Is the rate going down or staying the same?
Every day we are told about the numbers of deaths and the number of new cases of infection (based on those who test positive) with Covid-19. As interesting as these data are, they do not give any real indication of how likely any of us are to contract the infection and whether the rate of new infections is going down. In this post I will discuss what the current epidemiological studies tell us.
The ‘headline results’ are:
- Approximately 1/3500 people in England in the general population have evidence of being infected currently
- This is down from 1/450 around 8 weeks ago.
- There are probably around 1700 people newly developing the disease every day.
- Over the past month, the rate of new infections is levelling off.
Why the number of deaths is of limited value? Although the figure for the number of deaths provides an indication of the gravity of the epidemic, it gives no insight into the key questions above. A falling number of deaths is an indication of a decline in the number of infections but also of improvements in treatment. This statistic does not really give any robust information on the size of the epidemic because:
- Fortunately only (around) 1-2% of those infected die
- Death certification is not without problems-not easy to distinguish someone who dies directly from the virus as opposed to someone who dies with the virus (eg they were already serioulsy ill with another condition)
- Also (see below) infection rates are higher now in younger people, who are more likely to survive.
Why the number of positive tests is of limited value? People get tested for all sorts of reasons and the threshold for seeking a test is changing. The number of tests performed and the positive rate will tell something about the availability if resources. If the positive rate were falling that would be reassuring but the number of those who test positive cannot be used for any serious determination of risk.
We want to know two things from the information:
- What is the proportion of people in the general popualtion that have the infection (or how likely is it I would be in contact with someone with the virus)?
- What is the rate of new infections, and more specifically as the weeks go by, is the number of new infections going down?
These two measures give us different information but are connected by how long someone is infectious. For example, taking a week as the typical duration, then if there were 10,000 new cases per day, on average each week that would contribute a further 70,000 to the total number of infectious cases (though other cases previously infected ‘drop out’of the total).
Data to answer both the above questions can be obtained from the results of the swab test in large random samples of the population.
What is the proportion of people in the general popualtion that have the infection?
Such data are now available from two national surveys of random population samples, the first from the Office of National Statistics, the second from Imperial College. There is more information about the sources of these data and their limitations at the end of this post.
- Imperial College data, published last week, suggested that in May, 0.13% of the population had the virus (equivalent to 1 in 750)
- The ONS data shows how this rate has recently declined:
- 1 in 400 4-17 May
- 1 in 900 18-31 May
- 1 in 1800 1-14 June
- 1 in 2300 15-28 June
- 1 in 3500 29th June-12 July
- It can be calculated that currently there are 15000 people in the UK who are infected in the general population (eg excluding care homes, hospitals and very local epidemics in factories etc see below)
What groups have the highest risk?
- Similar rates in men and women
- Lowest risk in those aged over 64, highest rate in those aged 18-24 (less likely to be isolating perhaps)
- Those with Asian ancestry had 70% higher rates (after allowing for age)
- Compared to the general population in terms of occupation the risk is:
- 8 times higher in care home workers
- 5 times higher in health workers
- 2 times higher in other key workers
- Variation by region, London highest, South West lowest (Figure reproduced from Imperial data see footnote)
How many new cases are occurring? Based on the ONS data these are the number of new cases infected per day
- 8700 in week of 14 May
- 4800 in week of 21 May
- 2700 in week of 28 May
- 1700 in week of 11 June
- 1100 in week of 18 June
- 1200 in week of 25 June
- 1200 in week of 25 June
- 1500 in week of 2 July
- 1700 in week of 9 July
Comment: Caution: small numbers limit the conclusions when comparing between weeks. The data on new infections have been looked at in a number of different ways as the analysis is complicated and involves a number of assumptions. In the 6 weeks since the end of May there has been at least a halving in the rate of new infections. The rates have been a bit up and down since then, so will have to wait for more confirmation that the rate is continuing to fall
Another way of looking at these data is that each week now there are between 1 and 2 people /10,ooo population who are newly infected
Further information about the 2 studies
- The first is an ongoing survey of 20,000 individuals from 10,000 households who have agreed to provide a swab test every week for five weeks and then monthly to cover a full year. By repeating the survey, this study can tell us both how many existing cases there are and how many new cases develop between the repeat tests This study is organised by the Office of National Statistics and is done in collaboration with the universities of Oxford and Manchester1.
- The second is a larger study being conducted by Imperial College London2 This involved 120,000 individuals from across England. They compared the rates of infection recorded during May and this gave an estimate of any change over that month. With the larger numbers they were able to look at effects of region, age and employment type as well as the relation to symptoms. They are planning to repeat the study.
There are limitations to both these studies which may affect the results:
- It is always very difficult to get everyone who is approached to agree to participate. In the Imperial study, of those approached only about 30% agreed to take part who thus may not be representative of the whole population. As an example, it is possible that those with the greatest social advantage were more willing to participate. If they were also more likely to adhere to social distancing then they may have lower rates of infection.
- The studies relied on the participants to collect their own samples. As those who have had a swab taken know only too well, it is not an easy (or pleasant!) task to scrape the back of the throat. It is possible that if not deep enough then there may be tests that came back negative which should have been positive
- Although these are large studies, because of the low rates of infection, the actual number of people who tested positive were small. These numbers are then multiplied up to give estimates for the whole population. For example, in the latest ONS study there were only 40 people who had positive tests in the last 6 weeks. When these are divided into age, sex or geographical areas the numbers get very small indeed. The researchers are very aware that the estimates they provide are subject to imprecision and thus they give a range where the ‘real’ ie whole country figure might lie. For example for mid week around 9th July, ONS estimated there are 1700 new cases per day in England, but that is based only 4 actual new cases. ONS then calculated that the true figure probably lies between 700 and 4200 new cases a day-a big range!
Who do these numbers apply to?
Both these studies have aimed to study the ‘normal’ population ie people in the general outside world. They deliberately do not cover people in care homes for instance and would also miss out on concentrated local outbreaks, such as those reported in factories. Thus, when they report the number of new cases in the population they only refer to those arising in the general population
A positive test does not mean infection
The swab test detects the presence of genetic material from the virus which does not necessarily mean that someone is infectious. We know that people can test positive for some weeks after being infected whereas they are likely to be infectious perhaps for only a week to a likely maximum of two weeks after infection. This affects the interpretation of the number of existing cases with infection, whereas obviously the estimates of the number of new cases are not affected.