, 36% were avoiding public transport, and 31% were avoiding going out.
  • A study on the global impact of COVID-19 estimated that an unmitigated epidemic would infect 7.0 billion out of the world’s 7.8 billion people. This would lead to 40 million global deaths in 2020.
  • The latest modelling estimates that as of 27 April about 4% of the population of the UK has been infected with coronavirus.
  • The UK has strengthened capacity of the NHS to deal with COVID-19 by building field hospitals, but there is still a shortage of intensive care beds and intensive care nurses.
  • Various testing strategies are being explored for healthcare workers and the wider community. Testing each case and their contacts might require as many as 60,000 tests per day.
  • This breakdown of the Imperial College models is part of our rapid response content on COVID-19. You can view all our reporting on this topic under COVID-19

The first wave of documents released by SAGE went up to 16 March. By 29 April, there had been no further releases of the underpinning evidence. Assuming the Imperial College COVID-19 Response Team is still advising, their later work is summarised below.


Public response to UK Government recommendations on COVID-19: population survey, 17-18 March 2020

This is a summary of a YouGov survey that was commissioned by Imperial College London. On 16 March, public health measures were introduced in the UK. These included asking the public to:

  • stop non-essential contact with others.
  • stop all unnecessary travel.
  • work from home where possible.
  • avoid social venues.
  • isolate for 7 days if unwell.
  • isolate for 14 days if household member is unwell.

The survey was conducted online on 17 and 18 March and 2108 adults answered the following:

  • 77% were worried about the COVID-19 outbreak.
  • 48% thought it is likely they will be infected at some point.
  • 93% reported that they were taking measures to protect themselves:
    • 83% were washing hands more frequently.
    • 52% were avoiding crowded areas.
    • 50% were avoiding social events.
    • 36% were avoiding public transport.
    • 31% were avoiding going out.
    • 11% were avoiding going to work.
    • 28% were avoiding travel outside of the UK.
  • 88% confirmed they would self-isolate for 7 days if advised to by a healthcare professional.

Overall, 71% reported that they had changed their behaviour in response to guidance. For 18 to 24 year olds this figure dropped to only 53%. Hand-washing (63%), avoiding ill people (61%) and covering your face when sneezing (53%) seen as very effective.

Not going out (31%), to work (23%), to shops (16%) or to schools (19%) were less likely to be perceived as very effective. Only 44% of respondents said they would be able to work from home.

Evidence of initial success for China exiting COVID-19 social distancing policy after achieving containment

This non-peer-reviewed study was published on 24 March. By 20 March there had been over 254,000 cases and 10,000 deaths globally. When this study was published, public health measures in China had reduced the number of new cases to zero. To assess these measures, the authors examined the virus transmissibility compared with within-city movements.

Within-city movements were used as a proxy for economic activity. They were measured relative to the average travel in 2019 when about 5 daily trips per person was normal. Therefore if only 3 daily trips are made per person, this would correspond to a 40% drop in movements. R0 (average number of people infected by each infected person) was used for transmissibility.

As movement restrictions were put in place (late January to early February), movements and R0 were closely associated. For example, in Hubei fewer than 1 trip was being made per day. and R0 dropped from 6 to less than 1. However, as movement started again, this association was not maintained. This suggests that after very intense social distancing, moderate relaxation of restriction may be possible. This may be due to other containment measures being undertaken in China, such as testing and contact tracing. In Hong Kong, by contrast, movement and R0 were not related. Hong Kong was therefore able to maintain economic activity while still containing COVID-19.

The global impact of COVID-19 and strategies for mitigation and suppression

This was a non-peer-reviewed study looking at the impact of COVID-19 in 202 countries. The authors combined data on people’s contact with each other at different ages with COVID-19 severity. The mortality and healthcare demand estimates were based on high-income countries. Therefore these may be underestimated where health systems have less capacity.

Factors included in the model

A number of factors may influence the impact of COVID-19 on a country’s population. First, there is a strong link between the GDP of a country and its age structure. For example, higher income countries tend to have the oldest populations.

Co-existing disease patterns also vary. In low- and low-middle income countries there is a greater frequency of infectious diseases. In middle-income countries there are fewer infectious and more chronic diseases.

Low-income countries also have larger multi-generational households – the average size of households where one person is over 65 years old is much higher. In some countries the average household size is 13, which can lead to many more contact exposures for older people. This increased mixing of age groups in low-income countries may lead to higher rates of infection.

Finally, healthcare capacity varies substantially between countries. Average numbers of hospital beds are:

  • 1.24 beds per 1000 people in low-income countries. 1.63% of these are critical care beds.
  • 4.82 beds per 1000 people in high income countries. 3.57% of these are critical care beds.

Scenarios considered in the model

Four potential scenarios were explored:

  1. Unmitigated epidemic: no actions are taken.
  2. Mitigation: social distancing measures are implemented for the whole population but with enough contact for herd immunity (45% reduction in contact rates).
  3. Mitigation: social distancing measures are implemented for the whole population and over 70-year-olds are shielded.
  4. Suppression: intensive social distancing (75% reduction in contact rates).

The authors assumed these measures would be alongside testing and contact tracing. However, these interventions are not included in the models.

They estimated that an unmitigated epidemic would infect 7.0 billion out of the world’s 7.8 billion people. This would lead to 40 million global deaths in 2020. Low-income countries would see lower numbers of severe infections and deaths due to the younger average age of their populations. However, the authors assumed the same quality of healthcare in low-income countries and the same levels of coexisting diseases. In reality there are likely to be many more deaths in low-income countries than modelled.

Strategies including social distancing and shielding of the elderly could halve deaths to 20 million. In high income countries, demand for critical care beds would be sevenfold higher than capacity. In low income countries this may rise to 25-fold higher demand than capacity. While not modelled, this would lead to a high degree of excess mortality. This is because the health systems would likely fail in low-income countries under such strain.

Use of a suppression strategy will be most effective if used early. Implementing the strategy when death rates are at 0.2 per 100,000 people per week could save 38.7 million lives. If delayed until 1.6 people are dying per 100,000 then only 30.7 million lives will be saved. However, the authors do not account for the wider social and economic costs of suppression.

Estimating the number of infections and the impact of non-pharmaceutical interventions on COVID-19 in 11 European countries

In this report, the previous assessment of non-pharmaceutical interventions in the UK was expanded to Europe. Models assume that each intervention will have the same effect on transmission across countries and time periods. This means that the models are driven by countries that started measures early.

The aim of the measures is to reduce R0 (average infections per case) to less than one. If this happens then each case will, on average, infect less than one other person. This would stop the virus from spreading.

Testing in most countries is focused on severe cases, particularly those that are hospitalised. This makes R0 difficult to estimate. The authors used the observed deaths to estimate R0. However, this creates a time-lag as there is a 2 to 3 week gap between infection, symptoms and then death.

Italy was the first European country to take measures. Most measures began around 12 to 14 March. The authors examined the data on deaths up to 28 March which provides insights on the situation 2 to 3 weeks before that. Figure 1 from the report shows when different interventions started in different countries.

A figure showing when european started taking measures, and what those measures where. Switzerland advised for case based self isolation on 2 March. Italy ordered school closures on 5 March. On 9 Mrch Italy banned public events, encouraged social distancing and on the 11th ordered lockdown. Other countries started enforcing strict measures such as banning public events from 12 March onwards. In the UK, harsh measures such as the closure of schools, didn't come into effect until 21 March.
Intervention timings for the 11 European countries included in the analysis. Italy was the first European country to take measures beyond self isolation. Most measures began around 12 to 14 March

Cases and deaths

The authors estimated the numbers of cases in each country for 28 March. Italy had the most infected people, at 5.9 million (9.8% of population). Spain experienced a large increase in numbers of deaths, suggesting 15.0% of the population had been infected. Germany had one of the lowest rates of infection at 600,000 people (0.7%). These figures have all since been revised as more data have become available. For Italy, Spain and Germany the rates are now (27 April) 4.34%, 5.38% and 0.83%. The highest rates have been seen in Belgium (11.69%) and Sweden (6.33%).

The model has been put online and is updated on a daily basis. For 27 April, Figure 2 below shows the daily number of infections and deaths. By working backwards, this suggests that about 4% of the population of the UK has been infected.

Daily number of infections see an exponential rise until 23 of march were there's a sharp drop and gradual flatening. Daily number of deaths exponentially rises up till aproximately the 13th of April. After that there's gradual flattening.
Daily infection and death in the UK as of 27 of April. Source Imperial College London

Impact of interventions

The authors produced an initial R0 (average infections per case) of 3.87 averaged across the countries. However, this did change in response to the non-pharmaceutical interventions.

On 30 March, R0 ranged from 0.97 in Norway to 2.64 in Sweden. By 24 April R0 was estimated to be less than 1 for all countries except Belgium and Sweden. However, the figure for Sweden may appear artificially high. This is because, by 30 March, a total lockdown had not been ordered and so the model estimated R0 to be higher. At the time of publication, the authors were unable to conclude that measures had pushed R0 below 1 in any country. However, subsequent analyses suggest that this has now happened.

The following figure shows the change in R0 in the UK by date. Since 23 March the model estimates that R0 has been consistently below 1.

A short drop in Rt is observed from 23 March onwards
Rt is the variation of reproduction number R0 over time. Source Imperial College London

This is leading to a forecast gradual decline in the daily numbers of death shown in the next figure.

The logarithmic cure starts to flatten in late March/early April.
Forecast changes in deaths in the UK after 27 April. The lag between deaths and infections means that it takes time for information to propagate backwards from deaths to infections, and ultimately to Rt. A conclusion of this report is the prediction of a slowing of Rt in response to major interventions. To gain intuition that this is data driven and not simply a consequence of highly constrained model assumptions, we show death forecasts on a log-linear scale. On this scale a line which curves below a linear trend is indicative of slowing in the growth of the epidemic. Source Imperial College London

Many of the interventions were introduced close to each other in time. This has made it difficult to estimate their relative impact on R0. The final figure shows the latest estimated impact of each intervention. Lockdowns are thought to have had the greatest impact; a 65% reduction in R0. School closures may have had the least impact; an average reduction of 5%.

The figure shows the effect of intervention on transmission of coronavirus. The  most impactful intervention is Lockdown enforcement, with the banning of Public events in second place.
Estimated impacts of different interventions on transmission. Mean relative percentage reduction in Rtis shown with 95% posterior credible intervals. If 100% reduction is achieved, rt =0 and there is no more transmission of COVID-19. Source Imperial College London.

Strengthening hospital capacity for the COVID-19 pandemic

In this non-peer-reviewed report, the authors developed a pandemic planner for hospitals. This calculates how much capacity is achieved through different interventions. The tool assesses baseline capacity for beds, staff and ventilators. It then uses data from 12 European countries to estimate the impact of the interventions.

Baseline capacity

Baseline capacity focuses on hospital resources for COVID-19. It considers numbers of staff and their critical care skills, numbers of beds, and ability to offer breathing support.

Interventions to increase capacity

18 interventions were identified across 12 European countries. 13 of these were about increasing or reorganising the provision of care, such as:

  • rapidly constructing field hospitals.
  • repurposing general acute beds into critical care beds.
  • deploying newly qualified and final year healthcare students.

Five interventions were about managing admissions to care. These interventions require cancelling planned non-urgent, non-cancer elective surgeries and other trade-offs.

Case study: the NHS

A number of measures were taken to prepare the UK for COVID-19. For example the construction of the Nightingale field hospitals added 500 critical care beds and 8000 general acute beds to the NHS.

On 3 April there were 1,480 COVID-19 patients in critical care beds and an estimated 7,700 COVID-19 patients in general acute beds.

It was estimated that at that point, there would be 1.46 spare general acute beds per 10,000 people in the UK. This is compared to 0.03 spare beds before the construction of the field hospitals. But there would still be a shortage of 0.05 critical care beds per 10,000, despite an improvement of capacity of 64%.

But beds are only part of the solution. Before the epidemic there was a deficit of 500 critical care nurses. The construction of the Nightingale field hospitals has further created the need for another 500 critical care nurses. Upskilling could reallocate 647 nurses from general acute to critical care.

Role of testing in COVID-19 control

In this non-peer-reviewed study the authors develop a simple model to compare different testing strategies for COVID-19 control. This was driven by the observation that countries that have tested more have controlled the outbreak more quickly. There is a clear priority to target testing to patients in hospital. This enables timely treatment and infection control. However, it is important to identify what the impact of widespread testing could have on control.

The authors investigated PCR and point of care tests that identify active infection. They also examined testing strategies for when an antibody test demonstrating past infection is ready.

Testing healthcare workers

High levels of COVID-19 have been seen in healthcare workers. Available figures range from 4% of total COVID-19 cases in China to 19% in Spain. The Royal College of Physicians found that 11% of their members and fellows were self-isolating in March. 59% of them had symptoms themselves and 41% had a household member with symptoms.

Testing symptomatic or self-isolating healthcare workers

The current UK strategy is to test all healthcare workers with symptoms. However, there has been growing evidence that up to 50% of infections have minimal symptoms. The effectiveness of self-isolation depends on how much transmission occurs before symptoms develop. But up to 40% of transmission may occur before obvious symptoms appear. Therefore if healthcare workers self-isolate as soon as symptoms start they could reduce transmission by 16–57%.

It has been suggested that testing healthcare workers and their families may allow them to get back to work sooner. Current policy is for people with symptoms to self-isolate for 7 days. If family members are unwell, then isolation should be for 14 days.

With current test accuracy, the proportion of infected people that test positive is 80–90% (this is referred to as the sensitivity of a test). To increase accuracy, two tests could be carried out to ensure healthcare workers are negative before going back to work. But time constraints mean that, until more rapid testing is available, testing is unlikely to increase the speed with which healthcare workers get back to work.

Testing might be more useful if healthcare workers are self-isolating due to an infected household member. Testing the family member and finding that they are negative would be very useful in getting healthcare workers back to work. In 2003, in Hong Kong for SARS, all healthcare workers in the same environment as someone infected were tested at the same time. A similar strategy could be used for COVID-19.

Regular testing of healthcare workers

Another option would be to regularly screen healthcare workers with a PCR test. This would pick up workers who are infected but with minimal symptoms. The effectiveness of such an approach depends on the accuracy of the test, how often testing is and how quickly results are available.

The authors’ model suggests that weekly screening could further reduce transmission from healthcare workers by between 16% and 33%. This would be in addition to the 16–57% reduction in transmission from self-isolation.

At the end of 2019 there were 35,000 NHS staff working in intensive care, infectious diseases or respiratory medicine. Therefore to test weekly, 5000 tests will be required per day.

Immunity passports

Knowledge of a previous positive PCR test may mean that healthcare workers are less likely to be reinfected in future. If antibody testing becomes available it could determine immune status. Immune healthcare workers could then work in higher risk areas. The authors did not fully model immunity passports and these will be revisited in a later article. However, for now it is worth noting that there are some issues with them:

  • accuracy of tests may mean that people test positive (suggesting they are immune) when they are not. This risks non-immune people being given an immunity passport.
  • no one yet knows how long immunity lasts or what the best time to test for it is.
  • there are also ethical considerations such as discrimination and people actively seeking infection

Testing in the community

The testing of symptomatic individuals in the community will have less benefit as they should already be self-isolating. Community testing is more likely to be of use once contact tracing is reestablished. Once contacts of a case are identified early testing would help identify those with minimal symptoms.

Testing will become more important as COVID-19 cases go down and other respiratory viruses come back in winter. Testing in these cases will help exclude COVID-19 as a cause of any symptoms. This would release people from isolation earlier and avoid unnecessary contact tracing.

Between 18 March and 19 April, about 15,000 calls per day were made to NHS 111 from people with COVID-19 symptoms. Before lockdown, the average number of daily contacts for each person was 11. This reduced to three during lockdown. Therefore testing each case and their contacts would require 60,000 tests per day. During a typical influenza season, reported numbers are 2400 per day. This would require over 29,000 tests per day.


You can find more content from POST on COVID-19 here.

You can find more content on COVID-19 from the Commons and Lords Libraries here.