The government’s plan to ease the lockdown will be confirmed in an official review that Downing Street expects will give the all clear for schools to begin reopening next week. No 10 said the proposed steps for England should be outlined at the daily press conference on Thursday, though they are dependent on further scientific advice, including on the rate of transmission.
Easing the lockdown is a balancing act. The optimal strategy relaxes the right restrictions by the right amount to allow some return to normalcy without risking a second wave of infections. But because a swathe of measures were introduced in rapid succession, from physical distancing to school closures, it is hard to know what impact each had, and so what risks are associated with easing them.
As restrictions lift, experts are poring over data for signs that infections are rising. The number of new positive test results is clearly important. But regular contact surveys, Facebook and Google location data, workplace absences and calls to NHS 111 all provide an early warning. The data are analysed by the government’s scientific pandemic influenza group on modelling (SPI-M) which reaches a consensus each week on the R value, the average number of people an infected person infects. This estimate goes to the Scientific Advisory Group for Emergencies (Sage) for formal endorsement. If R rises above 1, the disease is bouncing back.
The new Joint Biosecurity Centre announced earlier this month aims to monitor the prevalence of the disease in England and provide real time analysis of outbreaks at the community level so that infected people and their contacts can be swiftly identified and isolated. But what sorts of data do scientists use?
Researchers conduct weekly online surveys of the UK population to gather details on how people mix and so how the R rate might be changing. The data is fed into a grid or “contact matrix” which shows, for example, how often a typical 25-year-old spends time with an elderly person or someone of the same age. These grids link directly to the R value of the outbreak. Because scientists performed similar surveys in the past, they can compare contact matrices at different times and work out whether the epidemic is growing or shrinking. In May, a team at the London School of Hygiene and Tropical Medicine used contact surveys to show how lockdown slashed the average number of contacts from 10.8 to 2.8, sufficient to reduce R from 2.6 to 0.62. One benefit of contact surveys is speed: they give an instant snapshot of how people’s behaviour changes as restrictions ease. The NHS contact-tracing app should be extremely valuable too as it can potentially record contact information for millions of people.
When people use Facebook with location services on, the app regularly records their longitude and latitude. In emergencies, Facebook provides anonymised co-location data through its GeoInsights portal which researchers use to see how much people go out, and how people from different areas are mixing. Google is using its own location data to produce regular, local “Covid-19 community mobility reports” that reveal trends in movement around places such as workplaces, shopping centres, cafes, pharmacies, parks and beaches. Public data on road traffic and public transport are also feed into models for calculating R.
R, or the ‘effective reproduction number’, is a way of rating a disease’s ability to spread. It’s the average number of people on to whom one infected person will pass the virus. For an R of anything above 1, an epidemic will grow exponentially. Anything below 1 and an outbreak will fizzle out – eventually.
At the start of the coronavirus pandemic, the estimated R for coronavirus was between 2 and 3 – higher than the value for seasonal flu, but lower than for measles. That means each person would pass it on to between two and three people on average, before either recovering or dying, and each of those people would pass it on to a further two to three others, causing the total number of cases to snowball over time.
The reproduction number is not fixed, though. It depends on the biology of the virus; people’s behaviour, such as social distancing; and a population’s immunity. A country may see regional variations in its R number, depending on local factors like population density and transport patterns.
Hannah Devlin Science correspondent
When patients call NHS 111, or see their GP with Covid-19 symptoms, the data is recorded and made available to outbreak modellers. Staff absences from work, in particular the NHS, are logged and can help reveal outbreaks in the community before testing has been done. The Zoe app developed by King’s College London to track the disease gathers details on users’ health and can potentially detect outbreaks through sudden spikes in people reporting Covid-19 symptoms.
The expansion of virus testing means more people can now get tests should they develop symptoms. But many people are asymptomatic or develop only very mild disease. To get a clearer picture of disease levels, several groups are collecting snapshots of infection rates in the community. Researchers at Imperial College have recruited a random 100,000 people to take swabs which are tested for virus in the lab to provide an accurate picture of how much infection is across the country. The testing can be repeated at a future date as needs require. Separately, the Office of National Statistics is testing thousands of households for coronavirus to monitor levels of disease. Antibody tests, which reveal if a person has been infected in the past, will also feed in.
If hospitals record a sustained uptick in confirmed Covid-19 patients then infections are on the rise. But other factors can influence patient numbers. During the outbreak, patients seeking emergency treatment slumped as people avoided hospitals. As lockdown eases and GP services return to normal, more Covid-19 patients may visit their GPs and be sent to hospital to be checked out. Other data, including admissions to intensive care, and deaths in hospitals, care homes and the wider community also feed into outbreak modelling. The downside of data like these is that they reflect the size of the epidemic some time ago. An infected person may be admitted to intensive care a month after being exposed to the virus. For this reason, scientists plug the data into their models and run them forwards, a process called “nowcasting”, which effectively predicts to the present rather than the future.