The city’s director of public health has said he has serious concerns over an app which claims the rate of coronavirus infections is rocketing in Brighton and Hove.
The Deckzero website compiles regional new confirmed cases and uses them to calculate the R number – an estimate of how many people each infected person passes the virus on to – in each area.
But Alistair Hill, director of public health for Brighton and Hove City Council, says the numbers of new infections in Brighton is so low that it’s impossible to calculate an accurate R number.
Last week, the app calculated the R number for Brighton and Hove was 1.68, the fourth highest figures in the country – but this was based on a rise of confirmed cases of just eight, from 10 to 18.
He says he expects the rate calculated by the app to rise in the coming weeks purely because results from the Amex testing centre will begin to be included in official figures
And tests – and therefore confirmed cases – will also ramp up when the test and trace system begins in earnest.
Mr Hill said: “We have very serious concerns about the deckzero.com website’s information about the R number.
“Quite simply, the alleged statistics they are using are misleading and potentially dangerous in terms of giving either false reassurance or creating unnecessary concerns.
“Calculating the R rate is very complex. It involves data from lots of different sources.
“The calculations include information relating to such things as confirmed cases, hospital admissions, deaths, patient surveys and testing.
“When these are all put together and interpreted by experts they give regional and national R numbers.
“Robust calculations of the R number come from Public Health England and the University of Cambridge and also from the London School of Hygiene and Tropical Medicine.
“It is not possible to calculate meaningful R values at a very local level, such as one specific for Brighton and Hove.
“This is because some of the data involves such low figures that even a small change in them can, in statistical terms, make a disproportionately big relative difference.
“For example, the increase from 1,000 and 1,009 is less than 1%. But the difference between 10 and 19 is still only 9 cases, but is a 90% increase.
“Locally we are reviewing early warning indicators on a daily basis. They include:
- Data on confirmed cases from Public Health England
- Data on the numbers of people contacting 111 and 999. This has been shown to be a good predictor of deaths with Covid 16 days later
- Data on individuals in hospital who have tested positive
- Data on outbreaks in community settings (for example care homes)
- Data on the number of Covid deaths
“To date, the rate of diagnosed cases has been relatively low compared to the national picture. The number of 111 and 999 calls and hospital admissions for Covid all currently remain low.
“However, there has been a small increase recently in the number of cases. We and our NHS colleagues are monitoring this situation very closely.
“We also know that data on tests at commercial laboratories, so for example those being tested at the Amex site, are not within the data made available and once this is included we will see an increase in cases due to more people being tested.
“In addition, in the coming weeks data from the national Test and Trace system will start to feed into these calculations. This system will start to pick up cases that aren’t currently being tested for.
“It is very likely that because of this there will be a rise in both local and national figures when these data start coming through.
“I’d also like to reassure you that we’re doing everything we can as a council to support residents and continue to deliver excellent services through the pandemic.
“We would urge all our residents to continue to take responsibility and respect the restrictions around social distancing. This is the best way for everyone to protect themselves and their loved ones.”
Editor’s note: On 12 June, Deckzero asked us to add the following statement to the story
Some of the initial reporting of Deckzero’s Covid 19 modelling was inaccurate and out of context.
Brighton and Hove Council has now withdrawn its initial statement and instead points to issues with the reporting.
The original disclaimer on the Deckzero website reads: “RZ Index is provided to indicate and signal changes in new cases, rather than to replicate the scientific R number. It is not official or government guidance and shall not be used in any scenario or situation where the government guidance should be followed, including, but not limited to policy-making by local organisations.”
Had all parties read this disclaimer we are sure the subsequent confusion would have been avoided.
Quite apart from the virus issue we have the problems that, firstly, statistics can be massaged in just about any way you choose (the old saying applies that “there are lies, d**n lies and statistics”), and then the problems that social media distributes these statistics as fact in double quick time with lots of people taking a single source of information as gospel truth and spreading it even further.
Yes indeed. Just look at the double counting of the number of tests and the number of people having those tests.
Carrying 100k tests isn’t the same as 100k people having tests.
Both are discerte but different numbers yet some are using them interchangably
And then you have the Government including tests posted out to people in the tests carried out figures…
It appears that not enough people want tests. Is that good or bad?
indeed, facts are in short supply but the lock-down goes on and anger/fear-factor jumps like a germ =
Id like to know if all of the 444 people reported to have the virus in Brighton and Hove have symptoms? Were they tested because they had symptoms or for other reasons. Are they young? Old?
If 70% of cases tested generally, do not show symptoms, does this mean that the true number of cases in Brighton and Hove is more likely to be 1200 or thereabouts?
Office for National Statistics – deaths registered in Brighton and Hove
Week ending – Number of deaths by location
Friday 20 March – 1 in hospital
Friday 27 March – 5 in hospital and 1 in a care home
Friday 3 April – 12 in hospital, 3 in care homes and 1 elsewhere
Friday 10 April – 15 in hospital, 5 in care homes and 1 elsewhere
Friday 17 April – 15 in hospital and 10 in care homes
Friday 24 April – 7 in hospital, 7 in care homes and 5 elsewhere
Friday 1 May – 4 in hospital, 8 in care homes and 2 elsewhere
Total 102 (59 in hospital, 34 in care homes and 9 elsewhere)
I think that antigen testing has been primarily restricted to patients in hospitals and care homes and also healthcare workers for some time now, and only those with symptoms. Try this link
https://covid19.healthdata.org/united-kingdom
You can see the estimated infections are in keeping with the deaths – accounting for the lag between infection and outcome. If you can use our local data and the maths from these graphs, you could make your own set of graphs and get an idea of the estimated infections in Brighton and Hove!
In the article I read, the author explained clearly that the small numbers made the calculation unreliable.
I suppose the R number is also influence by the test turnaround time. Mine was 8 days, this was at Amex, run largely by school-leavers, by the look of it. I had to test myself, they’re all self-tests, was unable to find my tonsils – I definitely have some – so just did the nasal swab.
If I’d had the virus (I didn’t), I would have had more than a full week to go round infecting people in my environment before I realised what I was doing.
My current understanding is if you have symptoms, assume you are infectious, as the tests are so unreliable.
Just out of interest. If the test results from Amex are only just now being included in the Brighton statistics what happened before ? Were they being included in any other official data? And is this similar in many other testing stations ?