Main Edition
US
FR
May 5, 2020 - 11:32am
The deadliness of Covid-19, measured by the “Infected Fatality Rate” or what percentage of infected people end up dying, has become an issue of global significance.
At UnHerd, we’ve spoken to experts at both ends of the range of estimates, from Neil Ferguson (who believes the IFR to be just under 1%, perhaps 0.8-0.9%) to Johan Giesecke who maintains that it is nearer 0.1%, or one in a thousand.
This may sound like splitting hairs — they are both under one percent after all — but in reality the difference between these estimates changes everything. At the lower end, a much more laissez-faire policy becomes possible, and at 30,000 deaths it starts to look like the UK has already been through the worst of it; at the higher end, a policy of continued ultra-caution is necessary because a more relaxed approach could mean hundreds of thousands of additional deaths.
That’s why the study conducted by Professor Hendrik Streeck of the University of Bonn is so significant: a representative sample population within Germany was tested and examined in great detail to determine what percentage had already been infected with Covid-19.
The headline result is that 15% of that population was infected, which implies an Infection Fatality Rate of 0.36%. This would put him somewhat in the middle of the previous experts we have spoken to. Professor Streeck was keen to point out, however, that he still believes this is a conservative estimate, and thinks it may be closer to 0.24-0.26% and may come down further still as we know more. He published the higher number to err on the side of caution: “it is more important to have the most conservative estimate and see the virus as more dangerous than it is,” he said.
To show just how significant every percentage point difference makes, if the 0.36% is correct for the UK, and we have had 30,000 Covid-19 deaths, that would mean around 8.3 million people have been infected, or 12.5% of the population — not enough to start feeling confident of much immunity in the community. If the lower estimate is correct, 0.24%, and there has actually been closer to 50,000 Covid-19 deaths (as per the FT’s speculations) then that figure suddenly rises to over 20 million, which at around a third of the population would fundamentally change the calculus of how bold we can be coming out of lockdown.
Join the discussion
Join like minded readers that support our journalism by becoming a paid subscriber
To join the discussion in the comments, become a paid subscriber.
Join like minded readers that support our journalism, read unlimited articles and enjoy other subscriber-only benefits.
SubscribeA big thanks to Freddiie for, once again, giving voice to one of the few sane actors in this whole farce. It seemed likely to many us from the beginning that the lockdown was a massive overreaction and that all we needed to do was:
– protect old people
– ban mass participation events
– get out there in the fresh air and sunshine
Cannot agree enough. But people like us are the voices in the wilderness and the Doomsday cult of Ferguson’s is at the wheel.
He should have thrown himself on the pyres of burning cattle that he ordered back in 2001. He’s done more damage to the UK than Napoleon, Hitler, Blair and Brown combined.
That sort of vicious insult of Prof Ferguson does nothing for your argument and should be condemned, just like those who wished for Boris Johnson’s demise when he had the virus.
I’m afraid Ferguson deserves it. It’s simply staggering that someone with his appalling track record could have been invited onto such a crucial committee. I noticed that Freddie didn’t raise this- I can only assume that Ferguson agreed to appear on condition that he saw the questions in advance. He seems quite media savvy.
His original coding for his model has never been released for others to scrutinise- only the cleaned up version- and no one can make that work.
He clearly has something to hide.
In interviews he adopts a faux diffident attitude all the way down to the studied anti – fashion singlet and jacket . But then he puts pics of himself looking stern and disapproving – with arms folded across his chest . They’re really saying this is hurting me more than you.
Can’t you see it? This guy is just an attention seeker- in the end it’s all about him.
Much tho’ I’d love him to be right, I think his logic is totally flawed:
The original study was of a very unrepresentative population, a fairly closed community where there was a “superspreader” event (carnival). His post-event and post-lockdown study indicated that c15% of that population had been infected, but there were only 7 deaths out of c12.5K in the community, giving his headline 0.37% IFR (=7/(15%*12.5K)). He then applies the same IFR to the UK to back-calculate the number of people who have been infected in the UK from the number of deaths.
The fundamental flaw is in the non-representative nature of the population that he studied. Few elderly or otherwise unwell (co-morbid) people would have been dancing, singing, kissing at the carnival. Some would have been infected later by transmission from carnival go-ers, but the infected population would be heavily biased towards the young and healthy, and as we well know, the risk of death from Covid-19 in young, healthy people is very low, hence he calculates an artificially low IFR. Indeed, given the heavy bias of his sample, I think his study gives a minimum for the IFR, not an estimate of it.
While it may well turn out that the UK lock-down is an over-reaction with costs (£ and health) that end up to outweighing the benefits, the alternative would have certainly been an overwhelmed NHS and very many more deaths in the short term – both politically totally unacceptable. The government had no choice.
You’ve hit the nail on the head here. A lot of the IFRs and CFRs being bandied about by medical scientists are not mathematically sound at all… Every single one I’ve heard trying to guess an IFR are based on a: the underlying numbers being statistically significant (you can’t calculate a representative IFR from 7 deaths!) and b: that the virus has spread evenly across the population. We know that R in a care home or hospital setting is multiple times higher than general population and we know that the viral load in these settings is substantially higher (which probably means more severe infections) – yet none seem to even attempt to account for this. It would be a more useful number if they were to try to work out IFR among the general population – i.e. exclude people who were infected in care homes or hospitals (excluding medical staff too). You can then base policy decision on a general population IFR while spending the billions on reducing cases in care homes and hospitals – regular tests of everyone entering, large grants to care home staff if they ‘live in’ for a few months. The IFRs would no doubt look much lower then and they could justifiably remove general lock downs more quickly while keeping infections out of places where there is a high density of vulnerable people. The positive take away from this is that countries will find that they can exit lockdown more quickly than they think if they can successfully protect the vulnerable – unfortunately it will be a lot slower than if they had done this from the beginning.
I agree that the extrapolations were very poor mathematics. However there are plenty of other better indicators that the IFR and CFR are much lower. I did one based on the discovery of the Santa Clara death from Feb 6th – which meant the infection happened about mid Jan (2 months before any lock-down) people were free to continue their normal lifestyles with no social distancing in a highly populated area (albeit with most people living in suburbs) for >7 weeks.
As of April 23rd California has 37,788 confirmed cases and 1,440 deaths attributed to Covid-19
which is a rate of 3.8%…. but widespread testing has not occurred so the cases could be much higher and the rate significantly lower.
Adjusted for cases and deaths per 100,000 people, California has very few – just 97 cases and 4 deaths – despite 3 months with the virus in the community and most of that time without social distancing.
We can hypothesise that many more people were infected but not yet counted. If we then take the 100,000 example and postulate that they might have a similar number showing antibodies as New York State then there would actually have been 13,900 cases with 4 deaths for a shockingly low death rate of 0.029. How could this be? A likely answer is that many deaths in February and March were not attributed to Covid-19 – but even if say 40 cases per 100,000 were missed (counted as flu or other illnesses) that would still only bring the Covid-19 death rate to just above the flu 0.29
I agree. It is an important study, but must be placed among many.
I sometimes think commentators should google “replication crisis”. Science cannot provide quick answers, and looking for them , while ignoring the precautionary principle, will get people killed. And tank the economy too.
In retrospect – most governments have overreacted. But given the (now shown to be wildly inaccurate) predictions of as much as 5-7% CFR the clamp-downs were justified. What’s difficult for them to admit is that they should now change policy – the delay is partly caution but also likely to be about confusion and saving face.
Again this confirms what Johan Giesecke stated is the true facts. Johan and Hendrik both confirm the virus requires limited contact and mitigation strategy. Actual infection:Death is < 0.5%. Similar to a bad flu.
Henrick stated that the complete lockdown instead of a graduated lockdown has robbed the scientific opportunity to evaluate which mitigations were needed.
I cannot say this enough the “Doomsday” cult of repression and mass annihilation of economy will effectively kill more people than this virus. The correlation between suicide and financial depression is 1:5. That is 20% not 0.36%. Before this 80% of people were living paycheck to paycheck with less than £400 in savings. 40% of small businesses will not reopen. Number of deaths due to refusal of access to cancer treatment will kill more people in the next two years than all corona deaths combined.
No one is saying not to use enhanced mitigation such as small groups, social distancing, hand washing etc. Unfortunately now the best return according to Hendrik is to apply each stage of reopening and wait 2-3 weeks to measure the effect.
Meanwhile, the real virus of death is the inability to respond to every other death related cause in our communities. The true virus was fear and panic and it pervades crippling a prudent scientific approach.
That IFR would be far in excess of seasonal flu. You are thinking of the CFR …the fatality rate associated with diagnosed cases,( plus a bit of guesswork from excess deaths each winter ) which is .1%. The majority of flu cases are unreported.
So the IFR for flu is much lower than .1%
Also, this IFR would not account for New York City. So I’m a little sceptical…but hopeful. It is good news.
… all good points well made and too true. The fear machine is still gearing up.
If the Liverpool School of Tropical Medicine preprint below is correct, it has huge implications. It points out that that different populations have different levels of exposure and susceptibility to the virus. So a care home, a hospital or a crowded tube station offers the perfect environment for large numbers to be infected very quickly. They then use complicated mathematics (that I can’t follow) to calculate what that could mean for the level of herd immunity needed to stop the virus spreading further. They use a measure called Coefficient of Variation (CV).
“Most CV estimates are comprised between 2 and 4, a range where naturally acquired immunity to SARS-CoV-2 may
place populations over the herd immunity threshold once as few as 10-20% of its individuals are immune.”
Recent research from Italy and Iran (and evidence from Iceland and the US) where serological tests have been undertaken show 20-30% of densely populated regions have already been infected. London almost certainly has at least 20% – just on the data from the Kings College Symptom tracker. So this would mean, at least in London, herd immunity has been reached. At the very least the likelihood for any second spike reaching similar levels to April 8th is vanishingly small.
https://www.medrxiv.org/con…
https://www.nytimes.com/202…
It talks about heterogeneity of susceptibility to infection (higher variation (CV) means lower herd immunity threshold).
I haven’t seen anything to suggest some people are more susceptible to infection than others – perhaps children – with Covid (although it is true for some other viruses – which is why you see they disappear when only a small percent are infected – eg Zika)
But….. if you re-used this model and changed it to heterogeneity to susceptibility to death… You would get some useful and measurable numbers… We already know CV of susceptibility to death… You still have high herd immunity threshold but if you can protect the vulnerable for a relatively short time you achieve herd immunity very quickly (with a much lower overall death rate) just by some simple measures to protect the most vulnerable while the virus burns its way through everyone else…. As whoever wrote this paper says though – if they persist with homogeneous lockdown they will unnecessarily prolong lockdown. The longer lockdown lasts the less likely that the vulnerable can completely isolate themselves.
Excellent interview and more good impartial information to consider. I hope our ‘experts’ are considering all opinions and research and not getting blinded by groupthink. Freddie, you really must teach your colleagues in the big legacy media companies the basics of interview technique! We would all be better informed if this approach to interviewing was standard.
The lockdown would never have been justified on a 0.34% fatality rate – never. Given that is probably a conservative estimate and other antibody surveys have come in lower it needs to be ended now.
That’s simply untrue. 1) People keep comparing flus CFR to to covid’s IFR. See my comments above. 2) Some ambulance services were stretched beyond capacity even with lockdown.
Thank you for the whole Lockdown series of incredibly enlightening interviews, where experts are asked open questions and given time to answer properly. As a GP working in the UK, I have seen clearly one aspect of the lock-down which has been a dramatic reduction peoples willingness to seek help for serious ‘other’ medical problems, and the attention of medical services diverted away to Covid 19. Dr Sikora the cancer specialist has echoed this concern (https://www.medscape.com/vi… Beyond this the economic consequences for the UK and world economy will be way more severe. Finding the answer to critical questions like the Case Fatality Rate (or Infection Fatality rate) seems absolutely essential to making the right public policy and I am surprised that there has not been more made of this. I am hearted that it would seem that the CFR and IFR are much lower than was first estimated by Prof Neil Ferguson. I hope more studies like this will be funded and the results and their import discussed openly.
Well, well, how the malignant have fallen. The vile Ferguson embodies all the serial incompetence and hypocrisy of the so-called experts and elites. Let us hope that we never hear from this noxious toe rag again
In more robust times, such as Ancient Rome or modern Japan he would have “done the decent thing” and rid the planet of his festering presence.
No such luck now.
Disgusting
Really? Why so?
Thanks for these excellent interviews. I am not an epidemiologist but I am a scientist quite familiar with experimental uncertainty and the perils of modelling sparse data. Also I live in Sweden so am following these things quite closely.
Could the divide between the Giesecke school of thought (low fatality rate) and the Ferguson school (higher rate) be explained by how much attention they have been paying to contact tracing such as presented here, and earlier studies from China?
When only limited data is available it is totally possible to fit a very good model to a wide range of outcomes including very high fatality predictions which would lead to the more pessimistic approach from Ferguson et al. Those who have looked closely at detailed studies of the outbreak and seen for example very little transmission via schools or outdoors despite relatively widespread infection have some important information which may not be captured in the large scale modelling by Ferguson et al.?
I’d be really interested if you’d put that question to one of your guests Freddie on this very informative series of interviews.
This is the first epidemiologist I have heard who is thinking like me. And if 1 of 3 in his study cohort(s) were implied pseudo-false negative antibody-test results, it will be much more than that IMO for the general population for reasons he discusses (and I have discussed for a long time).
That is because in this cohort a social event caused much higher initial exposure of virus than would be expected for the broad population, and this would lead to larger viral load and symptoms and more humoral antigenic stimulation.
If antibody testing has a large percentage of pseudo-false negative results (not due to limited sensitivity, but due to actual lack of antibodies in those exposed and infected) as I have postulated, that is very good news.
It would mean much more progression of herd resistance than the early serological testing studies would otherwise imply, as well as even lower fatality percentages.
If so, this should become more evident shortly but will be also confounded with arrival of summer and natural improvement of vit D status in northern hemisphere.
I agree that there is increasing evidence that available antibody testing will not yield the true prevalence of COVID-19 due to the antibody titers becoming undetectable very soon after recovery from the infection. If this is correct, it would suggest that the lethality rate of this virus is even lower. I am commenting as a physician in a semi-rural U.S. clinical practice that actually tested our staff who were exposed to confirmed COVID-19 patients. We saw thousands of patients in our acute care setting with COVID-like symptoms over the course of the last 4 months – many prior to appropriate PPE use. So far, all 37 staff members who were tested are negative for COVID IgG antibodies months after exposure or after they had an illness consistent with SARS-CoV-2. I believe the professor is correct and COVID-19 will turn out to be like many other coronaviruses and rhinoviruses with a very short-lived IgG immunity. I hope that he is correct that T-cell or other natural immunity will provide protection for future COVID exposures. I think that he is likely correct about this as well.
The decision to keep kids out of school is not supported by the data. There are still no deaths under 30 in South Korea and they tested early and often since January.
Thank you for producing this very enlightening interview. It’s refreshing to hear two highly intelligent people canvassing the science in a methodical non emotional manner. I have no comment or opinion to offer other than a sense of despair that our politicians probably will never take the time out to listen to this excellent interview.
In your interview with Neil Ferguson, the latter went to great pains to emphasize that there is no single/universal infection fatality rate (IFR) The IFR depends greatly on the demographics and age structure of the population.You are therefore misrepresenting what Neil Ferguson said. It’s difficult to make a judgement about this IFR from the German study without knowing the age structure of the sample considered. This idea of a single universal IFR is a massive red hearing. It’s also impossible to make extrapolations to the UK population without making adjustments for the age structure of the German study. However, the German study is too small to make reliable estimates of age-stratified IFRs.
I tried to make clear that Prof Ferguson emphasised how different the IFR would be for different populations, but he did estimate 0.8-0.9% for the UK as a whole, and Prof Streeck’s estimates are much lower, so wanted to probe that…
The problem is that all the models (including Imperial’s) assume that a population has a homogenous IFR… ultimately that’s how he gets to such a high number. (i.e. IFR in a care home is 50+ times higher than in a university but the imperial model doesn’t account for it). It also assumes that R is homogenous when we know it is multiple times higher in care homes and hospitals . If the models looked more at protecting vulnerable settings, and dramatically lowering infection rates in those, the population IFR would be dramatically lower – and population wide lockdowns would have been shown to have a much lower effect on the end result.
Bingo!
I find it hard to believe Ferguson neglected something so simple. Did you do the math? I took the IFR vs age table (table 1) from Ferguson’s report and applied it to the US population distribution and get around 3 million deaths. Then I apply an assumed herd immunity percentage of 66% to get 2 million deaths. Here’s the google sheets:
https://docs.google.com/spr…
Peter,
Thanks for this and your helicopter money article last year linked in the article.
I think (hope) you underestimate Boris and the relatively young team he has – they may be a bit naive politically, but they are not stupid. Boris said before he went down with Covid that it won’t be like 2008 and he said when he came back he did not recognise austerity (we never really did austerity anyway). Look out for terms like re-balancing and balancing off. Boris has a plan and a team on side to deliver that plan that would have seasoned Tories choking on their caviar canapes and Maggie turning in her grave.
Helicopter money is happening already. It has been untargeted to a large extent due to the need to get it out quickly, but the attempts are there, eg the up to £5m loans are for “good” businesses and directors needing to put up personal guarantees for over £250k will make them think how good am I really.
Going forward the targets need to be those who really need it and those who will do the most sensible things with it ie the exact reverse of 2008 QE.
I think the reason why we have not seen the expected effects of QE are because banks were not just lending money to people who had no hope of paying it back, they were not just lending money they did not have to lend in the first place, they were lending money which did not officially exist because it had not been “printed” yet. How much unprinted money is still out there and needs to be converted to printed money is anyone’s guess.
While I think it is good to hear from outliers, could we balance this a little? We’re beginning to sound a little too like Peter Hitchens at the moment. Could we have an interview with the likes iOf Carl T Bergstrom, perhaps?
Also…that’s an IFR far in excess of seasonal flu. (A CFR of.1% is often incorrectly compared with the IFR for covid).
Very basic data seems to be absent from the study.
1. Demographics of the study: (number of people in age ranges, for male/female).
2. Demographics of infection: (number of people in age ranges, for male/female).
3. Demographics of fatalities, observed in the area.
Also, whatever we believe this tells us, still has to answer the South Korea question…
What I’m still curious to know about is whether the countries being compared by IFR have a difference in how they count deaths. There’s been plenty of controversy around policies that include deaths with a present Covid infection as equal to deaths from a Covid infection. Even medical professionals have gone on record to demonstrate their disapproval of that approach. How can we compare IFR stats unless the policies are the same or similar?
And I want to thank Freddie for being one of the handful of clear, non-partisan and sane voices amidst all of this chaos.
The guest almost suggested inoculating with the virus (probably politically incorrect to suggest, but hey, I don’t have his career to protect) to achieve herd immunity. I think that would be a viable strategy. To start you inoculate school children and isolate them to prevent the virus spreading uncontrolled in society. After they recover they should be reasonably safe to “roam” without precautions. This group can be done relatively en-masse because their IFR is so low. If you apply that strategy to young adults you have go more slowly because IFR increases exponentially with age which will in a larger hospitalization rate. Other enhancements include screening for risk and “treatments” to mitigate severity without preventing sero-conversion.
In a sense the COVID-19 pandemic does provide an opportunity to set things straight. As Peter Franklin is aware of the BoE blog “Helicopter money: setting the tale straight”, he must see that helicopter money is not an acceptable alternative to the policies followed to get out of the last recession. The best hope for the future lies in a reformed inflation targeting regime for the Bank of England. The UK more than any other developed country has been subject to housing booms and busts. The last recession came from one of them and the next one could as well. It is time for the UK to reform its HICP (a gross premiums approach to insurance services, as in the RPI, makes better sense than the existing net premiums approach) and add an owner-occupied housing component based on the net acquisitions approach. This won’t help much in the recovery, but it will mean that as recovery turns to expansion the UK will be less likely to be hit by another housing-induced recession.
What I don’t get is this:The Bonn university study measured how many people were infected, but in order to compute the infection fatality ratio you also need the reliable number of deaths. How did the researchers deal with this issue? For if in the region of their study the true number of deaths is double the official number (as is the case in many parts of Europe) then the IFR is doubled also.
The number of deaths is being counted with two variables which work against one another – one increases the count by including deaths where people died ‘with’ COVID-19 (if you died in a car accident you’d still count), the other decreases the amount it’s said because people are dying at home and being buried without testing and therefore not being counted. Much bigger and consistent samples are needed, but extrapolating from areas like California the rate still looks like a ‘bad flu’ year – see detail in my reply to will and doug mccarthur above
Why are we not looking at the international disparities?
49 countries report less than 1 death/1M population.
64 report between 1 and 10 deaths/1M.
32 between 10-30.
16 between 30-50.
7 between 50-100.
5 between 100-200. (Monaco, Portugal, Bermuda, Canada, Luxembourg)
3 between 200-300.(Switzerland, USA, Ireland)
4 between 300-500 (UK, France, Sweden [319], Netherlands [316]).
4 between 500-740 (Italy, Spain, Andorra, Belgium)
World average: 36 deaths/1M population
World Median: 3.5 deaths/1M population
Remember, we are talking about deaths per million.
Details are here: https://www.worldometers.info/coronavirus/
Those countries with more than 100 deaths/1M are all, except for Bermuda, either Western European or North American.
This is a ‘crisis’ only for a portion of the First World. However, by bringing those economies to a sudden halt and keeping them locked-down, we are very likely to create an economic crisis in the rest of the world which will lead to far, far more damage than that caused to them by C-19
I just saw the interview with the German professor and found it very interersting. The only thin thats disappointed me was you were only talking about Germany and UK, even if you also have talked to Swedish doctor Giesecke.
Being a swede as i am is to live in a country where the “ the controversial corona strategy” has a a support from 80% of the population. When i open medias like BBC or Guardian there is only accusations and talk about “ the idiots in Sweden”.
But as doctor Giesecke said in his interview: “ You can make a total lockdown but not very long periods in a democracy” And dr Giesecke also said that he thinks that in the end all countries will end in same percent of fatally rate.
I am only saying that i am disappointed that this interview didnt say a word of Swedens strategy. That had widened the interview a bit more for us who want a wide coverage and knowledge.
Is it possible to get a reference for the study or studies Prof. Streeck referred to showing that the larger viral load a person was exposed to, the greater the infection rate? This is of course common sense but I was not aware there was a study substantiating it. Also, I would like the reference to only 15% infection in a household regardless of household size.
You describe Prof. Streeck’s study as “a representative sample population within Germany was tested and examined in great detail” but this is not what his preprint at doi 10.1101/2020.05.04.20090076 says: its title is “… a German community with a super-spreading event”. Its conclusions start “While the number of infections in this high prevalence community is not representative for other parts of the world”. The community was chosen to estimate the IFR, not to estimate the prevalence of the infection in the population at large.