Cookies on this website

We use cookies to ensure that we give you the best experience on our website. If you click 'Accept all cookies' we'll assume that you are happy to receive all cookies and you won't see this message again. If you click 'Reject all non-essential cookies' only necessary cookies providing core functionality such as security, network management, and accessibility will be enabled. Click 'Find out more' for information on how to change your cookie settings.

Emma Pritchard, Statistician and PhD student shares her work with ONS on Coronavirus statistics.

Developing a process to monitor populations at increased risk from coronavirus in the community


You may have heard the saying “canary in the coal mine” before. It is often used to express that something is an early sign of danger. The saying originated when caged canaries were sent into coal mines. If there was methane gas present, the birds would keel over and this would act as an early signal for miners that there was danger….and they should get out quickly!

We wanted to see if we could design our own “canary” which we could use to tell us where, and in whom, coronavirus was the rising fastest. So we set out to develop a process (i.e. make a canary) we could use to identify any characteristics which could act as an early warning signal for rising coronavirus cases in the community. We could then use these warning signals to target testing in certain groups of people or to inform public health policy, such as offer advice on current coronavirus restrictions (getting the miners out of the mine!).

The miners knew that the threat of methane gas may change over time, so they often took the canaries down the mine with them. Similarly, we wanted our process to be easy to use on a regular basis as the risk of catching coronavirus can change over time in different groups of people. This can happen for lots of reasons but a couple of examples could include new variants appearing or lockdowns being enforced. We therefore set out to design a process that we could use each week to alert us of groups of people in the community who were more likely to test positive for coronavirus.

How we did it

We used data from the Office for National Statistics COVID-19 Infection Survey. This is a large survey of individuals living in the community. Participants get visited regularly by study workers who supervise a swab test to check for coronavirus particles in the nose and throat. When the swab is taken, participants are also asked about:

  1. Demographics e.g. age, sex, ethnicity,
  2. Work and employment e.g. where do you work? how do you travel to work?
  3. Health e.g. have you been vaccinated against coronavirus? Do you have any long-term health conditions?
  4. Social behaviours e.g. how often do you socialise outside your home?

Altogether, there were 60 characteristics across these four groups! We designed our process to check every fortnight which groups of people, based on the 60 characteristics, were more likely to have coronavirus.

To assess how well our process worked, we took one year of data from the survey (July 2020- July 2021) and chopped it up into fortnightly chunks. In each of these chunks we had information from all participants who have a visit from a study worker within that fortnight, including the result from the swab test and answers to all questions asked about their demographics, work/employments, health, and social behaviours. We implemented our process in each of these fortnights and summarised which groups of people were more likely to test positive at that time. For a detailed description of the process, see the paper linked at the end of this blog. 

What we found

We found that our process picked out groups of people who were more likely to test positive for coronavirus in each fortnight. Some of these characteristics were persistent across all the whole pandemic, whereas some came and went as the pandemic changed, and others rarely appeared important at all.

The paper linked at the bottom of this blog summarises which of the 60 characteristics were most important in each fortnight over the last year. We don’t have time here to talk about all 60 characteristics, so we’ve summarised five interesting ones in a bit more detail below….

5 important characteristics driving coronavirus positivity

1)    Geographical region

Where people lived in the UK had a big effect on whether they had coronavirus. During September and November 2020, we saw more cases in Northern areas of England. When the Alpha variant began to circulate in December 2020, we instead saw an increased chance of having coronavirus if you lived in Southern areas of England, and especially in London.

2)    Vaccination

The vaccination roll-out programme began in December 2020 and we quickly saw that people who had been vaccinated were less likely to test positive for coronavirus. We have seen this in every fortnight since then!

3)    Social distancing at work

Lots of jobs require staff to work away from home, and some jobs require you to be closer to colleagues/patients/clients than others. When coronavirus cases were high across England (mid-September 2020 to February 2021), we saw that those working outside of home were more likely to test positive for coronavirus, compared with those working from home. Additionally, those in jobs where social distancing was hardest were even more likely to test positive for coronavirus.   

4)    Travel Abroad

Those who were lucky enough to jet off for a holiday abroad in July, August, September, or October 2020 were also more likely to test positive for coronavirus soon after they returned, compared with people who stayed in the UK.

5)    Sex

We didn’t see any difference in how likely you were to have coronavirus between men and women during most of the pandemic until June 2021 where we saw that men were more likely to test positive than women. If you remember back to June 2021 (or if you’re an England football fan, maybe you’d rather forget…), this coincided with the European Football Championship (EUROs) where men may have been engaging more in social mixing compared with women. 


We managed to develop a process which we could use every two weeks to alert us of rising coronavirus cases in different groups of people. This process is now being used regularly to check which characteristics are driving rising cases of coronavirus in the community.

Link of full academic paper:

Link to the Office for National Statistics technical article showing recent results: