Lockdown is going to bankrupt all of us and our descendants and is unlikely at this point to slow or halt viral circulation as the genie is out of the bottle. What the current situation boils down to is this: is economic meltdown a price worth paying to halt or delay what is already amongst us?
Tom Jefferson and Carl Heneghan, Centre for Evidence-Based Medicine, March 30th 2020
This is probably the most significant ‘known unknown’ when it comes to trying to understand the crisis and work out how best to respond. Simply put, if the Imperial College modelling by Professor Neil Ferguson and his team is correct and only 3-5% of the UK population has been infected, that’s a powerful argument for prolonging the lockdown. If we start relaxing social distancing measures, tens of millions of people will become infected, the NHS will quickly be overwhelmed and hundreds of thousands will die. This was the assumption built into the March 16th model which estimated the death toll at 510,000 if we took no precautions, 250,000 if we followed a mitigation strategy and 20,000 if we moved to a suppression strategy. In effect, the lockdown is preventing 230,000 unnecessary deaths, although that’s an underestimate if Professor Ferguson’s model is right because his 250,000 figure assumed that all those requiring critical hospital care would receive it when, in fact, the demand for critical care in the mitigation scenario would be eight times greater than the NHS’s emergency surge capacity. And even if we inflate that 250,000 to allow for this, that still doesn’t account for the total number of deaths that pursuing a mitigation strategy would result in because it doesn’t include the increase in the number of people dying from other diseases because the NHS would be overwhelmed.
But what if Professor Ferguson has underestimated the number of people who’ve been infected? A paper written by a team of scientists led by Professor Sunetra Gupta at Oxford University published on March 24th included a range of estimates of the percentage of the UK population that has already been infected, one putting it as high as 68%. (This was widely reported as a claim that half the UK population may have already been infected.) If that’s true, it suggests we’re well on our way to acquiring herd immunity and if we end the lockdown tomorrow the NHS will be able to cope, particularly as it has over 2,000 vacant intensive care beds compared to about 800 before the crisis. As of April 13th, 290,720 UK citizens have had swab tests, of which 88,621 were positive, or about 30%. True, this isn’t a representative sample, but against that some of the people tested will have been negative because they’ve already had it. In general, the fact that only a small minority of the population has been presenting with symptoms doesn’t mean a majority haven’t been infected because data out of China suggests four-fifths of those who get COVID-19 are asymptomatic. (Patrick Vallance, Chief Scientific Advisor to the British Government, thinks the real figure is likely to be closer to 30%.)
The Oxford paper was criticised on the grounds that many of the assumptions made by Professor Gupta were “speculative” and had no “empirical justification”, but the same is true of the Imperial model. The FT’s Jemima Kelly said Oxford’s research should be taken with a large dose of salt because it was “not yet peer reviewed”, but Imperial’s paper hasn’t been peer reviewed either. As John Ioannidis, professor in disease prevention at Stanford University, has pointed out, some of the major assumptions and estimates that are built into the Imperial model “seem to be substantially inflated”. But others are much more sceptical, such as Gregory Cochran, who argues that half the UK population cannot possibly have been infected since, if they had, you’d expect the percentage of people testing positive after being swabbed to be far higher. What if they’ve already had it and flushed it out of their systems? Cochran thinks that’s implausible because the virus is so new.
One of the reasons it’s so important to accurately gauge how many people have been infected is because without knowing that we don’t know what the infection fatality rate (IFR) is. That’s different to the case fatality rate (CFR), which is the number of people who’ve tested positive divided by the number of deaths. The CFR varies enormously from country to country. In Italy, for instance, it’s 11%, while in Germany its 0.79%. In Iceland, which has carried out more testing per capita than any other country (it only has a population of 364,260) it’s 0.2%, just above seasonal influenza. In the UK, the CFR is around 9%. But it’s a safe bet that the IFR, whatever it turns out to be, will be significantly lower than the CFR. If it turns out that 30% of the UK population has been infected and 20,000 people end up dying, that’s an IFR of 0.12%, or just above the IFR of seasonal flue. Knowing the IFR matters because we won’t know how much demand there’ll be for critical care in the NHS if we relax the social distancing measures until we know both what percentage of the population has been infected and what the IFR is. We should start to build up a more accurate picture of both once we start doing large scale serological testing – something like an opinion poll, i.e., a large, nationally representative sample of the UK population. A team at the University of Bonn tested a randomised sample of 1,000 residents of the town of Gangelt in the north-west of the country, one of the epicentres of the outbreak in Germany, and found that 15% either were or had been infected, yielding an IFR of 0.37%. For what it’s worth, Oxford’s Centre For Evidence-Based Medicine (CEBM) estimates the IFR to be between 0.1% and 0.26%.
In the US, research released on April 17th by Dr Ioannidis of Stanford University on actual infection rates in Santa Clara county using a serology approach to test for antibodies on over 3,300 residents suggests that the number of people actually infected is a staggering 50 – 85 times higher than the 956 cases that have been documented (see video link below and the research here). As he goes on to explain, this would make the fatality rate “in the same ballpark as seasonal influenza”. A second report, covering Los Angeles County, was released on April 19th with similar findings, with actual infection rates estimated at 28-55 times higher than the 7,994 documented cases.
One note of caution: we don’t know for sure that people who’ve had COVID-19 are immune, not in perpetuity. There is at least one instance of someone catching it twice – a Japanese woman, although she may have been immunocompromised. Even if you’ve had COVID-19 as a result of being exposed to SARS-CoV-2, coronaviruses have a nasty habit of mutating, so you could catch it for a second time from another strain that you’ve got no immunity to. However, cases of reinfection are extremely rare to date and when viruses do mutate they tend to become less deadly, not more. Why? Because more deadly strains kill off their hosts faster and hence are less successful at replicating themselves. As a rule, the most successful coronaviruses in evolutionary terms are the least harmful, like those associated with the common cold.
Further Reading
‘“Trump Is Right About the Coronavirus. The WHO Is Wrong,” Says Israeli Expert‘ by Oded Carmeli, Haaretz, March 21st 2020
‘Fundamental principles of epidemic spread highlight the immediate need for large-scale serological surveys to assess the stage of the SARS-CoV-2 epidemic‘, Sunetra Gupta et al, MedRxiv, March 24th 2020
‘Coronavirus may have infected half of UK population – Oxford study‘ by Clive Cookson, Financial Times, March 24th 2020
‘How deadly is the coronavirus? It’s still far from clear‘ by Dr John Lee, The Spectator, March 28th 2020
‘Covid-19 – The tipping point?‘, Tom Jefferson, Carl Heneghan, Centre for Evidence-Based Medicine, March 30th 2020
‘It’s very rare to catch Covid-19 twice‘, FullFact, March 31st 2020
‘How likely are you to die of coronavirus?‘ by Tom Chivers, UnHerd, April 1st 2020
‘Covid-19: four fifths of cases are asymptomatic, China figures indicate‘, British Medical Journal, April 2nd 2020
‘Coronavirus, Castiglione d’Adda is a case study: “70% of blood donors are positive”‘ by Monica Serra, La Stampa, April 2nd 2020
‘Population-level COVID-19 mortality risk for non-elderly individuals overall and for non-elderly individuals without underlying diseases in pandemic epicenters‘, John Ioannidis et al, medRxiv, April 8th 2020
‘Covid antibody test in German town shows 15 per cent infection rate‘ by Ross Clark, The Spectator, April 10th 2020
‘1-in-7 New Yorkers May Have Already Gotten Covid-19‘ by Justin Fox, Bloomberg, April 15th 2020
‘Has SARS-CoV-2 Fooled the Whole World?‘, Mikko Paunio, LockdownSceptics.org, April 16th 2020
‘COVID-19 Antibody Seroprevalence in Santa Clara County, California‘, John Ioannidis et al, medRixv, April 17th 2020
‘Stanford study suggests coronavirus is more widespread than realized‘ by Ross Clark, The Spectator, April 17th 2020
‘Global Covid-19 Case Fatality Rates‘ by Jason Oke and Carl Heneghan, CEBM, April 17th 2020
‘The end of exponential growth: The decline in the spread of coronavirus‘ by Issac Ben-Israel, The Times of Israel, April 19th 2020
‘Early results of antibody testing suggest number of COVID-19 infections far exceeds number of confirmed cases in Los Angeles County‘, University of Southern California and Los Angeles County Public Health Department, April 20th 2020
‘Getting a handle on asymptomatic SARS-CoV-2 infection‘, Daniel P Oran and Eric J Topol, Scripps Research, April 20th 2020
‘New York antibody study estimates 13.9% of residents have had the coronavirus, Gov. Cuomo says‘, CNBC, April 23rd 2020
Further Viewing
To join in with the discussion please make a donation to The Daily Sceptic.
Profanity and abuse will be removed and may lead to a permanent ban.
There is zero chance that most people, or many people, in the UK were already infected. As of March 10, 25,888 negative, 373 positive. That tests for presence of the virus, rather than antibodies – but in a rapidly growing epidemic, most cases are very new and would test positive on a PCR-type test like this. They didn’t, so it wasn’t.
That simply does not follow. You can only apply random testing logic to a random test.
That’s correct. In the early days of the infection even those who had been infected without symptoms would have shown up in the initial bulk PCR tests. We did NOT see that, so NO hidden mass of ex-infected existed at that time.
We cannot know how many people had this virus prior to lockdown. Anecdotally, I can think of no one who did not have very unpleasant cold symptoms between late December & early February. So many probably had it as have not had symptoms again. Ferguson is not reliable. His work on Foot & Mouth left many of us with doubts about his modelling.
Jennine, how do you explain, then, that nobody died from these unpleasant colds in late December and early February, yet now we have nearly a thousand dead a day in hospitals and hundreds more in homes? The disease progresses from infection to hospital admission in an average of ~10 days and death an average of ~7 days later – there is abundant data supporting this timeline. The pattern of the disease’s arrival in each geography also fits exactly with the idea that nobody had the disease in the UK in December and very few in February.
I met someone who nearly died from a virus with respiratory complications in November 2019.
4,700 people died from respiratory illnesses in November 2019. Of those, about 90% were over 65 years in age.
The worst numbers this year for respiratory disease are in January 2020, 2,477 in the second week of January, higher than the 2,106 registered last week.
Numerical analysis seems not to get us anywhere, particularly since the whole of Europe is having fewer deaths this year than for the same period in 2016, unless Covid 19 has been massively overhyped.
Hmmmm………
This individual’s son was an estate agent showing Leeds University students around properties to let. Leeds University has an international partnership with Wuhan University of Technology.
In the clip below, Dr Peter Forster, a geneticist from Cambridge University explains how the earliest the virus could have crossed over to humans in China was 13 Sept 2019
https://www.youtube.com/watch?v=AQQf2yoymu0
Chinese nationals from Wuhan were hospitalised in Britain 24 Jan 2020 having been in the country for two weeks.
https://www.theguardian.com/science/2020/jan/24/coronavirus-uk-universities-issue-quarantine-warning-china-chinese-students-in-slugs
The autumn session of Leeds University began Wed 25 Sept 2019.
Knowing what we now know about the infectiousness of the disease, and the mildness of its symptoms, asymptomatic transmission in the able bodied, a great deal of transmission could have occurred in this country before red flags were raised concerning the virus in January 2020.
The Cambridge University paper commentary and link is here:
https://www.cam.ac.uk/research/news/covid-19-genetic-network-analysis-provides-snapshot-of-pandemic-origins
Looks like you’ve needed to revisit your thinking.
As always looked most likely, around 5%-10% of badly affected countries have had it so far and the elevated death rates were pretty much what the national modellers said they were: 0.5%
Funny how doctors around the world were able to spot a new disease though, isn’t it. Seems that if you work in an ICU, the Covid patients are obvious straight away. Again, nobody had it here in November, the pattern of accelerated deaths exactly fits the National models not your pet theory.
Anecdotal evidence but I work in an urgent care centre and last autumn there were a lot of people coming in with persistent cough lasting several weeks, including younger people. I had a persistent cough for 6 weeks around the same time, I have diabetes. I remarked to patients that there were some nasty viruses going around that seemed to affect people with diabetes and asthma worse. This coincided with the start of term at our local university with a large number of overseas students. A whole group of customer facing staff at the university library went off sick in January. Finally the virus was found in effluent in Italy and Brazil in November 2019. People’s reserves of vitamin D were higher than in January/February therefore more likely that their innate immune system could handle the virus. Whereas in January/February vitamin D is low unless people take supplements, it is also the time when other viruses strike. You cannot say categorically that the virus wasn’t present in one form or another late last year. Any admissions to hospital would not have been tested for the virus, simply a Lower respiratory tract infection. No autopsies have been performed so there’s no… Read more »
The fact that no autopsies were not being done should have sounded warning bells that governments were hiding something.
I know of no one who has had Covid never mind died from it, and after 14 months that is still very much the case!
It is great to see all the arguments addressed without hysteria or censorship.
It is very difficult to see how every area of the UK could be infected so quickly, with a very small percentage of the population infected. A model that spreads from a very few infected people at an R0 of 2-3 would take some time to reach across the UK in January/February.
It was delivered, that’s why.
The Finnish National institute for Health and Welfare estimates the IFR to be less than 0.2%
https://translate.google.fi/translate?hl=fi&sl=fi&tl=en&u=https%3A%2F%2Fwww.hs.fi%2Ftiede%2Fart-2000006473402.html&sandbox=1
How then do they explain how too many people have died in Bergamo? More than 0.2% of the population already. Or that Sweden is at ~50 deaths a day today in Stockholm when only 2.5% of Stockholm had the disease (from a representative sample tested) on the 3rd of April?
The only data that’s available for contained populations, or representative tests supports the Imperial model at 0.5-1.5% IFR.
Hi Will, possibly excess deaths ascribed to COVID but actually caused by limited medical facilities due to temporary overload.
Most papers seem to be taking the antibody samples as a lower bound – this is because positives tend to be retested to confirm but not negatives. Also, it’s possible that asymptomatic cases don’t generate very many antibodies (or any at all). Either/both of those would explain your conundrum.
More proof for a low IFR can be found with the COVID tracking app from Kings College and St Thomas’ Hospital which had up to 2 million cases on 1st April in the UK. That’s symptomatic cases.
Neither of those arguments stack up.
The Covid tracker as a data point for actual cases is a joke, surely. Some 10% of people report a cough at any time over the winter. It’s hard to find someone who doesn’t think “they probably had it in December”.
The Bergamo explanation stretches credibility to say the least. They were taking bodies away in Army lorries. 4,500 people died in a month in a tiny region having been treated for pneumonia in their 1000s. You doubtless saw the pictures and video and testimony of the doctors. 50 working age doctors and nurses died in Lombardy alone
Questioning the efficacy of the lockdown is totally reasonable. But to still try to make this a bad flu you have to turn to the slightest of evidence and implausible explanations and ignore the vast body of evidence from China, Italy and now Sweden. Plus the ice rinks used as morgues in the few places this ran free before social distancing and reduction in R. Occam’s razor says that the IFR is 0.5-1.5% and that only 5%-10% of people in the uk have had it. It fits all the data including the regional concentrations.
Is it a joke? They’re reasonably sure about it – it’s the best way we currently have to get any idea of cases. They’re not looking at coughs but clusters of associated symptoms and they’re using the (large number) of people with the app who have been tested to make judgements based on symptoms. So they’re confident they can discern between the Covid symptoms and a normal cold. If you read the blog https://covid.joinzoe.com/blog then you can see their methodology, and how they claim to distinguish different illnesses.
It’s a joke. Because we have had zero testing outside hospital for many weeks, this may represent our best data in the UK but it’s still a joke, it’s producing results laughably out of line with places that have actual Covid sampling data. In many respects it presents as flu, so how could it be accurate? Fortunately other countries have had better investigative testing regimes and know how many people actually had it at certain points in time.
Hi Will, do you have any sampling data from places that have also used the app? If so please link it!
Now that we have good antibody tests, it’s clear that it was as lethal as originally thought. Still only 5% have had it in Spain.
I am not an immunologist, but I have a degree in computer science and in order to get a handle on what is going on here, I have created an epidemiological model. As a result, it has led me to wonder about the criteria for ‘exposed’, ‘infected’ and ‘immune’. In immunology there seems to be the concept of ‘minimum infectious dose’. If considering influenza (presumably better understood than Covid-19) we read the following: “Influenza Virus Aerosols in the Air and Their Infectiousness. Influenza is one of the most contagious and rapidly spreading infectious diseases and an important global cause of hospital admissions and mortality. There are some amounts of the virus in the air constantly. These amounts are generally not enough to cause disease in people, due to infection prevention by healthy immune systems. However, at a higher concentration of the airborne virus, the risk of human infection increases dramatically…. The human infectious dose of the influenza A virus, when administered by aerosol to subjects free of serum neutralizing antibodies, ranges between 1.95 × 10^3 and 3.0 × 10^3 viral particles…. it is important to consider that the risk of acquiring influenza is determined by both the concentration of the… Read more »
Great work Caswell. I’ve been writing about this possibility of a 15% infection limit on Conservative Woman for the last couple of weeks. Can you publish your model somehow – maybe Toby could host it here? Did you see that the Charles de Gaulle aircraft carrier reached 33% infections, maybe more once all results are in? I don’t think this invalidates your model as you allow for different conditions driving greater spread but it would be good to include it and maybe reflect on the difference between a cruise ship and a military vessel in terms of the amount of close interaction.
More data: antibody tests show 22% in Robbia, Lombardy were infected. That further corroborates the model I think.
Many thanks, Will. There certainly does seem to be a picture emerging. I saw a nice quote from Dr. John Lee in the latest Spiked:
“The UK models, as I understand it, said that 60 to 80 per cent of the population would be infected by the virus in a short period of time. Whereas in fact some of the evidence from enclosed populations, such as the Diamond Princess cruise ship, suggests that only 15 per cent of the population may be susceptible. Maybe that is because the virus spreads in a different way than we thought. Maybe some people actually have immunity based on other coronaviruses that are already out there in the population. “
That’s a great interview. Well done spiked. John Lee stopped short of saying that in his last Spectator piece so I’m glad he’s voicing it now. I hope you can get the model published somewhere – this seems to be very significant, especially with government listening to all the wrong people who don’t seem to be interested in the actual data or considering other possibilities.
Quick calculation: If 15% of the population are susceptible ~10M. Lowest figure I’ve seen for developing herd immunity is 62% so ~ 6M need to be infected. Most consistent average figure I’ve seen for IFR is 0.19%. Implies around 116K UK deaths with no action. By the end of this lockdown we will probably see about 25 to 30K. Summer may help, but we will need to smooth / delay the remaining 90K. Social distancing, masks and protecting the vulnerable will still be required to lower any possible subsequent peaks.
This doesn’t look right. The idea is that a form of herd immunity would develop after around 15% infections. So given 0.19% death rate the number of UK deaths would be 66.4m x 0.15 x 0.0019 = 18,924 ie similar number to what we seem to be heading for.
Derrick – where have you seen IFR of 0.19%? Only surely on skeptical sites like this – that level is very much an outlier. Vast majority of international models seem to assume 1-3% – the upper end coming when appropriate treatment isn’t possible. And that is because that level is supported by the only widescale representative sample studies eg Stockholm and Wuhan and because places where the virus ran free for weeks, like Bergamo, an implausibly high proportion of the population have died for a 0.19% IFR. – 0.45% of the entire Bergamo region have died. Together with other data it suggests that something like half the population was infected.
The figures are from the Iceland study where a high proportion of the population have been tested:
“Iceland’s higher rates of testing, the smaller population, and their ability to ascertain all those with Sars-CoV-2 means they can obtain. an accurate estimate of the CFR and the IFR during the pandemic (most countries will only be able to do this after the pandemic). Current data from Iceland suggests their IFR is somewhere between 0.01% and 0.19%.
and Finland
https://translate.google.fi/translate?hl=fi&sl=fi&tl=en&u=https%3A%2F%2Fwww.hs.fi%2Ftiede%2Fart-2000006473402.html&sandbox=1
Impossible to be definitive though until the pandemic has run its course or accurate mass anti-body testing has been widely achieved and “excess” deaths are known.
Where’s the data from Iceland showing only 0.01-0.19% IFR?
Stockholm area did a rep sample test on the 3rd April and found 2.5% infection rate with cases doubling every 6 days at that point and 50 deaths/day today, it gives an IFR of 1-1.25% (in line with all major international models).
There are a number of studies that support this kind of death rate including a Danish antibody survey, a German one and an American one from Colorado.
Will, that’s very interesting, would you mind posting links to those?
https://www.t-online.de/gesundheit/krankheiten-symptome/id_87680236/lockerung-der-corona-massnahmen-ergebnisse-der-heinsberg-studie-machen-hoffnung.html
https://www.dr.dk/nyheder/indland/doedelighed-skal-formentlig-taelles-i-promiller-danske-blodproever-kaster-nyt-lys
https://reason.com/2020/04/08/mass-antibody-testing-in-this-rural-colorado-county-sheds-light-on-covid-19s-prevalence-and-lethality/
Also https://www.medrxiv.org/content/10.1101/2020.04.01.20050542v1
Will, the German study shows 0.4%, just outside the 0.5-1.5% consensus.
The Colorado study is not complete and not worth its name yet, it only has 17 positive Covid cases, clearly not enough to draw an infection rate.
The Danish study is a strange sample of only blood donors and it’s author warns not to draw any conclusion from it.
That compares to representative studies from Stockholm giving 1-1.25%, the entire region of Bergamo giving 0.45% as a minimum *if every person was infected*. And more generally, the patterns of spread and death fitting easily with standard models with a 1% death rate.
It’s still possible just about that the IFR could be as low as 0.2% but the evidence stacked up against that is massive, the evidence for is sketchy, at best, it’s not what you’d call an educated best guess. It’s more of a guess with an agenda having to be very selective with data collection.
Question about Bergamo, Will – are you using total excess deaths or just deaths recorded as from COVID? Can you link your numbers? I can’t easily find recent statistics from that town.
Are the deaths spread equally amongst the population of Stockholm?
Do you have a link for Stockholm? Lombardy is generally regarded as atypical for local reasons such as poor air quality and high antibiotic resistance https://www.ansa.it/english/news/science_tecnology/2019/11/19/italy-top-in-eu-in-antibiotic-resistance_369e0123-0107-445e-8c17-f11932c9d27c.html.
You’ve rounded Gangelt up from 0.37 to make it seem closer to 0.5.
A new study from Stanford concludes 0.12% to 0.2% https://www.medrxiv.org/content/10.1101/2020.04.14.20062463v1.
Using data from the cruise ship Diamond Princess, Stanford Professor John Ioannidis showed that the age-corrected lethality of Covid19 is between 0.025% and 0.625%. https://www.statnews.com/2020/03/17/a-fiasco-in-the-making-as-the-coronavirus-pandemic-takes-hold-we-are-making-decisions-without-reliable-data/
The Centre for Evidence-Based Medicine (CEBM) at the University of Oxford argues that the lethality of Covid19 (IFR) is between 0.1% and 0.36% https://www.cebm.net/covid-19/global-covid-19-case-fatality-rates/.
The Finnish epidemiology professor Mikko Paunio from the University of Helsinki has evaluated several international studies in a working paper and comes to a Covid19 lethality (IFR) of 0.1% or less. https://dailysceptic.org/wp-content/uploads/2020/04/How-the-World-got-Fooled-by-COVID-ed-2c.pdf
And another ship, USS Theodore Roosevelt.
“As of today, 94% of USS Theodore Roosevelt crewmembers have been tested for COVID-19, with 655 positive and 3,919 negative results.”
https://navylive.dodlive.mil/2020/03/15/u-s-navy-covid-19-updates/
I believe that’s 14.3%.
Is it the case that you think this supports the idea that only 15% are susceptible? It surely supports a much higher percentage than that.
And another boat:
“Meanwhile, a Dutch navy submarine, MS Dolfijn, has returned to its Den Helder base two weeks early because of a coronavirus outbreak on board. Eight of the 58 crew tested positive and the submarine, which had been sailing near Scotland, is now in quarantine.”
https://www.bbc.co.uk/news/world-europe-52308073
That’s 13.7%.
Also, referring to the same Dutch submarine:
“Fifteen members of the MS Dolfijn’s 58-man crew developed mild flu-like symptoms last week and were tested on Monday. Eight were revealed to have antibodies for the virus.”
https://sputniknews.com/europe/202003311078781646-dutch-sub-cuts-mission-short-after-sailors-catch-coronavirus/
Thus suggesting that only half the people with mild flu-like symptoms went on to develop Covid antibodies..? Maybe it was just coincidence that the other half also fell ill, of course, or maybe they were tested too early in the illness.
Will/Caswell, reading this thread there seems to some confusion of concepts. The herd immunity infection ceiling is determined by the r0 of the population. A high r0 will see much higher infection ceilings, a low r0 a much lower one. No two populations will have the same r0 and even within the population factors will vary the r0 and infection ceiling, like population density, London will have a much higher ceiling than rural areas. Importantly a lockdown is an r0 influencing measure, it is trying to suppress a community’s natural r0 to one below 1 to stop the spread. Further, simply measuring infection at a certain point in an outbreak, especially when r0 has been manipulated by introducing a lockdown does not mean that the measured infection rate represents the ceiling for that population. In all the studies so far (Diamond Princess, USS Therdore, etc) the populations were lockdown and measured, so all the infection proportion reflects is how far the infection had got before this was done. The only comparison you can draw is how quickly the leaders reacted. Without knowing what the r0 of that community was prior to lockdown we can make no estimate about whether it… Read more »
Hi Sinichol, that’s exactly what they’re discussing – possibility that there is a limit on susceptible people in the population.
Hi sinichol. I think I’m aware of all that, but my point is that when you say “65%-70% for herd immunity” you are allowing only for a binary ‘immune’ or ‘not immune’, and I think you would be counting anyone not showing antibodies as ‘not immune’. I’m suggesting that there may be graduations of ‘exposed’, ‘infected’ and ‘immune’. In the real world, with some people receiving small exposures to the virus, they may already be resistant to infection to a certain level, or they may develop a degree of immunity that falls short of an illness with symptoms and the production persistent antibodies. Nevertheless, their immune system may develop a memory of the virus ready for next time. This might explain cases where a person seems to become re-infected having already fought off the illness. Maybe the re-infection comes when they receive a much higher dose. Maybe they really were infected previously as shown by the PCR test, but didn’t develop full, binary, ‘immunity’. I’m effectively taking the virologists and immunologists at their word when they say ‘We know nothing about this; this is brand new’ and suggesting that there might be intermediate states that modellers are possibly not allowing… Read more »
With my limited understanding of virology, I think you’re taking them at their word that this is completely new. Viral load is a know concept, and Chris Whitty certainly suggested a few weeks ago the reinfection capacity is not a new thing, it happens with other viruses too. I don’t think any antibodies are binary with any virus, and I don’t think this is new to Covid. I’d imagine it is quite analog. As in you’re either: 1) a poor healthcare worker who just breaths in loads of virus, or you have a weak immune system and it gets chance to replicate loads and spread around the body, but either way you have large system wide infection before you finally overcome it. 2) you get a small infection or have a very good immune system and still clear the infection with a limited local reaction. I’d imagine in the former case leads to lots of antibodies being present potentially being better spread out across the body. Whereas the latter see far fewer potentially locally provisioned antibodies, so that with a large viral load reinfection the immune system is still overwhelmed a second time before enough antibodies are created to overcome… Read more »
I haven’t ended up with a ‘ceiling of infection’ of 15%. I have ended up with 15% of the people in the population showing antibodies. You are equating antibodies with ‘infection’ and no antibodies with ‘no infection’. Maybe that’s right, but maybe it isn’t. In this version of the model, I am equating antibodies with ‘heavy infection’. I have a second tier of people who have been infected differently/less, resulting in a lower level of immunity and no antibodies. They have still been notionally infected, though. If I add the antibodies people to the infected-but-no-antibodies people it is, perhaps, 70% of the population. So my ‘ceiling of infection’ might be said to be 70% on that basis. The point is not to claim that this is definitely how it really is. It is to attempt to tie together a few ‘anomalies’ that we seem to be seeing and suggest that perhaps the modellers could be simplifying it too much. Neil Ferguson might instruct the government to wait until we see 60% of the population showing antibodies before relaxing social distancing, for example, but I am suggesting that we don’t actually know that it would ever come. We might need more… Read more »
Yes, you have created a model that has an infection ceiling of 70% by trying to model two well the analogue nature of infection!
If possible, please explain to Will Jones that you have really led him a merry dance with his mission to prove infection ceilings are all 15% regardless of what r0 the population in question is… I’ve done my best to explain!
By infection rate I mean infection that leads to antibodies, which is what is implied by the antibody surveys purporting to show infection spread and hence IFR. I understand that in Caswell’s model there is a larger group of a presumably lighter form of infection/ exposure. You could count that as infection but then you’d need to be consistent when stating IFR to avoid giving a false impression of death rate. If these ‘light infections’ don’t show up in antibody tests then there’s also no way to identify them, especially if they don’t show up in infection tests either. It may be helpful to use a different term to avoid getting confused between the two types of infection/exposure when they have quite different properties especially in relation to identifying them – one showing up in tests, the other not.
Will, if the infection ceiling is 15%, then the IFR must be at least 2.7% to explain the number of dead in Bergamo. I expect you won’t like that outcome, either.
A number of explanations have been put forward by specialists to explain the
high number of deaths in Lombardy including:
– greater susceptibility to pneumonia because of eg poor air quality or water contaminants (the area is known for this)
– older population
– deaths from isolation
– deaths from other conditions not being treated
– the health service being overwhelmed – this has happened before in bad flu seasons buthas been made worse in this crisis by healthcare staff needing to self isolate or not wanting to turn up to work.
See https://swprs.org/a-swiss-doctor-on-covid-19/ (search the page for Italy to find the relevant parts).
If you read my original comment, you’ll see that I explained precisely what I meant, and Will Jones understands it, too. Maybe you’ve leapt in half way down.
I am suggesting that the boundaries between ‘exposed’ and ‘infected’, ‘infection resistant’ and ‘immune’ are likely to blurred, not binary. A study has apparently found a number of Covid patients with mild symptoms who went on to produce no detectable antibodies. Were they originally ‘infected’? By your definition they weren’t. I say that they were infected in a way that didn’t provoke the production of persistent antibodies. I am also open to the idea that there are other levels of innate ‘resistance’ or maybe we could call it ‘immunity’ that act before ‘infection’. These may also contribute to ‘herd immunity’.
This discussion is, in fact, very illuminating and if Neil Ferguson also thinks along rigid, binary lines then I can see we’re in big trouble!
Caswell/Will, I think you’re confusing the discussion by adding in to the model a concpet that virologist already call something else. The concept of genetic immunity is well understood by virologists. I think you’re trying to add that to your infection model in a way that threw me. As an example the way that viruses attack cells can not affect certain people, so they don’t show infection as the virus can’t affect them so they don’t produce antibodies because they simply don’t actually get infected. These people are simply already immune, or need taking out of the total in some way – e.g. 2/3 don’t get infect so with an r0 of 2.9 and a ceiling of 65% for those that can be infected you will get a population ceiling of 65% x 1/3 = 21.7%. I believe an existing genetic study in the UK has added Covid and is looking into this… https://www.the-scientist.com/news-opinion/dna-could-hold-clues-to-varying-severity-of-covid-19-67435 It is entirely possible that this is the case with Covid, but I stress again none of the “have it now” numbers can show this, and none of the “have had it” (Gangelt and Robbio) numbers show either way at the moment, we either need an… Read more »
The natural r0 of a population is what determines an infection ceiling, but within it will vary based on factors like population density, so London has a higher ceiling than rural areas. A lockdown is an attempt to suppress the natural r0 of a population to below 1. This study looked at pre lockdown r0 by country. https://www.journalofinfection.com/article/S0163-4453(20)30154-7/fulltext The UK very similar to Germany at 3. All the studies (Diamond Princess, USS Theodore, etc) locked down populations (changing their natural r0) then assessed the level the infection had got to at that point. Without knowing the r0 for that community prior to lockdown it is impossible to determine if it had got to it’s own natural ceiling, and wrong to take the infection level measured at that point as its ceiling, or to try to apply it to the UK as ours. Ahead of our own serology survey the best we can do is to find a country as similar to ours as possible in lifestyle factors, healthcare provision, etc, that has done a serology survey and estimate from their figures. At present the Gangelt one from Germany, is the closest; similar r0 of 3 so infection ceiling 65-70%, although… Read more »
Studies and surveys keep coming out showing 15% infection rate, including Shenzhen in China and Gangelt in Germany, often after the virus has been circulating for a while. No studies yet show much more than that, with the Charles de Gaulle quite an outlier at 33%. Why then do you assume herd immunity must come at 60% when no study yet has found a population at anywhere near that level of infection?
Very few places have had the virus circulating for long enough to reach more than 15%, have they? In Northern Italy in certain towns, it clearly was circulating with no change in people’s behaviour for many weeks (not the 2-3 months needed for reaching peak) and it’s clear that either the infections were well above 15% OR IFR was much higher than 1%…..
In Robbia, Lombardy 22% were infected. That’s more than 15% but not much more compared to 60%. The death rate in northern Italy may be elevated for a number of reasons including greater susceptibility to serious pneumonia and the consequences of lockdown for other health treatment. I don’t know if the 15-20% ceiling idea is true but given the data that keeps coming in it needs to be properly explored not dismissed because it doesn’t fit with the current models.
All the 22% reflects is how far the infection had spread before lockdown stopped the spread. It is not the natural infection ceiling, it is an artificial one created by the lockdown.
When they lift if the don’t keep r0 below about 1.3 infection will rise.
Interesting numbers they have measure 0.7% mortaility, almost twice that of 0.37% in Gangelt in Germany, the first real sign of the relative degree to which the Italian health service was overwhelmed. Very sad.
I think it is Robbio or Commune Di Robbio:
https://www.tgcom24.mediaset.it/cronaca/a-robbio-pv-il-22-ha-o-ha-avuto-il-coronavirus-ok-del-sindaco-ai-test-per-tutti_17285128-202002a.shtml
Some people, e.g. Suneta Gupta at Oxford University are estimating high R0 numbers, etc. That translates, for simple amateurs, to a doubling time every 3 days.
e.g., https://philip.greenspun.com/blog/2020/04/09/evidence-that-sunetra-gupta-and-her-oxford-team-were-right-r0-for-coronavirus-may-be-5-7/
https://www.medrxiv.org/content/10.1101/2020.04.04.20050427v2
The most interesting case to date is the French aircraft carrier that was probably infected when it visit port in mid march and had 1081 cases about 30 days later. 2 to the power 10 is 1024.
1) Infection ceiling I don’t assume anything about the study groups infection ceilings, they will all be different. It is not some fixed number, the infection ceiling changes with the r0 – e.g. 1.25 will have an infection ceiling of roughly 15%, 2.9 will have a ceiling of 65%. The less aggressive the spread, the fewer immune people you need to stop it overall. All the populations in the various studies will have had natural r0s and different infection ceilings, for the Theodore or de Gaulle likely really high as they all pack in like sardines, say r0 of 10 and infection ceilings of 95%. Diamond Princess probably lower than you think as lots of separate cabins, the London tube probably more like Navy’s! The UK was measure to have an r0 of 2.9, so a 65% ceiling. The virus spread, and then authorities (governments or captains) intervened and lockeddown the population in question as people started dying – e.g. the Diamond Princess confined to quarters, the UK semi-lockeddown. Once lockeddown the r0 changes and drops below 1 which makes the theoretical infection ceiling 0% as in it can’t grow. So whatever the infection level was at that point is… Read more »
You’re forgetting Shenzhen which found 15% of those who lived with infected people had developed antibodies – that was unaffected by lockdown. So that’s Gangelt, Shenzhen and Diamond Princess – although the last may have had some who had recovered before they were tested. Then there’s Robbia at 22%. You are right that this could reflect a spread stopped in its tracks – but as Caswell says it’s quite a coincidence all stop around the same time, and match the Shenzhen rate which was unaffected by lockdown. Why not consider the possibility that there is a lower infection ceiling? You keep quoting natural R0 figures and corresponding ceilings but those are just the product of the assumptions put into them. What if this virus behaves differently to those assumptions, such as in the way Caswell suggests?
Diamond Princess did not do a serology survey, they simply sampled active infections as everyone came of the ship after 2 weeks of quaratine to the cabins, bear in mind the lockdown will have lowered active infection counts, especially after 2 week, and we don’t know how many had antibodies. I’m not aware of a serology survey in Shenzen, if you are please share the link? Gangelt and Robbio are the only two reasonably large, unbias sampled, serology studies that I’m aware of from countries I would trust numbers from. It is no coincidence, when you notice people dying, you stop them seeing each other and you lock in whatever infection level has occurred by that point, and that this happens early in infection. We are humans, we act to protect. As to thinking about infection ceilings, just look at the de Gaulle numbers, 33% active infections, active, not serology survey, so we have to assume more had had it and tested negative on Gangelt numbers of 2% active and 14% with antibodies. The de Gaulle r0 was probably like 10 and probably had the capacity for an infection ceiling of 95% or something mad like that. I really wish… Read more »
They tested for the virus rather than antibodies in Shenzhen but concluded ‘the household secondary attack rate was 15%, and children were as likely to be infected as adults.’ https://www.medrxiv.org/content/10.1101/2020.03.03.20028423v1.full.pdf
So the same as the Diamond Princess.
I memory serves the best guess is a 5 day test window with 70% accuracy. The study quotes 6.3 day mean serial interval meaning many tests of household members will have gone negative by the time they were sampled, and what if they gave it to the person that caused their contact trace test? They would definitely test negative for an active infection. Chances that 15% was the actual secondary infection rate seems really low, easily far higher, and you still need to actually catch it. Were talking about people that knew about it and will have self-isolated within the home so will have been actively trying not to catch it. This dramatically changes the r0.
Sorry, just don’t buy that number as meaning anything.
Did you read the paper? You appear to be dismissing their research and conclusions because it doesn’t fit with your assumptions.
These will also be of interest to you. Knut Wittkowski in a new paper concludes: ‘In China and South Korea (and the first wave in Iran), incidence peaked after about 2 weeks and then declined. In Europe it took and in the U.S. it takes twice as long. The shorter duration of the epidemic in China and South Korea, however, does not demonstrate the effectiveness of social distancing, because the social distancing started too late to be effective. Instead, the longer duration in Europe is consistent with premature interventions in Europe prolonging the epidemic.’
On Sweden: ‘There was no difference in the shape of the epidemic or the height of its peak to the other Scandinavian countries.’
https://www.medrxiv.org/content/10.1101/2020.03.28.20036715v3.full.pdf
In terms of lockdown reducing reproduction rate, it’s clear that in Germany the reproduction rate had already fallen below 1 before the lockdown https://www.rki.de/DE/Content/Infekt/EpidBull/Archiv/2020/Ausgaben/17_20_SARS-CoV2_vorab.pdf and the same in Switzerland https://bsse.ethz.ch/cevo/research/sars-cov-2/real-time-monitoring-in-switzerland.html.
Yes, I read the paper, it makes no claims to be asserting an infection ceiling, you’ve taken a number and decided to imply that yourself. I’m just trying to point out to you why it is wrong to do that. Please read my response to sunchap below which tries in the clearest way possible explain why you can’t read PCR “have it now” surveys results like that. The Knut Wittowski paper, which I’ve just read, is terrible. You can’t compare cases, they are just a function of testing levels. So he just compared a whole load of policy driven almost random numbers that don’t even provide him with comparatively correct proportions of true infection levels between countries, meaningless. Until we have serology numbers, you can only compare deaths/1m between comparable countries and even then only to gauge the relative degree of spread. Everything else is a fools errand given testing strategies are all so different. Sweden has the same measures in place as us, the only steps they did not take is to enforce working from home and close schools. People keep talking like it is so different, go to any park in the UK, it really isn’t that lockeddown… Read more »
When the paper concludes ‘the household secondary attack rate was 15%’ I hardly think I’m inserting the idea myself – they don’t say ‘15% of those tested tested positive at the time of testing,’ they draw the wider conclusion. I agree that it seemed odd that Wittowski relied on test data. Test data obviously doesn’t show the true spread, and when test numbers are changing the raw number of positive tests is not a useful statistic and has been widely misused. However, once the number of tests becomes more constant then it can be used to detect a decline in infections, if the testing strategy also remains broadly the same. News reports say Germany imposed lockdown on 22 March not the 15th. Do you have a link for the earlier date? Sweden does not have lockdown in anything like the same way as other countries, though yes it has introduced some social distancing, much of it voluntary but being observed. Willow has an interesting comment on yesterday’s post: “The first response of the immune system to an invading pathogen is mediated by what’s called innate immunity. Innate immunity is not pathogen-specific. It is a range of bodily defences that include… Read more »
Yes you are, they are simply saying that was the number of active infections measure at on average 6.3 days after the original contact in other people in their household. They make no claim this is in infection ceiing. That is your theory. … and as data science it is very flawed. I’m not saying anything different to Caswell. I’ve said all along that genetic immunity is entirely possible in subset and they will NOT present antibodies. If 50% have such immunity, then you simply start the population infection at that level, so for an r0 of 2.9 to achieve 65% herd immunity you would need to measure 15% antibodies to get to overall herd immunity, and Gangelt is done. Robbio 22%, could mean the same just that the r0 was slightly higher in that case making the infection ceiling 72%. I entirely accept this is possible. I really want this to be true, but I’m just saying you can’t interpret PCR test study numbers the way you are to argue this, and you need to be looking for medical study proof the virus can’t infect certain people. All the quotes above are rhetoric not medical study results. We can… Read more »
Thanks. I don’t really disagree with this comment, except that the claim isn’t that people have genetic immunity that means the virus can’t infect them. The claim is they have an innate immunity based on a first response of the immune system that gives them a measure of resistance. They can be infected if they are sufficiently exposed to it. This might be (if this is true) why some surveys in some contexts are coming back with higher proportions of positive tests.
I have just posted a fresh thread with all the unanswered questions I see open from this and other broader discussions, grouped into 4 strands. I hope this means you accept why PCR results can not be read like this. We all accept that antibody numbers may NOT be recording enough immunity in a population that we might all have been exposed to the virus, at the same time we need to remain open to the fact that no numbers disprove that the lockdown is simply suppressing the spread. There should be no ego here, just a search for evidence. As to your point, if their response does not show up in the antibody tests then it is in the same group as genetic immunity as it represents a group people who won’t show up in the antibody test numbers, so this strand could be added to group 3 on my open questions list? I agree if merits further investigation, although it could just be a theory that explains the same numbers you would see if people are just getting the weaker of the two strains of the virus, and the Gangelt/Robbio antibody test does not register people with immunity… Read more »
There seems to be a lot of ‘bootstrapping’ in epidemiology. Dodgy stats (that the same scientists would accept were dodgy if asked about them) are ‘laundered’ to produce values like ‘R0’, then everything else is derived from that, resulting in statements such as “…the estimated number of people who could potentially die from COVID-19, whilst the population reaches the Pcrit herd immunity level, may be difficult to accept.”
I can’t help but start from the position that the ‘confirmed cases’ figure is dodgy – we know it is. So I wouldn’t even get to the next stage of fitting a model to it and declaring I knew the value of ‘R0’. And I wouldn’t run the model based on this dodgy ‘R0’ and then make a sort of moral judgement on the output of my own dodgy model!
Is the R0 used derived from confirmed cases? That’s obviously going to be inaccurate.
I think you guys are on to something. Covid 19 clearly turned down BEFORE the lockdown in Wuhan and South Korea : Wittkowski on. MedArxiv. It also went down naturally in Japan and Sweden. Something must have caused this and it must be immunity in the population.
The only accurate studies as to maximum infection rate are stated above and point to a “saturation” rate of about 15-20%. Being cooped up on the Diamond Princess for two to three weeks must maximise infection. It is highly possible that many humans have developed cross-immunity from other coronaviruses, i.e. the common cold.
It also appears unlikely that 60-70% have been infected from all the data.
In terms of cross-immunity, that’s exactly what German virologist Christian Drosten has now suggested. https://www.watson.de/!324026684
They would still show antibodies so would be included in the 14% or 22% from serology surveys to date, just part of the mild group. They wouldn’t be a reason to argue that even with an r0 of 3 we won’t see those numbers rise to 66%.
The silver bullet we are all after is genetic immunity, as in some chunk of the population that can be exposed to the virus, don’t get it and therefore never need to produce antibodies.
If they can find that, say 2/3 of us are like that then all of a sudden only 1/3 of us can get it and will develop antibodies so that 66% ceiling becomes 22% and we know that Robbio is all the way there, pub time!
… but, we need that kind of silver bullet.
Not just genetic immunity, cross-immunity too. And maybe some blurring of the boundaries between ‘exposed’ and ‘infected’, ‘infection resistant’ and ‘immune’, as Caswell suggests.
Yes, I mentioned both too, my point was that cross-immunity will still lead to antibodies, as it means you do become infected, but you have similar antibodies you can adapt to fight the infection.
By genetic immunity I mean you don’t even become infected, you have no need to develop antibodies.
The former you would get counted in a serology survey, the latter you would not, so modelling and accounting for them in interpreting number is very different.
The Diamond Princess did not do a serology survey. You’re really misreading the numbers and thinking about it wrong. Firstly, let’s consider the measure you’ll get from a “have it now test” as the spread starts you’ll read low cases, this will rise to a peak halfway to the infection ceiling then reduce again up to infection ceiling as the virus finds it harder and harder to propagate – e.g. a population with a ceiling of 65% at 63% infection will probably show 0.5% (guess) active infections, much like it did at 2% spread, but at 30% probably 10-15% (guess). The height of the peak will depend on the r0 and window length of a positive test. So without knowing these figures comparing “have it now” samples is pointless. Secondly, think about the dynamics of r0 and infection ceiling. An r0 of 1.25 will have an infection ceiling of 15%, 2.9 will have 65%, 6.6 of 85%. It is just harder for a slow spreading infection to spread with fewer people being immune. So comparing active infection measures without knowing the r0, or how progressed it was, is meaningless. Thirdly, think about how the active infections changed over the quaratine… Read more »
This is an interesting podcast, with professionals talking among themselves about SARS-Cov-2: https://www.microbe.tv/twiv/twiv-602/ Maybe the most interesting bit starts at about 57.45 The main points I take away from the podcast are: 1. The most common phrase is “We don’t know” (that’s not a criticism of them, it’s just how virology seems to be) 2. Mild infection/illness can result in low antibody counts, or antibodies that fade over time. Not much is known about it. 3. Reinfection might be possible, or maybe not; they don’t really know. 4. Antibodies for a previous respiratory virus were found to be present in about 30% of workers in a Chinese market. 5. Everyone seems to agree that the only practical outcome is natural herd immunity. “The virus is so widespread”. 6. Then they start talking about vaccine trials and the problems with possible vaccine induced enhancement. It seems to be superfluous, given point 5 – as though they are following a professional road map that must end in a vaccine, and the bit about herd immunity was just a moment of offline speculation. 7. Nice people can do very nasty things to mice. 8. A very low dose of the virus can act… Read more »
The apparently blind handover to epidemiologists seems very worrying and may be where things are going awry. Did you see Willow’s comment on yesterday’s post? It seems very helpful for your model: “The first response of the immune system to an invading pathogen is mediated by what’s called innate immunity. Innate immunity is not pathogen-specific. It is a range of bodily defences that include local inflammation, fever and the engagement of cells called Natural Killer Cells to literally consume the pathogen – as well as others I won’t go into. It’s complex! “Healthy young people and adults under ~40 seem to be able to see off the virus with just their innate immunity. This is good, it means the virus didn’t present much of a threat to them. They don’t have antibodies only because they didn’t need to produce them. They are not “immune” in the antibody sense, but they aren’t particularly vulnerable either. “If the innate immune system doesn’t defeat the pathogen, that’s when the adaptive immune system kicks in and the chain of events that leads to antibody production is kicked off.” It means aiming for herd immunity defined by 60% antibody presence may be unrealistic and unnecessary… Read more »
That’s brilliant. Many thanks.
Yes, I’m beginning to realise that some scientists are very uninterested in fields other than their own – seemingly going from total expertise in their own field to sub-member of the public awareness of others, even if closely related to their own. I find that fascinating, because you’d imagine that the philosophy of science would run through their veins; that they wouldn’t be able to stop themselves from absorbing science in general like a sponge.
They might not know exactly how a self-driving car works, for example, but you’d think they might understand perfectly well that there’s a crucial philosophical and political question about who the car should kill in the event of an unavoidable accident. The Covid-19 crisis seems to suggest that this decision would be left in the hands of people like Neil Ferguson, simply because that person would be an ‘expert’ and the rest of us couldn’t possibly have a valid view on it.
The interesting questions that remain open for investigation to me (please feel free to add) can be grouped roughly into these strands: 1) The New Scientist reported a while back that there were two strains to Covid, one weak, one strong. UK infection could well be a mix of both, Caswell’s model could be seen as trying to model both at the same time, but only recording immunity to the smaller one. https://www.newscientist.com/article/2236544-coronavirus-are-there-two-strains-and-is-one-more-deadly/ … there is an S and L type, the L was more prevalent in China. a) Is there evidence of the UK tracking the two different types? b) Does the PCR test pick up both? c) Are the really sick testing mainly for one type? d) Does immunity to one confer immunity to the other? Or, do they need modelling as two totally separate viruses? e) Does the Gangelt/Robbio test look from antibodies for both? f) Did Gangelt and Robbio use the same test? And how accurate is it, might 14/22% have large margins for error? e) Why is the Porton Down so slow to release antibody results? Are they stuck with all these problems too? … this is worth a read… https://www.sciencemediacentre.org/expert-comments-on-different-types-of-test-for-covid-19/ 2) The de Gaulle… Read more »
Yes, a good list of questions. At the end of the day, though, doesn’t the size of the list show simply that ‘life finds a way’? (to quote the Jeff Goldblum character in Jurassic Park)
Even if we answered all the questions, a third strain could pop up tomorrow. And a universal vaccine for the first two might react badly with it, making the current infection look like a (government-sanctioned) walk in the park.
Maybe we should just accept that we need to carry on living as a free society, and although we don’t like the idea of the threat of this virus in the background, we’re just going to have to get on with it?
Charles De Gaulle ended up at 59% active infections – huge number.
Yes, I really hope they do a serology survey on them. This will help fill in the picture, if the antibody numbers are lower then we know either the test does not measure all antibodies, or that there simply might no be antibodies. If it does, then we know it is much more likely we have simply slowed the spread. Plus, I really don’t like the way no ones discussing the different strains. For me it fits with Caswell’s model, and potentially what we are observing. Take this scenario… Strain1 is deadly, but harder to spread so only has an r0 of 1.25. Strain2 is mild and spreads easily with of 3. The PCR test picks up both. The antibody test only shows antibodies for 1. All the results we have fit this scenario, as would Caswell’s model. As an r0 of 1.25 would limit Strain1 to an infection ceiling of 15-22%, so Gangelt/Robbio results would be for populations at their infection ceilings. Strain2’s r0 would mean an infection ceiling of 65% and the de Gaulle PCR results. If no immunity confers from one to the other, we basically just have two viruses, but if it does then it gets… Read more »
Did some digging on the Santa Clara serology survey numbers of 0.12-0.2% mortality. https://www.medrxiv.org/content/10.1101/2020.04.14.20062463v1.full.pdf Unlike Gangelt/Robbio the author concedes there is bias in the population sample “mostly white women (63%) contacted via Facebook who could drive to test centres, age restricted to 19-64”, and the sample failed considerably to match the demographic of the County. He has had to massage the data to try to correct for this. What worries me more is he makes no observation about the possibility that this will have changed the r0 and therefore spread of antibodies in his sample. If they had a higher r0 (they certainly are able to travel more than the background population was willing to do) then they will have seen higher spread and his denominator in the mortality calculation wil be too big… More worryingly though, he has also had to estimate deaths to rush his results out. He based them on a sample of only 50 deaths in hospitals in Santa Clara by the 10th of April projecting what that number might be by the 22nd, but does not seem to make any mention of needing to make allowances for in-community deaths. He estimates this will lift to… Read more »
On the antibody test which is what matters for mortality ratios, I’ve found this article…
https://www.aruplab.com/news/4-21-2020/How-Accurate-Are-COVID-19-Tests
… which suggests 90% accuracy at 14 days increasing after that. So the 21 day window they recommend makes sense, no discussion of strains.
This was one of the studies they looked at from China…
https://www.ncbi.nlm.nih.gov/pubmed/32221519
… still don’t like the “of symptomatic cases” bit. What about asymptomatic cases? How do they know you can’t be exposed, be asymptomatic, and develop no antibodies like we suspect? Does not jump out of the work.
A very weak article. As the author points out, we simply don’t know a lot of basic data that would allow us to model. We do know that this pandemic has the capability to swamp our health care systems. We do know it kills people old and young. We do know it is killing our healthcare workers. We do know it is quite contagious – but still have work to do on exactly how it spreads. We do know reducing our social contacts cuts transmission and saves lives – and costs us dearly. We also know that we are not equipped with enough PPE or testing. Until we have some more answers which should be a matter of weeks & have sorted out PPE supply and testing, it seems sensible to cut the spread, reduce the load on our healthcare systems, understand the facts and formulate a proper plan. So yes, the Imperial work is not correct – they are some best estimates with the data we have. I’ll tell you something though, 68% of the population are not infected. I do wonder if the Oxford team quoted this to warn that their models and analysis can throw out odd… Read more »
“it seems sensible to cut the spread”. Quite a statement – if you look at what it means in practice.
Quite a result if you do not act too. Every country is taking some measures to different degrees. Many have gone for limiting interactions – the so-called lock-down which it isn’t.
On reflection, what is so controversial about cutting the spread of a deadly, novel virus for which nobody has natural immunity? Sure complain about how it’s done but not the goal surely?
Of course we have some innate immunity – we do to most pathogens. We may also have some acquired immunity via cross immunity from other coronaviruses. In any case though Sweden is having a better run than us with much weaker restrictions. And it’s now clear the epidemic plateaued before the lockdowns. Plus the lockdowns are themselves killing people. So they really aren’t worth it.
Latest research from Sweden suggests lockdowns to be a complete waste of time.
‘The study indicates that for every confirmed case of COVID-19, a further 999 people are likely to have been infected with the virus without knowing it.’
https://www.forbes.com/sites/davidnikel/2020/04/21/sweden-600000-coronavirus-infections-in-stockholm-by-may-1-model-estimates/#2f5e717f78d6
The virus may well have crossed to humans in China as early as mid September 2019.
Some universities in Britain have exchange programmes with Chinese universities including The Wuhan Institute of Technology. Students started work in Britain after the summer break on 25th September 2019.
It seems increasingly likely from the research that the virus has been with us for a great deal longer than originally thought (and modelled).
Sorry but the Forbes article does not conclude that social distancing is useless. It is stated that the work done that is the centre of the article is too small. There is a lively debate in Sweden about what their experience is. I am not poo-pooing tegh study. I am saying that we just don’t know yet. It’s early days. The data is thin and they medical community is working hard to find the answers. You nor I have them.
Of course social distancing is useful for avoiding contagious viruses.
No-one has suggested otherwise.
But the Swedish study referenced by the Forbes article very much suggests that lockdowns are a waste of time, as I point out above.
Professor Johan Giesecke, former Chief Scientist of the European Center for Disease Prevention and Control (ECDC), further suggests that lockdowns are politically motivated rather than evidenced:
“We take into account the scientific evidence. Norway and Denmark have a more political leadership that want to show strength by implementing restrictions.”
Hmmmm……….
In the John Ioannidis interview above, at about 33 minutes in, he talks about the phenomenon of immunity without antibodies. He says that “preliminary data suggests” that perhaps the majority of younger people may be able to clear the virus without developing detectable antibodies.
https://off-guardian.org/2020/04/21/watch-perspectives-on-the-pandemic-4/
Therefore, you would think he might mention this in his recent serology study on the streets of Santa Clara. If the prevalence of antibodies shows that “the infection is much more widespread”, leading to revised epidemic and mortality projections, then presumably it may be even higher than calculated due to young people having been included in the tests.
Clearly, there is no hint of such shades of grey in the standard epidemiological models, nor in the calculations of what constitutes herd immunity. It is conceivable that policies could be based purely on prevalence of antibodies such as deciding when to lift the lockdown or the issuing of “immunity passports, or decisions on the need for a universal vaccine. It does seem to be a rather important detail.
Of the utmost importance, yet seemingly being only mentioned in asides and not being given the prominence it deserves in models and policy. Bizarre, and incredibly frustrating.
I wrote about this on TCW yesterday to try to raise awareness of the point but I don’t know if anyone is listening. https://conservativewoman.co.uk/we-might-be-more-immune-than-we-thought/
New York antibody results in https://twitter.com/NYGovCuomo/status/1253353516803993600:
13.9% across state, breaks down:
Long Island: 16.7%
NYC: 21.2%
Westchester/Rockland: 11.7%
Rest of state: 3.6%
These are all in the low range we’ve been discussing, even NYC.
Deaths plateaued only 9 days after lockdown suggesting the lockdown was too late to stop the spread. https://www1.nyc.gov/site/doh/covid/covid-19-data.page
It’s around that magic 15% again…
I’d like to see some actual proof that mild cases or younger people don’t produce antibodies… Do you know if age statistics of antibody studies have been published?
On the other hand even if this is the limit of the infection it puts the IFR at 0.5 – well within bad flu epidemics.
0.5% of the UK population of 66m with a infection ceiling of 65% is 215k deaths…
Remember, Flu has a much lower r0 0.9–2.1 (vs 1.4–5.7 for Covid), so has an infection ceiling of about 20%, which means on top of its CFR of 0.1% killing 1/5 the people it only needs to do so with 1/3 of the people they think Covid will to burn itself out.
On these measured numbers (not models) 15x more deaths than Flu.
Given de Gaulle shows you can measure at least 59% active Covid infections… we need:
1) substantive proof lockdowns don’t curtail spread, Sweden is looking dodgy this week…
2) proof serology testing misses big enough proportions of people who were infected to mean that the CFR is too high, and that we are much closer to the infection ceiling than they think in these places…
… for us to have a concrete case this is an over-reaction…
Check out swedenborg’s post on ‘latest news’ article here.
‘Today you can see the most likely day of maximum deaths is most likely 8th April or in the week after. There is a clear downward trend the last 5 days. The day of maximum new intensive care cases seems to be around 2nd April and it also seem to be a downward trend the last days. On the general website you can see that of the 2152 deaths (Swedish population 10 mil) 87% are over 70 years in fact a staggering 24% of all corona virus deaths are 90 years and over (Life expectancy Sweden 82 years). There are no deaths under 20 and only 36 deaths under 50.’
So explain to me why looking at Sweden’s total weekly measures up to and including yesterday’s numbers you get:
NewCases Deaths
Mar 27-Apr 02 2728 231
Apr 03-Apr 09 3573 485
Apr 10-Apr 16 3399 540
Apr 17-Apr 23 4215 688
… don’t believe the rhetoric of people who have a vested interest in proving themselves right… LOOK at the data.
Deaths will not fall if new cases don’t… looks to me like they have an r0 > 1.
Now it is slow growth, they may make herd immunity without busting through ICU limits, I have not forecast beyond this. I say good luck to them, but don’t fool yourself because they choose to fool you… they have not peaked.
Would it be fair to say that the curve is slowing though? I don’t see any evidence it’s about to go skyhigh in the next few days from the intensive care numbers.
Arguably keeping the disease on a ‘slowburner’ is the best way to get through.
I have no crystal ball, I’m just very keen we interpret data correctly.
Trouble is they’ve locked in their next 3 weeks already. We’ll just need to see what happens.
… but, the next Fri-Thu weekly total started cumulating today up on last Friday’s numbers at 812 new cases and 131 deaths, the former their largest ever single new day new case total… but, I agree possibly one day’s noise…
I’ve just checked your numbers Simon, I only make 400 deaths reported from the week ending 23rd thus far. Surely it’s fairly useless going by number reported on a particular day, which you appear to be – especially as Sweden has a huge break on weekends. It appears to me that a lot of the numbers of deaths in the last couple of days have been prior to the current week – at least 288 of them thus far. Ignoring the current week, as that is the most likely to be revised upwards, I get
Mar 27-Apr 02 318
Apr 03-Apr 09 609
Apr 10-Apr 16 673
We’ll have to wait another week to see how far up those 400 go – but if they’re anything like our own stats, by far the highest number are recorded from n- 2 where n is the current day. It tails off quite rapidly. I’d guess that we might see around 650 again.
but you’d also expect to see higher weekly numbers in the ‘plateau’ stage than in the week with the peak, as that’s been the case in all the other countries.
What are you talking about? I’ve given you weekly totals (not daily) entirely to remove the noise of their weekend reporting lag. Putting commas in to make it clearer for you…
Mar27-Apr02, 2728, 231
Apr03-Apr09, 3573, 485
Apr10-Apr16, 3399, 540
Apr17-Apr23, 4215, 688
As to ignoring the large numbers this week because you’re worried they’ll be reported UP!? WTF… the observation is only going to get worse so we can observe it now without fear…
I don’t see how you get the 17th to the 23rd to total up to 400…
Date NewCases Deaths
Fri17, 676, 67
Sat18, 606, 111
Sun19, 563, 29
Mon20, 392, 40
Tue21, 545, 185
Wed22, 682, 172
Thu23, 751, 84
… totalling up, and I really can’t make this any clearer,,,
Cases: 676+606+563+392+545+682+751=4215
Deaths: 67+111+29+40+185+172+84=688
Making a week Fri-Thu minimises the effect of their weekend lag… but, it really makes no difference which day of the week you use as cutoff as long as you do the same thing each week…
Where have you sourced your numbers?
My source worldometer:
https://www.worldometers.info/coronavirus/country/sweden/
From Swedenborg’s post, previously referenced:
‘The absolute majority is using the worldometer which is of very doubtful value especially as regards deaths which is not recorded on the day of death but reporting date.’
‘I recommend everyone to got to the Swedish Public Health website and you can see for yourself update each day https://experience.arcgis.com/experience/09f821667ce64bf7be6f9f87457ed9aa
When you see the map of Sweden on that web site you should click below on “information om datakallor”
There you would be able to download an excel file with daily updated figures for day of death (avlidna) and day of intensive care.
Today you can see the most likely day of maximum deaths is most likely 8th April or in the week after. There is a clear downward trend the last 5 days.’
Hope this helps.
I have sourced my numbers from the swedish health ministry. You should know that the numbers on worldometer are the total REPORTED that day. I added up the numbers in the graph on Avlidna/dag – deaths by day – from the swedish portal, linked below.
https://experience.arcgis.com/experience/09f821667ce64bf7be6f9f87457ed9aa?fbclid=IwAR2aLwJtRHm-CPooa1q1S8La8TsFq01u2RyS5uIyNCHiGYVHq1pvjMM_yt4
This is obviously a better metric because they are adding in deaths in hospices & care homes when they happened (the lag for them is greater)
See this chart: https://twitter.com/gizhkoandrii/status/1253658265310957573
Now it’s quite obvious that the tailing off is a product of the lag for reporting, but it’s also quite obvious that they have been at a plateau for a while.
The worldometer numbers are from their official hospital daily mortality numbers, like ours they are always LOWER than ONS (which is what this site is the equivalent of), and these all-mortality stats always lag hospital stats.
I think it morally quetionable given their last offical update date was to the 12th to show partial numbers beyond that will always tail off to very small numbers in recent days. It leads to misleading rhetoric like your twitter post.
You’re also wasting our time here, the data lags by up to 2 weeks, and in the end makes no difference for measuring the trend. BOTH yours and my totals for the 3rd-9th and 10th-17th show the same 11% increase in deaths not a plateau.
Regardless, my entire point was about NEW CASES, for these two ranges it looked as if new cases HAD plateaued in Sweden. BUT, this week new cases are up, 25%, and there is no rhetoric or dramatic change in their testing strategy this week, which means this is r0>1, these are from infections 15-20 days ago and in 2 weeks deaths will go up too. This needs watching this week.
No, their worldometer numbers are lifted directly from the total increase each day on the site I link to above. Not my twitter post, and he adds in (in grey) the average increase expected, which is not unreasonable.
I don’t see a huge increase in ‘Sjukdomsfall per dag’ beyond previous weeks?
It would appear they also backdate cases/day to the day the case was reported – for example Worldometer has 17/04/20 at 676, whilst the government portal has it at 688. The last 4 days are quite high, so we might be seeing an increase in cases, but it’s too early to tell.
If you think that numerically it will lead to a big difference, it won’t. Sweden have marked all figures before the 12th closed, which means all the people recording deaths (and cases) are always trying to make sure that they clear backlogs in 2 weeks. Looking at one overnight sample… Date, 24th, 25th, %change Apr12, 96, 97, 2.5% Apr13, 84, 84, 0.0% Apr14, 91, 90, -2.5% Apr15, 108, 109, 2.5% Apr16, 107, 107, 0.0% Apr17, 72, 74, 5.0% Apr18, 72, 73, 2.5% Apr19, 73, 76, 7.5% Apr20, 70, 70, 0.0% Apr21, 42, 46, 10.0% Apr22, 40, 43, 7.5% Apr23, 31, 41, 25.0% Apr24, 3, 19, 40.0% … 93% of the change (40) affected the last 7 days. This daily measure totalled up to a week, will reflect in the same proportion the number of deaths added to the last week, it essentially does a bit of smoothing at the same time. There will be the same total number of deaths reported either way. In my view the arcgis representation is a morally far more questionable, their plot of it will always tail of to low numbers more recently over the last 2 weeks making it always look like deaths have… Read more »
New infections are a product of more testing. Antibody tests show there is a great deal more of the virus winging around, doubtless started a great deal earlier, than anyone had appreciated when they started modelling. Mortality figures are ‘with’ not ‘from’. The data is junk. As I keep repeating, there is no international consistency in recording cause of death so comparisons are pointless.
It is not a measured number of 65% ceiling except in one context. (Also you’re assuming all infected develop antibodies.)
Hi Simon, I didn’t mean to compare with normal flu, but with particularly nasty strains, which can come up higher than that in CFR. What was the CFR of the Hong Kong Flu? This paper:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3357941/
estimates it at 0.5%, and Spanish Flu at 2%:
“The CFR of Spanish influenza is sometimes thought to be approximately 2.0% [5, 24] and those of Asian and Hong Kong influenza pandemics are thought to be up to 0.5%”
Swiss doctor has this:
The German Network for Evidence-Based Medicine reports that the lethality of a severe seasonal influenza (flu) such as 2017/2018 is estimated by the German Robert Koch Institute [government] to be 0.4% to 0.5%, and not only 0.1% as previously assumed. https://www.ebm-netzwerk.de/en/publications/covid-19
… but the important bit is:
“In 2017/18, 25,100 people died of influenza in Germany [12]. This death rate corresponds to 5 million infected people based on the CFR of 0.5% calculated by the RKI for 2017/18. According to the RKI report on the epidemiology of influenza in Germany in 2017/2018, the influenza season lasted 15 weeks, from the 52nd calendar week of 2017 to the 14th calendar week of 2018 [11]. In order to reach 5 million within 15 weeks, the number of infected people would have to double every 4.4 days”
Which means the natural infection ceiling with no mitigation measures at all was 5m/83m = 6% (not 65%) making the r0 about 1.08, so CFR the same, but was 12x less infectious so it did not kill 300k people…
We are talking 1968!? You’ve instantly promoted Covid from seasonal Flu comparsions to once in 100 year events.
From that study, the r0 of Hong Kong Flu was measured to be similar to that of seasonal flu in the first wave 1.06–2.06, and slightly higher in the second 1.21–3.58. The second wave look between seasonal Flu and Covid.
This means it will have burned out at a lower infection ceiling infecting fewer people in the process. So as much as they have the CFR, there would just have been fewer caes with out mitigation strategies… at the moment we’ve certainly lowered the r0 with hand washing/social distancing and made it more like this.
No going to the pub, restaurant or club though…
Did they shut down in 1968? Was there mass hysteria? I think Covid might be on a par with 1968 (possibly) but the lockdown is still a huge problem.
As I’ve already said it is not just about having equal CFR, you have to factor in that a lower r0 creates a lower infection ceiling too…
Hong Kong Flu killed 1m people globally:
– the global population was 1/3 of what it is now
– we were far less interconnected
– and it had a lower r0, probably infecting 1/3 to 1/2 of people Covid might
All your telling us so far is that Covid mortality globally could be in the region of 6-9m when seasonal flu only kill about 650k.
Higher numbers in CDC homeless shelter study: San Francisco 66% residents; 16% staff, Boston 36%; 30% (both with big Chinese communities), lower elsewhere.
https://www.cdc.gov/mmwr/volumes/69/wr/mm6917e1.htm?s_cid=mm6917e1_w
The modelling assumptions (on which the now plainly silly lockdown is based) are looking less and less sensible as more and more data comes in.
Added to which, U.S. Covid 19 mortality rates are unreliable:
‘At Tuesday’s White House coronavirus press conference, task force member Dr. Deborah Birx said that while some countries are reporting coronavirus fatality numbers differently, in the U.S. you are counted as a victim of the pandemic if you die while testing positive for the virus, even if something else causes your death.’
https://www.realclearpolitics.com/video/2020/04/08/dr_birx_unlike_some_countries_if_someone_dies_with_covid-19_we_are_counting_that_as_a_covid-19_death.html
CFR was 0.5% so bang in the middle of Gangelt (0.37%) and Robbio (0.7%) for a healthcare response that creaked more than Germany’s, but not as badly as Italy’s, so consistent, and makes John Ioannidis numbers look even more like an outlier… but, to be expected given he used estimates and had sample bias. I don’t understand how you make it 9 days? I think you’re misremembering what they did… they closed everything on the 16th and then locked down like Italy (don’t leave your house) on the 22nd. As this article on the 18th discussed from the 9th crowds were thinning out… https://www.bbc.co.uk/news/business-51880799 This piece really shows the fall off in subway travel from the 9th… https://talkingpointsmemo.com/edblog/when-did-new-york-city-shut-down From your link: hospitalisation plateaued 30th-6th, and deaths peaked on the 7th Apr. Even from the 22nd to the 7th that is 17days to deaths peaking. BUT, given they closed everything to the same level we have on 16th, and subway travel had more halved by the 14th it is easily 3 weeks since r0 would have been heavily impacted by their measures. So you can’t dismiss the lockdown as having caused it. Entirely concede we need somewhere without a lockdown to… Read more »
The NYC hospitalisation plateau (drastic slowdown in growth rate) begins on 25 March, surely too soon after the 22 March lockdown to be attributed to it? And then there’s Switzerland and Germany doing similar. And Sweden of course. Surely adds up to a strong case.
Their shutdown in terms of subway rides began on the 9th down 10%, by the 13th down 35%, by the 14th they were at 1/3 of the daily journeys, sliding down from then to a 1/4 by the 20th, and at a roughly 1/10 beyond that. Incubation period of 5 days, expect hospitalisations to peak 10 days from mitigation measures, which they do by the 30th, but it records 3 further days equally high admissions days through till the 6th. Your assertion of the 25th is that of a novice. It does briefly plateau and dip 28th/29th, but these are a Sat/Sun, and if you look the week before on the 21st/22nd the admissions numbers dip too, in fact if you look at the plot all Sat/Sun days are lower, suggesting some kind of weekend reporting periodicity. I do this for a living, I’m a data quant, I look for trends and patterns in sources of data, and work out how to predict statistically likely future events from them. Germany, by the survey you sent me, when we actually analysed the data had only got it’s r0 to 1.8 before the full lockdown, and this had only happen from 2.4… Read more »
What numbers are you suggesting as average time from infection – hospitalization – death? I’ve seen longer numbers than 10 days (6 day average from infection – > symptoms then 7 days for symptoms -> hospitalization.) I’ve also seen 22-24 days from infection to death. Would you broadly agree with those?
I’d also like to know whether you think the same with the UK data? To me it looks like the peak of deaths (8th April) was about 5 days too early for the lock-down to have had much effect, assuming the 22-24 days is correct, and that previous voluntary efforts were sufficient to cause cases to peak. Do you disagree?
I was going on 5 days incubation, 10days to present, 10days till death, as averages I’d read from various places. Hospitalisations start from 10 days in peaking at 15 days.
So we have reduction in numbers from the 9th, peak hospitalisation at the 30th with a slowing ftom the 25th, a plateau to the 6th, then peaks deaths at the 7th.
Tracing back 25 days you get to the 14th, which seems perfectly plausible to me as the lockdown taking effect.
But the lockdown has to coincide with the slowdown of infections not the peak to be regarded as the cause. And while I take your point about subway rides declining from 13th that was far from lockdown, with much remaining open.
Patrick Vallance has said a 75% disruption to community movement is all that is needed to get r0 < 1. This is not digital, so any reduction up to that will slow the rate of hospitalisation.
7th/8th a weekend saw 45% of journeys, people enjoying time at home.
Mon 9th at 90% … dropping to 64% by that Fri 13th (caused by social distancing measures).
This 7th-13th range 15 days later would be the 22nd-28th, exactly the range you are saying saw a slowing in admissions, upto the peak on the 30th then a plateau through to the 6th.
Yes, it is not conclusive proof lockdowns work, as it could be an entire coincidence, but if anything it shows it working exactly as they intended.
But social distancing that stops short of lockdown to flatten the curve is what many lockdown sceptics advocate. So this analysis, crediting the pre lockdown social distancing, supports their position not the lockdown zealot position.
Will, that does not follow. No one denies social distancing will lower r0, the question is simply whether for a particular population it lowers it enough to avoid more invasive measures. One size will not and does not have to fit all. In my view the evidence from around the world seems to be pointing toward population density being the biggest deciding factor. Sweden is 1/11 our density, looking at cities… NewYork: 10k people per square km London: 8.6k Stockholm: 4.8k … and from the looks the rate of virus spread pre any measures London and NY were growing too quickly and they needed the extra measure to get their r0<1. As to Sweden, they were growing slower and so far have got away with not needing the extra measure. That said it looks to me like they have a very slightly positive r0, say 1.04, and have seen a 25% growth in new cases in the last week. They might be able to get that under control again without new guidance and no lockdown, good luck to them. Great if they can ride to herd immunity without needing to do so, we just need to get ourselves to the… Read more »
Mistake! … should read…
“I simply don’t agree that the peak, plateau and fall off between the 30th and the 6th CAN be attributed to these lighter measures.”
I agree population density appears to be a big driver of rate of spread. I imagine it will drive the overall infection rate as well. I also imagine air quality and population healthiness (eg obesity), among other things, will affect mortality rate. How do you know what the case growth rate is in Sweden? It obviously isn’t the reported case numbers which are just a reflection of the number of positive tests. Is there a reliable data source? This sounds reasonable: ‘In my view London and NY (and some other UK cities) needed their lockdowns to create time to think and plan, the rest of the UK, maybe not so much, but it is hard to sustain different policies like that in terms of people doing it.’ But when you think about what that costs you have to ask whether the cost can be justified. If the health service can be boosted to cope, what is actually gained that can possibly justify the immense cost of lockdown? You seem incredulous that lighter measures in the UK/London could reduce the reproduction rate to below 1. Yet peak UK deaths occurred on 10 April, 21 days before that is 20 March –… Read more »
Will, choose your experts better, that intepretation of London is as flawed as the one for NY…
… tube journeys started to decline heavily on the 10th of March… down to 50% by the 15th… 25% by the 20th… even from the 20th that is 3 weeks to the 10th.
… kick the tyres on other people’s theories before taking them as your own when they are clearly not very good analysts… you’re filling the debate with falsehood and I’m having to spend far more time than I would like keeping us honest, and spending far less time being able to research real data on lockdown mortality.
In fact the CEBM data with deaths on day of death shows that the UK peak was on 8 April and London’s on 4 April. That suggests London’s infection peak occurred on 14 March, 10 days before lockdown.
… what, matching the day by which tube travel in London had halved?
Please think about what you’re saying.
If only obeying social distancing caused the London peak, then on the 14th the data would need to show that 100% of people were still travelling to work and only following those rules… BUT, it clearly does not.
You’re ignoring data staring you in the face.
Even by the 14th 50% of tube users had clearly changed behaviour before the official lockdown, people thought Boris was being wreckless, they had self-isolated themselves working from home.
It peaked around that day and was slowing before it. If you think that fact supports lockdown then that’s up to you.
Will, I merely think the data doesn’t disprove it could have. For me two things remain possible from the London and NY data.
1) the lockdown had the effect
2) errors in antibody surveys mean these populations were hitting their infection ceilings at the same time, and the lockdown was conincidental.
I’m pulling you up for trying to make conclusive statements about 1, that I would not put money on.
What is increasingly frustrating me, is you’re occupying all of our time discussing what you’d like to be true, but in the data plainly isn’t.
If you want people reading this site looking to see if we are credible to judge we are, you need to make sure you really kick the tyres on what you’re saying, before you say it.
I’d like to be able to look at information you’re posting and trust your judgement on it, and I can’t.
Isn’t this the point though? Voluntary action had a strong effect? I don’t think any (sensible) lockdown skeptic would suggest otherwise; Sweden might be more normal, but it’s hardly fully normal.
Why do you suggest that 100% of people following the rules is necessary? What if it only took 50% reduction in public transport to lower the R0?
We should be encouraging people to work from home, reduce public transport. We shouldn’t be putting them under house arrest, we shouldn’t have enabled a police state, we shouldn’t have crashed the service economy (where social distancing was possible, as it is in most highstreet shops, garden centers, restaraunts.) We shouldn’t have closed the schools – there’s relatively little data that this changes much, especially in the case of Coronavirus.
We should have continued to quarantine anybody with symptoms
And we should certainly be interrogating the science around outside transmission.
I don’t suggest that. I was just pointing out the conditions that would have needed to be present in the data for me to take the interpretation that Will has.
I do think a 50% reduction is enough to significantly impact transmission, they recon 75% is enough to get r0<1. I don't think 50% of people no longer going to work on the tube represents following just the advice of washing hands and keeping 2m apart, which is what I take Will to mean when he says "the lockdown definitely had no effect".
Maybe he mean't something else.
According to your graph above the tube was at 60% on 14 March and other forms of transport had barely reduced at all. And that was the peak – it slowed before that. I regard this as possibly showing the power of some modest social distancing – assuming it wouldn’t have happened anyway. It certainly speaks against the necessity of anything stronger.
Any idea where the data on London hospital admissions is published? I see it referred to in various places but haven’t been able to find it.
No need to be rude! We’re on the same side here and if you disagree then it doesn’t need to come with insults or accusations of dishonesty.
I can recognise a drastic reduction in growth rate when I see it and that occurs around the 25th, starting five days of similar numbers after a period of fast growth.
We have discussed Switzerland before. This is the link. https://bsse.ethz.ch/cevo/research/sars-cov-2/real-time-monitoring-in-switzerland.html
You said: ‘I agree Switzerland’s paper is puzzling. On worldometer, they certainly seems to continue to see new case growth like Germany until the 28th, but the paper does seem to show r0 dropping to 1 around the 15th. It would require much deeping digging, understanding all the measures they had already introduced, but they weren’t in anyway “running around freely” state for the few weeks before the 15th, and we don’t really know if testing strategy changed, or exactly what happened, but things do look inconsistent.’
My apologise, even to the 25th, we are still talking 10 days back inclusive being to the 16th inclusive at which point subway journeys were down to 25%, and Patrick Vallance talked about 75% being the disruption level needed to get r0 < 1. Given the peak lasts to the 6th, it seems entirely plausible the lockdown could take credit.
I just don't think it is conclusive evidence.
Apologise again, I recall Switzerland now, worldometer seemed to support lockdown, the paper was confusing and did not. I think we need more evidence (are there any other studies?) to be sure…
Sweden is the centre of the lockdown focus now, they have locked in the next 3 weeks, we just need to see what happens.
In a spirit of helpfulness (from Swedenborg’s post):
‘One of the most disputed issue on twitter threads or blogs is the reporting from Sweden. The absolute majority is using the worldometer which is of very doubtful value especially as regards deaths which is not recorded on the day of death but reporting date. The most important figures are always the date of maximum deaths and date of maximum use of intensive care. There is a predicted time lag from infection peak to COVID deaths peak of approximately 21 to 28 days. The peak of intensive care is most likely a week before peak of deaths.
I recommend everyone to got to the Swedish Public Health website and you can see for yourself update each day https://experience.arcgis.com/experience/09f821667ce64bf7be6f9f87457ed9aa
When you see the map of Sweden on that web site you should click below on “information om datakallor”
There you would be able to download an excel file with daily updated figures for day of death (avlidna) and day of intensive care.
Today you can see the most likely day of maximum deaths is most likely 8th April or in the week after. There is a clear downward trend the last 5 days.’
My point was about new cases… and this ticking up again this week… and your website shows exactly the same problem. As to the entirely secondary you have created about deaths, assessing the general direction of travel from hospital daily figures by assessing weekly totals is entirely reaonsable, and as per the other post both sets of numbers when used to just look at direction of travel show the same 11% increase. I repeat again, that website (for deaths) have marked numbers up to about the 12th as official. They have a similar 2 week lag to the ONS in gathering all the data, the ONS has only reported to the 10th. From then owards up to now daily counts will always tail of to zero as reporting from different regions comes in at different rates… I think it is morally very questionable for them to display partial figures as it will ALWAYS make it look like deaths have peaked. Which is entirely the trap you have fallen in to. The ONS don’t do such a live stream for exactly this reason. It just isn’t honest about data, and for them to actively say it is a better source is… Read more »
Well done for playing devil’s advocate, providing good perspective on all the numbers. Very much appreciated. The state media narrative shifted this morning, big time, so I think the panic is over. All that remains is a politically stage managed retreat from lockdown to grab headlines now likely to be full of mud hurled in the general direction of the government, much of it, alas, well merited.
Thanks, I just want arguments to be sound.
I have no issues with making sure they have their numbers right on mortality, etc, but once it is clear that they aren’t fooling themselves it is really a case of working out the impact of the unintended consequences and demonstrating we may have the wrong lesser of two evils.
Caswell. I’ve noticed that antibody prevalence seems to vary in surveys so far between a few per cent up to a third of the population. In NYC for example it was 21 per cent but in NY state overall around 14 per cent. It seems to be higher in denser population centres. I imagine that different communities have different curve profiles for the virus not just in terms of rate of growth but also height of peak. I guess it’s because different conditions interact with pre-existing resistance differently to give different requirements for herd immunity. So a city like New York has a higher requirement than the wider state. Is there any way to reflect this in the modelling?
Hi Will. I think the Imperial College model already has this structure in place: “… individuals reside in areas defined by high-resolution population density data. Contacts with other individuals in the population are made within the household, at school, in the workplace and in the wider community. Census data were used to define the age and household distribution size. Data on average class sizes and staff-student ratios were used to generate a synthetic population of schools distributed proportional to local population density. Data on the distribution of workplace size was used to generate workplaces with commuting distance data used to locate workplaces appropriately across the population. Individuals are assigned to each of these locations at the start of the simulation.” The aim of such a simulation might be for effects such as those due to population density to fall out of the model i.e. to actually see different levels of immunity developing for people within cities. However, this would depend on realistic core assumptions, such as the type of illness varying with viral load. Different ways of living (e.g. urban versus rural) might then produce different levels of immunity once the epidemic was trailiing off. But if the core assumptions… Read more »
Thanks Caswell. Now that much more data is available there should be much more effort put in to developing models that conform to the data. As you say, instead there just seems to be a stubborn sticking with the seemingly simplistic models already developed. Or rather, if they are being refined or revised in light of the real world situation then I have not seen any evidence of that. It looks to me from the antibody surveys (and antigen testing) that have come back so far that the basic assumption that herd immunity requires 60 per cent antibodies (and the virus will continue spreading in every community until it reaches that point) may be incorrect. Instead it may be that communities with different profiles tend to peak at different infection rates, varying between perhaps a few per cent up to, well, the highest result so far is around 33 per cent in a district in Boston I think. What would be helpful is if someone was collecting all the antibody results that come in along with the profile of the community they are from and trying to develop models that produce those levels of antibodies in those communities (taking into… Read more »
The model I developed does contain the idea of type of illness being related to viral load – but that’s because I didn’t ‘know any better’, having come to this without reading anything at all about epidemiological modelling.
It really does seem to me that there’s a barrier between the immunologists and the modellers. The immunologists know there’s a lot of unknowns, even regarding established viruses, and they may be all too pleased to look at the beautiful apparent certainty that the modellers produce. And the modellers need something definite to work with that can’t be challenged so they stick to the ‘tried-and-tested’ models.
Everyone seems happy with the arrangement.
Susceptibility!
If you assume that 1/3 of the population, i.e. young people, is naturally “immune” (or something approximating to immune); doesn’t develop measurable quantities of antibodies; and that this proportion is also not contagious when it gets exposed, then it follows that herd immunity is achieved at around 33% antibody-positive for an overall R0 of 3.
Yes indeed. And it could vary in different places depending on the conditions such as population density and age profile. But all I hear is people parroting the 60 per cent figure as though it is a solid datum rather than an estimate based on a model based on assumptions.
The 60% figure is just the estimate of the average herd immunity ceiling for a pathogen with a measured initial r0 of 2.6. As per Caswell’s observation Ferguson’s model already varies the strength of the r0 for different parts of a country by taking detailed demographic data and modelling that those outside a city like NY would see lower penetration and a lower ceiling, but I’d imagine in NY city higher penetration and a ceiling more like 70%. Creating an average of 60%. He seems to have data on things like classroom and office density. He has really gone to town on this. Bear in mind this would also allow him knowing the demographics of these sub populations and the distribution of CFR by age to predict deaths in each setting in a similarly more accurate way. I’d imagine this will have been fit out of years of research looking at antibody studies and the densities brought about by different pathogens in different settings. Getting that data and replicating would I’d imagine take years. Add to that he can probably calabrate any r0 reading given to him based on the demographics of the sample, so he can probably even make… Read more »
“We simply don’t have scientific proof infections are happening without antibody response” I rather suspect we won’t get that proof (if ever) until long after the epidemic has passed, difficult to prove a negative etc. although we do seem to know 95% for sure that younger people don’t generally get ill and aren’t very contagious. What we might get is a much better idea which model variables fit the pattern. My non-expert meta analysis of sero studies to date throws up lots of data suggestive of a significant proportion of prior immunity (either with or without antibodies) and very little data suggestive of anything approaching 100% susceptibility. This also fits with what we know about pretty much every other virus; even with HIV there’s a natural immunity of c. 10%. E.g. in the worst hit towns in Bergamo, approx 61% found to have antibodies but the area was absolutely overrun and the natural herd immunity threshold would have been overshot by a mile; it’s far from inconceivable that everyone who was susceptible was affected. Similar on the Charles de Gaulle, and in the hotspots in the US. In other areas, a seroprevalance of c. 15% seems to lead to a… Read more »
I believe we can already calculate the IFR for healthy under 65s. And we can do so with a high degree of confidence, using evidence already published and with no need for further widespread antibody testing. If I’m right – and I’d love people’s opinions on this – then the argument that this cohort – which also happens to be the economic engine room of any nation – should, with a few sensible precautions, be set free to go back to work and normal life, becomes powerful. My approach is neither revolutionary nor particularly insightful, I just decided that rather than having to know the IFR for the general population I would simply choose one of the more pessimistic IFR estimates – the 0.9% IFR assumption of the Imperial college group. From there it is very straight-forward to calculate the IFR for under 65s. Using data produced by Dr Ioniddis’ Stanford team on death’s of healthy under 65 in Netherlands, Italy and New York, and also the actual number of deaths in the UK of people with Covid-19 (itself an inflated representation of deaths from Covid-19) we can extrapolate the following: As of April 23rd the number of deaths associated… Read more »
One factor I have not seen discussed anywhere is the effect of people being told there is a deadly virus circulating among us. Particularly for older people, who tend to be more likely to believe what the established public authorities tell them, relative to other demographic sub-groups, and equally more physically and mentally fragile, i.e. more likely to be strongly affected by such effects. In the Chernobyl melt-down, for example, the fatality predictions ranged from thousands (immediate effects) to hundreds of thousands to millions of secondary deaths (CNN 1995: 3.5 million), whereas the actual deaths caused by that event was something like 56 (different sources make different claims – it’s not possible to be absolutely certain), and over the course of succeeding decades several thousands have died from the secondary effects of the event, e.g. the pollution of the local water and food supplies, and etc. Whatever the actual numbers, what is reasonably well established is that: 1. the actual numbers are several orders of magnitude below the direst predictions. (Sound familiar?) 2. a significant proportion of the dead ended up that way because they had been told by public authorities that they were going to die. (This is more… Read more »
Is this of interest? “Memory T cells undergo different changes and play different roles in different life stages for humans. At birth and early childhood, T cells in the peripheral blood are mainly naive T cells. Through frequent antigen exposure, the population of memory T cells accumulates. This is the memory generation stage, which lasts from birth to about 20-25 years old, when our immune system encounter the greatest number of new antigen. During the memory homeostasis stage that comes next, the number of memory T cells plateaus and is stabilized by homeostatic maintenance. At this stage, the immune response shifts more towards maintaining homeostasis since few new antigens are encountered. Tumor surveillance also becomes important at this stage. At later stages of life, at about 65-70 years of age, immunosenescence stage comes, in which stage immune dysregulation, decline in T cell functionality and increased susceptibility to pathogens are observed… …Currently, the mechanism behind memory T cell maintenance is not fully understood….” https://en.wikipedia.org/wiki/Memory_T_cell If I could bold the text, I’d be highlighting the bits that talk about three distinct stages of life: 0-25 years, 25-70, post-70. I’m very ignorant of this stuff, but as I understand it, T-cells are the… Read more »
Someone has finally done a model that includes susceptibility – and it confirms yours. From the Speccie:
Research: Covid-19 vulnerabilities
How many people would have to fight off Covid-19 to achieve collective (or ‘herd’) immunity? So far, we have heard that it’s around 60 per cent. But this assumes a uniform population with everyone mingling equally and being equally at risk of catching the virus. A new study https://www.medrxiv.org/content/10.1101/2020.04.27.20081893v1, led by academics at the Liverpool School of Tropical Medicine, adjusts for the fact that they don’t and they aren’t. The virus, they argue, could quickly infect the more susceptible part of the population – which will then become immune and stop spreading the disease. As Matt Ridley puts it, ‘if the virus runs out of highly-susceptible segments of the population (elderly, hospital settings etc), it may struggle to keep going in the rest of the population’. Adjusting for this, says the study, and the threshold for Covid-19 herd immunity falls to between 10 per cent and 20 per cent.
Hi Will. I just saw it in the Spectator and was about to post it here!
So it seems quite lot of people are now arguing that lockdowns are unnecessary but voluntary social distancing is what does the trick eg https://www.thepublicdiscourse.com/2020/04/62837/ https://arxiv.org/abs/2004.10324. I looked at London and New York to test this. London deaths peaked on 4 April and began to plateau on 27 March. Assuming average 16 days from infection to death (I used to say 21 but I now understand the latest data suggests 16) that puts peak infection on 19 March and the beginning of plateau on 11 March. The thing is, that matches pretty well when social distancing began. Almost exactly in fact – see graph here https://dailysceptic.org/2020/05/02/. New York deaths peaked on 6 April and hospitalisations peaked on 1 April. Hospitalisations began to plateau on 26 March so we infer infections began to plateau 11 days prior on 15 March. Again, that corresponds almost exactly to the beginning of social distancing – see the NYC graph here https://www.thepublicdiscourse.com/2020/04/62837/. I had hoped this exercise would show that the peaking and plateauing of infections preceded social distancing and hence was due to some sort of saturation point/herd immunity. But it ended up strengthening the case for social distancing, leaving open the widely-believed possibility… Read more »
A new interview with Hendrik Streeck that’s very relevant to what we’re talking about here.
https://www.youtube.com/watch?v=vrL9QKGQrWk
It’s all good, but very interesting from my point of view at about 5.15 and several other places where he’s talking about the possibility that lower exposure (and lower viral load) is likely to produce fewer symptoms and lead to partial immunity/resistance, possibly being based on T-cell response.
Did you see that the final results for NYC found 19.9% had antibodies? Within the Liverpool model range of 10-20%. https://www.governor.ny.gov/news/amid-ongoing-covid-19-pandemic-governor-cuomo-announces-results-completed-antibody-testing
Have you had the chance look at the spreadsheet that someone made of the surveys done so far? https://docs.google.com/spreadsheets/d/1zC3kW1sMu0sjnT_vP1sh4zL0tF6fIHbA6fcG5RQdqSc/edit#gid=0
Interesting that most are within the Liverpool model range, except for a few that are of smaller communities such as prisons or schools that had a bit (or a lot) higher. Four that stood out to me were number 9 in Iran (33%), but that turned out to use households rather than individuals so that would inflate it (they’ll catch it off each other), number 11 in Italy (67%) but that seemed to be an estimate based on assumptions rather than an actual survey, number 44 also in Italy (70%), which I’m not sure about as I can’t read the article but seemed odd and to do with blood donors (were the tests accurate?), and number 28 in Chelsea in Boston (33%) which does seem to show a badly affected neighbourhood with a higher proportion infected.
It doesn’t have the surveys from boats so it’s not complete.
What are your thoughts?
I’m puzzled.
Some here are suggesting that a large proportion of the population have been quietly infected without adverse effects.
Others are proposing an ‘infection ceiling’ which blocks further spread … a sort of super herd immunity.
I have seen serological based immunity estimates for various countries ranging from 1% – 15%. The top values were for mega hot spots : one in Germany and NYC. Certainly nothing like 30% or more immunity has been reported as far as I have seen.
As for the idea of a magic infection ceiling, this site is the only place where I have seen this discussed.
I suspect that a lot of wishful thinking is going on.
Actually there’s been a recent paper from Liverpool which deals with exactly the sort of questions we have here. Populations are not homogenous, diseases do not spread through everyone equally. There may be a number of reasons for this, prior immunity, genetic immunity – those we just don’t have any data on. But social behaviour is a big part of this. Liverpool basically argue that we only need 10-20% infection because ‘superspreaders’ and the vulnerable – in care homes and hospitals where it’s hard to prevent spread – get it first. After they have got it, the effective reproduction rate drops much faster.
My understanding, such as it is, is that most people just have sufficient infection resistance to fight off covid 19 before ever it gets established. Even the black death only killed half the population I think. I can’t imagine that half of them were somehow never exposed to it.
Yeah it’s all speculation by the ill-informed. We don’t know how many ppl have had it, because we can’t know: current numbers are based on hospital admissions, then hospital staff, then care home data. We don’t know how many ppl have antibodies, because the tests are still being refined. We don’t know what proportion of ppl have immunity. We don’t know whether antibodies, if they can be detected, confer immunity. We don’t know whether asymptomatic infection leads to immunity. My contribution to the speculation: 5% of ppl have had it and most will be immune. 95% of ppl still to have it. It’s a stretch to argue that this virus will evolve to lesser virulence on Darwinian grounds, a) because most infections are minimally symptomatic, b) because even for more serious infections it’s not the virus that kills you it’s the day 8 immune response, and by then you have passed the infection to 3 or 4 others. It’s not like flu, or plague. My guess is any release of lockdown will lead to resurgent cases – duh – we just need to decide as a society whether we want to succumb to the consequences of lockdown (economic meltdown, indirect… Read more »
One of the questions that everyone should be asking is how effective is the lockdown? I guess it is hard to answer as we don’t yet really know quite how fast covid 19 would have spread if the lockdown had not happened. But we do know how fast normal flu viruses spread from many years of experience. If I am right in thinking that the transmission method of covid 19 is similar to ordinary flu viruses could we not check the success of the lockdown by following the incidence of ‘flu’? It looks almost unchanged to me on the ONS page, although the exact numbers are hard to seperate out from the data we are given. I have looked quite hard on the internet and I havn’t found any comment on this yet, has anyone else?
We really do now have enough clues to estimate the number of people infected in UK. Fergusson estimated 3-5% had been infected at start of April, with a doubling time that was in the process of falling from 3 days to 10 days due to lockdown through that month. This suggests 20% or more infected to date (and Fergusson wants this number low!) Italy had 3% in villages where widespread testing occurred around mid-March, and we’re ~1 week behind them with a doubling time of 3 days back then. This suggests 50% or more infected to date. UK Governments own testing data for Pillar 2, which is now almost all about testing worried well not symptomatic key workers) is showing ~5% currently infected. Assuming virus clearance in 3 weeks or so, would reduce current point prevalence of infection by 5% per day (or far less if ongoing transmission is occurring). Guven R0 is now 0.5-0.9, we can estimate infection rate will be falling perhaps 2% per day. Extrapolating back gives us a peak point prevalence for end March (when the effect of lockdown kicked in and started reducing prevalence in UK datasets) of 20-25%. Adding in all infections that would… Read more »
Do you think that when Prof Sunetra Gupta makes assumptions, considering that she has published over 100 papers since 1994, that the difference between her assumptions and those of the less experienced members of the epidemiological academitribe, would be a bit like the difference between the assumptions I make with 35 years driving experience, and those of a learner driver
Errors hopefully become less over time, as the assumptions become more useful and applicable
https://www.zoo.ox.ac.uk/people/professor-sunetra-gupta
I don’t understand why Ferguson would be invited to be a member of SAGE and not her. Having watched one of her lectures on Youtube I think she has the exactly the right insights for the job.
Someone has confirmed the Liverpool study: https://judithcurry.com/2020/05/10/why-herd-immunity-to-covid-19-is-reached-much-earlier-than-thought “Incorporating, in a reasonable manner, inhomogeneity in susceptibility and infectivity in a standard SEIR epidemiological model, rather than assuming a homogeneous population, causes a very major reduction in the herd immunity threshold, and also in the ultimate infection level if the epidemic thereafter follows an unconstrained path. Therefore, the number of fatalities involved in achieving herd immunity is much lower than it would otherwise be.” Of course, when I say “confirmed”, I don’t mean that they have confirmed that real people’s susceptibility or infectiousness are more variable than previously thought. Nor that the biological mechanisms involved have been identified. All that has been established is that modelled variability changes the model’s dynamics. But I think this is showing modelling at its best. Where detailed biological mechanisms are almost impossible to examine, measure and quantify quickly (not least because a virus and people are moving targets), all we have are snippets of population-level information. Here, the model is modified in a perfectly plausible manner and shown to match those snippets of information. The result is counterintuitive and would not have been apparent without running the model. The model has served its purpose of testing… Read more »
Yes, I believe the research showing about one third of humans are immune to C19 (due to T cell immunity from prior corona virus infections) and the models are very important. Especially when combined with the fact that the IFR of C19 is only about 0.2%. This is because this scientific evidence and the modelling together confirm that herd immunity can be achieved with a low percentage of infected and with low, flu level, mortality. Here in New Zealand, I am expecting a fresh outbreak this winter or next as we have hermetically sealed our border and it is now summer. This bug does not like UV. The evidence and models are clear. We could have left our border wide open, like Belarus, and had about 60 deaths with a quick eight week outbreak, as Wittkowski recommends. Instead, we have had 20 deaths now, and will probably have 40 deaths in the future. We have thus unnecessarily destroyed our economy. Saint Jacinta, our PM, is not going to open our tourism based economy for months. The bug, I believe, will return when more susceptibles are “created” ie the very old with co-morbidities. The lockdowns do not, I think reduce the… Read more »
Re today’s announcement that 0.27% of UK were carrying the virus in last weeks or so: * This survey only tested people from families who have previously volunteered to be part of ONS research. So probably not the poorer or less well educated members of society. And it excluded all health care families, and all key worker families. I.e., those tested were not typical individuals, but the kind of people who would be able and likely to keep themselves very protected throughout this lockdown. * The test is not highly sensitive (i.e., misses many affected) and samples were taken at home by test subjects without expert help, so could easily have been done badly (missed virus carry material). * Indeed, the government openly admit the 0.27% is ‘test positives’ not infection prevalence, because of the above reasons. * Given the above, I’m going to assume the more general recent infection rate is about 1% * Subsequent to lockdown starting (23 March) and R0 falling to 0.5-0.9 (governments numbers) the percent of the population would fall daily. Assuming 2 week or so is needed to clear the infection, then the infection prevalence would fall by as much as 50% every 2… Read more »
Even more evidence…
Infection prevalence is now LOWER in London that elsewhere in the UK, even though it started off FAR higher in London when lockdown began and we’ve all been under the same lockdown regime.
The ONLY way to explain this is that HERD IMMUNITY WAS ACHIEVED IN LONDON SOME TIME AGO (a few weeks ahead of other places) and so the virus is naturally fading away there – and almost has disappeared. Elsewhere, the level of infection needed to rise some more during the early lockdown period before herd immunity was achieved. It may or may not quite have been achieved elsewhere, but probably has been by now given that infection prevalence is now falling everywhere.
Yes, great second to last paragraph. In Wuhan the peak was Feb 12 – two weeks before their lockdown. IMHO we need legislation to forbid lockdowns as in Japan.
Something I’d like to run past you. _____________________________________________________________ “The R0 for novel influenza A (H1N1) has recently been estimated to be between 1.4 and 1.6” https://pubmed.ncbi.nlm.nih.gov/19545404/ _____________________________________________________________ “The novel 2009 H1N1 influenza strain has been found to have… an R0 of 1.3…. Seroprevalences against pandemic 2009 H1N1 influenza varied by age group, with children age 10-19 years having the highest seroprevalence (45%), and persons age 70-79 years having the lowest (5%). The baseline seroprevalence among control samples from 18-24 year-olds was 6%. Overall seroprevalence against pandemic H1N1 across all age groups was approximately 21%.” https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2828126/ _____________________________________________________________ “one may predict the final size of epidemic, z, the proportion of those who will experience infection by the end of epidemic, by using the final size equation (based on a homogeneous mixing model), z = 1 – exp(-zR0) https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2821365/ (Neil Ferguson’s model estimated 81% for Covid-19 at R0 of 2.4 (I think). The formula gives 88%. For an R0 of 1.5, the formula says that 59% of the population should be infected and showing antibodies. For an R0 of 1.3 it is 42%). _____________________________________________________________ MY POINTS AND QUESTIONS So for H1N1, the above study found that the actual measured seroprevalence averaged at 21%,… Read more »
R0 cannot be employed so simplistically. The ‘R’ number of people on average that an infected person infects changes dramatically over time and in different situations. E.g., for the same virus this number could 0.5 in the general population but 10 in a prison population. Also, as more people become infected there are less people to infect, and so this number decreases from above one and goes below one. If people change their behavior, this too affects the number, as will things like demographics, average health status, degree of inherent immunity in a population. A typical virus will also mutate with time to become more infectious but less sever, again affecting the R number. R0 (note the 0) is the R number when the virus first starts spreading. For SARS-CoV-2 R0 was about 3 or 4. The current R number has now fallen to something way below 1 – because it has by now infected many/most (who are hence immune), because there is a significant degree of inherent immunity in populations (e.g., infected prisoners at Marion Correctional Facility have 98% zero symptoms), because we are now socially distancing etc, and therefore, overall, because herd immunity has been achieved (infection prevalence… Read more »
Re-enforcing the comment made above re downward bias of running serological tests in blood donors, this report (https://www.medrxiv.org/content/10.1101/2020.05.10.20097451v1 – download the pdf) shows that for the same region of France and using the same antibody test, the blood donor individuals demonstrated 10-fold lower sero-positivity that the general population.
In many situations in different countries, many tens of percent of the general population people are sero-positive. Then one must add those that fought the virus via T-cell rather than by creating many antibodies, plus there are those that have inherent immunity.
So herd immunity must now be working to largely eradicate this virus. Ignoring this blatant evidence whilst ruining lives and economies by maintaining draconian lockdowns (with the hundreds of thousands of excess deaths that will cause) is surely immoral unjustifiable if truly “being led by science”.
This is rather dated now. In fact, it was dated when it was published a month ago. Look, this is a trans-species virus. We would not expect it to be well adapted to a human host as the flu virus is, for example. We would expect that small changes in cell proteins due to genetic variations in humans would make a big difference in whether the virus can actually lock onto and infect those cells or not. With that in mind, we can see that the big mistake in the Imperial College model was to assume everybody was susceptible to the disease, whereas in fact that is not the case. Models need an additional silo of people who will never be infected simply because the virus cannot infect them. A number of studies on ‘petrie dish’ cases like the Diamond Princess and the USS Theodore Roosevelt (where there was substantial testing done and where respectively eight died and one died in a population of around 3500 who where exposed to the disease) and others show that about 80% of us will never get infected because we are simply not susceptible. In consequence, the famous “R zero” number defining the likelihood… Read more »
I disagree on a few details: (1) Original ‘R’, i.e., R0, was greater than 1 or the disease would not have spread as it did, and it was about 3-4 whereas ‘R’ has now fallen to <1; and (2) lab evidence indicates this virus infects human cells more efficiently than any other species, which is good evidence that it was selected for in a lab somewhere, whether or not it was manipulated by genetic engineering.
But the general thrust of your argument is certainly correct, that – for genetic diversity, antibody cross-reactivity, and other reasons – many people will not be infectable and/or will clear the virus almost immediately via T-lymphocytes without generating detectable antibodies. This is not theory, a wide range of evidence now exist to prove all this. Hence as you say – the disease will burn itself out, there will be no rebound, there will be no second spike, there is no need for a vaccine….. there never was a case for [such a draconian] lockdown.
Today (28 May), as I warned last week, the government announced that “Of those individuals providing blood samples, 6.78% tested positive for antibodies to COVID-19“
[I was actually told this number was going to be nearer 4%, but no matter]
What they DID NOT make clear, is that this finding relates ONLY to BLOOD DONORS. To quote ONS…
“Public Health England also publish an estimate of the prevalence of antibodies in the blood in England using blood samples from HEALTHY ADULT BLOOD DONORS.”
https://www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/conditionsanddiseases/bulletins/coronaviruscovid19infectionsurveypilot/28may2020
In France, they previously ran serology tests on healthy and on people going to the supermarket. The percentages were 3% and 30% respectively
So this 6.78% figure is a complete curve ball, intended to mislead. Extrapolating from the French data, the real fraction of UK population likely to have been infected by now is >>50%
I’m only aware of one study of archived U.S. blood samples (collected by the Red Cross over 4 days in mid-December 2020 and then six other states in mid-January 2021). This seroprevalence study found that abut 1.47 of the 7,000 or so units of blood tested “showed signs” of previous COVID infection. The blood collected in mid-December from just California, Oregon and Washington showed that 2 percent of the archived blood vials showed signs of likely previous infection. Well, extrapolating to the USA, this would mean that somewhere between 4.5 and 6 million Americans had probably already been exposed to the virus by mid-January. Of course the CDC study said that results “may not” be “generalizable” to the population. But “may not” also means “may be.” I don’t know why the results wouldn’t be generalizable to the population at large as blood donors would seem to be a good, representative sample. This study probably UNDERESTIMATES prevalence as blood donors tend to be much older … and those most likely to contract the virus would be younger. Anyway, this “Red Cross” blood study – and the inferences which can be drawn from same – didn’t get nearly enough attention. Also, it’s… Read more »
I have a contact in one of the closed down antibody testing labs who says they have several 1000 results sent in by people who bought kits online and 14% are positive. The population might be a bit biased but it seems surprisingly high.
…and even the CDC admits those tests are wrong half the time! So double that 14%
https://www.zerohedge.com/health/cdc-admits-covid-19-antibody-tests-are-wrong-half-time-virus-isnt-deadly
Professor Karl Friston, ranked by Science magazine as the most influential statistician in the world says: (a) social distancing rules are unnecessary; and (b) “the true portion of people who are not even susceptible to Covid-19 may be as high as 80%” https://unherd.com/2020/06/karl-friston-up-to-80-not-even-susceptible-to-covid-19/
Science magazine reported two studies that “found some people never infected with SARS-CoV-2 have these [T cells], most likely because they were previously infected with other coronaviruses.” https://www.sciencemag.org/news/2020/05/t-cells-found-covid-19-patients-bode-well-long-term-immunity
“Importantly, we detected SARS-CoV-2-reactive CD4+ T cells in ∼40%–60% of unexposed individuals, suggesting cross-reactive T cell recognition between circulating “common cold” coronaviruses and SARS-CoV-2.”
https://www.cell.com/cell/fulltext/S0092-8674(20)30610-3
Science Magazine (11 Nov 2016) ranked Prof. Friston as the most influential neuroscientist, not statistician. More to the point, his model postulated an unknown and unknowable form of protection. It may be true that 80% of people in the UK are not susceptible, but how would we know? That 80% figure certainly cannot have been true in New York City, where the death rate has reached 0.2% of population, and 0.6% in one zip code, unless the IFR were at least 3%.
We are living through Covid Hysteria. I lived through Hong Kong flu in 1968 – killed 80,000 in UK, no lockdown – and checked with contemporaries that no one remembers any ‘lockdown’ then. The big difference now with Covid is that politicians have got involved, hence the media too, and none have any grasp of evidence-based thinking, so act only to spread panic and make ruinous decisons.
Research has moved on, as has testing. We now know that about 60 to 75% of cases are broadly asymptomatic. These people can get infected and test positive for a while; but don’t get ill and such antibodies they have don’t seem to last that long, except possibly with repeated exposure. There is no good reason however to doubt that future exposure will have the same nil effect. Nor is there any sign that such asymptomatics infect others to any significant degree. So then? Antibody tests suggest 7 % actually getting ill enough to get them. 4 to 5 million then of which perhaps 10 % needed intense medical intervention and 30 to 45,000 died. How many more to herd immunity? If that is 80%, then the asymptomatics have to count to that total whether they have been infected or not. So we aren’t quite there yet but still close enough for this to be the aim of Government policy with a vaccine a suitable back up for the most vulnerable.
We have a massive government department (Office for national statistics) crawling all over data and telling us daily how many people are being tested positive, deaths, etc.
They apparently test 20, 000 people a week to see how many people really have it, not just those going for tests. However, can anyone please find an up to date figure for how many people have actually had the disease??
If half the population have had it, then how bad can it now get? There would then be no need for any of this control over the population?? For some (strange😂) reason don’t want to give us this figure….
Btw… only 7% of people going into hospital with this disease are now not surviving (how poorly do we think this 7% are already?)
This won’t stop until we say NO!
Here’s an epidemiological study that hasn’t been done that should be done: Simply survey all of the people who think they may have have had COVID before, say, March 1st 2020. What percentage of this group did NOT subsequently “test positive” for COVID? If this percentage is significantly higher than the population at large this would strongly suggest these people DID have COVID (before it was said to be spreading) and had developed some kind of “natural immunity.” BTW, as others have pointed out, the population that thinks they may have had COVID undoubtedly numbers in the MILLIONS. Comment sections of articles are full of people who report they were sick with “signature” COVID symptoms and, went to a doctor, and, significantly, tested “negative” for influenza. I’ve done my own amateur study in my hometown (via personal inquires and Facebook solicitations) and quickly found about 50 to 75 people who think they may have had COVID before the virus was supposed to even be in America. I’ve actually followed up with about 40 of these people. Only one or two later came down with (confirmed) cases of COVID. I’m not saying everyone who thinks they had COVID did have COVID… Read more »