The case for C19 being widespread
Epistemic status: very, very uncertain. Please continue following official advice like social distancing etc.
I’ve seen all the previous discussions (including in this thread and on Tyler Cowen’s blog, but remain unconvinced]
--
A preprint from 24⁄03 by Gupta et al. at Oxford[1] suggests that the current data on C19 is consistent with both:
Few infections, low infectiousness (r0), and high infection fatality rate (IFR)
Widespread actual (asymptomatic) infections, high r0 and low IFR
For instance, the authors suggest that if the IFR is low and C19 is very infectious, it is possible that by 19⁄03, 36%-68% of the UK population would have already been infected with C19. The ongoing epidemics in the UK and Italy started at least a month before the first reported death.
This study has been criticized, but scientists agree that serological assays will show whether this hypothesis is true.[2], [3]
Here I make the strongest possible case that C19 is widespread, r0 is underestimated and IFR is low.
I base this on the following:
Much C19 transmission might be asymptomatic[4] and presymptomatic.[5],[6]
Tim Spector, professor of genetic epidemiology a King’s College London, finds that
10% of 650,000 UK users of their C19 symptom tracker app showed mild symptoms. Thus 6.5m people in UK are infected, not taking into account asymptomatic cases
A preprint from 26⁄03 by epidemiologists Gutierrez et al.[7], Professor and Chair of Mathematics at University of Texas at San Antonio (Google Scholar Profile) suggests
An R0 between 5.5 and 25.4[8], if you account for asymptomatic spread. In this scenario, the peak of symptomatic infections is reached in 36 days with approximately 9.5% of the entire population showing symptoms.
The authors argue that it’s unlikely for a pathogen to blanket the planet in three months with an R0= ~3 and that it has to be more contagious than measles, which has an R0 of 18.
A preprint from 13⁄03 by Chowell et al., Professor & Chair—Georgia State University School of Public Health suggesting (GScholar profile first author who has written 2 paperson the Diamond Princess, senior author quoted in the NYT) suggests that
r0=5.20 (95%CrI: 5.04-5.47)
IFR=0.12% (95%CrI: 0.08-0.17%), several orders of magnitude smaller than the crude CFR estimated at 4.19%.[9]
~20% of the all people in Wuhan were infected on Jan 23rd (~2 million infections)
A preprint from 24⁄03 by French epidemiologists[10] (Google Scholar profile) suggesting:
“The actual infections France is probably much higher than the observations: we find here a factor ×15 (95%-CI: 4 − 33), which leads to a 5.2/1000 mortality rate (95%-CI: 1.5/1000 − 11.7/1000) at the end of the observation period. We find a R0 of 4.8, a high value which may be linked to the long viral shedding period of 20 days r0=4.8”
Oxford University Evidence Service meta-analysis suggests that as of 22⁄03 that the IFR=~0.29% (95% CI, 0.25 to 0.33).[11] Widespread testing (which isn’t random) in Iceland suggests an even lower IFR.
A British Medical Journal editorial from 20⁄03 arguing that C19 fatality is likely overestimated [12]
The Imperial study is based on “thousands of lines of undocumented C [code] from 13+ years ago to model flu pandemics”[13]
Dengue tests react to C19 and many could be false positives according to a Lancet paper[14]
Through the week ending March 13, Paraguay has reported 203,922 total dengue fever cases, including 51 deaths. This compares to 669 dengue cases reported during the same period in 2019.”[15]
Singapore which is said to have very good containment of C19, reports a recent dengue outbreak (4000 cases) doubled from previous year[16]
There were a few dengue in Australia and Florida where it is unusual[17]
High proportion of special populations are infected (celebrities, athletes and politicians).[18] For instance, very many Iranian politicians have C19.[19] This suggests that if the whole population had access to frequent tests like those special groups would have, then we would see many more cases. Fatalities are also very high amongst people with very high age and many comorbidities, suggesting that there are many asymptomatic infections amongst the young.
C19 has been detected in wastewater in the Netherlands. If the test is not very sensitive, this would suggest C19 is widespread.
“From 17 February 2020 on, weekly wastewater samples were taken at Schiphol Airport. During the first two weeks, the virus that causes COVID-19 was not detected. However, the genetic material from the virus was detected in the airport wastewater samples taken on 2, 9 and 16 March. The first sample containing the virus was taken four days after the first person in the Netherlands tested positive for COVID-19 on 27 February. Genetic material from the virus was detected in wastewater samples taken from the wastewater treatment plant in Tilburg on 3, 10 and 17 March.”[20]
Wastewater-based epidemiology biomarkers: Past, present and future
PCR test have a high false negative rate
They can only detect the virus for ~1 week
Difficulties in False Negative Diagnosis of Coronavirus Disease 2019: A Case Report. Note that this was a highly symptomatic person.
One person had persistent negative swab, but tested positive through fecal samples.[21]
“If the samples are not correctly stored and handled, the test may not work. There has also been some discussion about whether doctors testing the back of the throat are looking in the wrong place. This is a deep lung infection rather one in the nose and throat.”[22]
71% accurate the first time people are tested. The other 29%, the test showed negative even though they really had it.[23]
Infections in China might be underestimated because:
Hidden infections challenge China’s claim coronavirus is under control
“Chinese journalists have uncovered other cases of people testing negative six times before a seventh test confirmed they had the disease.”
“There have been reports that early Chinese tests may have had a false negative rate as high as 50%.”
Analysts have doubted the near-zero transmission rate in China for various reasons such as people with no symptoms being denied testing.[24]
Early Chinese tests may have had a false negative rate as high as 50%.[25]
Czech Researchers claim that Chinese do not work well[26]
[1] Lourenco J, Paton R, Ghafari M, et al. Fundamental principles of epidemic spread highlight the immediate need for large-scale serological surveys to assess the stage of the SARS-COV-2 epidemic. https://www.dropbox.com/s/oxmu2rwsnhi9j9c/Draft-COVID-19-Model%20%2813%29.pdf
[2] “Covid-19: experts question analysis suggesting half … - The BMJ.” https://www.bmj.com/content/368/bmj.m1216.full.pdf. Accessed 25 Mar. 2020.
[3] “expert reaction to unpublished paper modelling what ….” 25 Mar. 2020, https://www.sciencemediacentre.org/expert-reaction-to-unpublished-paper-modelling-what-percentage-of-the-uk-population-may-have-been-exposed-to-covid-19/. Accessed 25 Mar. 2020.
[4] “Presumed Asymptomatic Carrier Transmission of COVID-19 ….” 21 Feb. 2020, https://jamanetwork.com/journals/jama/fullarticle/2762028. Accessed 18 Mar. 2020.
[5] “Potential Presymptomatic Transmission of SARS-CoV … - NCBI.” https://www.ncbi.nlm.nih.gov/pubmed/32091386. Accessed 18 Mar. 2020.
[6] “Transmission interval estimates suggest pre-symptomatic ….” 6 Mar. 2020, https://www.medrxiv.org/content/10.1101/2020.03.03.20029983v1. Accessed 18 Mar. 2020.
[7] “Investigating the Impact of Asymptomatic Carriers on COVID ….” 20 Mar. 2020, https://www.medrxiv.org/content/10.1101/2020.03.18.20037994v1. Accessed 25 Mar. 2020.
[8] Investigating the Impact of Asymptomatic Carriers on COVID-19 Transmission
[9] “Early epidemiological assessment of the transmission ….” 13 Mar. 2020, https://www.medrxiv.org/content/10.1101/2020.02.12.20022434v2. Accessed 25 Mar. 2020.
[10] “Mechanistic-statistical SIR modelling for early estimation of the ….” 24 Mar. 2020, https://www.medrxiv.org/content/10.1101/2020.03.22.20040915v1. Accessed 27 Mar. 2020.
[11] “Global Covid-19 Case Fatality Rates—CEBM.” 17 Mar. 2020, https://www.cebm.net/archives/covid-19/global-covid-19-case-fatality-rates. Accessed 27 Mar. 2020.
[12] “Covid-19 fatality is likely overestimated | The BMJ.” 20 Mar. 2020, https://www.bmj.com/content/368/bmj.m1113. Accessed 27 Mar. 2020.
[13] “neil_ferguson on Twitter: “I’m conscious that lots of people ….” 22 Mar. 2020, https://twitter.com/neil_ferguson/status/1241835454707699713. Accessed 27 Mar. 2020.
[14] “Covert COVID-19 and false-positive dengue … - The Lancet.” https://www.thelancet.com/journals/laninf/article/PIIS1473-3099(20)30158-4/fulltext. Accessed 27 Mar. 2020.
[15] “Dengue outbreak in Paraguay: More than 200K cases ….” 24 Mar. 2020, http://outbreaknewstoday.com/dengue-outbreak-in-paraguay-more-than-200k-cases-reported-to-date-57202/. Accessed 27 Mar. 2020.
[16] “Dengue infections hit 4000, doubling from same period last year.” 22 Mar. 2020, https://www.straitstimes.com/singapore/dengue-infections-hit-4000-doubling-from-same-period-last-year. Accessed 27 Mar. 2020.
[17] “Dengue cases in Townsville could potentially be ruled … - ABC.” 27 Mar. 2020, https://www.abc.net.au/news/2020-03-27/dengue-cases-in-townsville-could-potentially-be-ruled-outbreak/12096104. Accessed 27 Mar. 2020.
[18] “Celebrities, Athletes and Politicians With Coronavirus ….” https://www.nytimes.com/article/coronavirus-celebrities-actors-politicians.html. Accessed 27 Mar. 2020.
[19] “Iran’s Coronavirus Problem Is a Lot Worse Than It Seems ….” 9 Mar. 2020, https://www.theatlantic.com/ideas/archive/2020/03/irans-coronavirus-problem-lot-worse-it-seems/607663/. Accessed 27 Mar. 2020.
[20] “Novel coronavirus found in wastewater | RIVM.” 24 Mar. 2020, https://www.rivm.nl/node/153991. Accessed 26 Mar. 2020.
[21] “COVID-19 Disease With Positive Fecal and Negative ….” https://journals.lww.com/ajg/Citation/publishahead/COVID_19_Disease_With_Positive_Fecal_and_Negative.99371.aspx. Accessed 26 Mar. 2020.
[22] “Are coronavirus tests flawed? - BBC News.” 13 Feb. 2020, https://www.bbc.com/news/health-51491763. Accessed 26 Mar. 2020.
[23] Modes of contact and risk of transmission in COVID-19 among close contacts
[24] “Life after lockdown: has China really beaten coronavirus ….” 23 Mar. 2020, https://www.theguardian.com/world/2020/mar/23/life-after-lockdown-has-china-really-beaten-coronavirus. Accessed 26 Mar. 2020.
[25] “How China’s Coronavirus Incompetence Endangered the World.” 15 Feb. 2020, https://foreignpolicy.com/2020/02/15/coronavirus-xi-jinping-chinas-incompetence-endangered-the-world/. Accessed 27 Mar. 2020.
[26] “Eighty percent of coronavirus tests ‘donated’ by China to ….” 25 Mar. 2020, https://www.washingtonexaminer.com/washington-secrets/80-of-virus-tests-donated-by-china-to-czechs-are-faulty. Accessed 27 Mar. 2020.
I’m sorry but this simply isn’t possible. The Diamond Princess cruise ship data alone proves it.
The Diamond Princess had a total of 3,711 people on board. By the end of its odyssey, a total of 712 of them tested positive, about a fifth. This population was observed in great detail, and puts severe limits on the fraction of people who have various outcomes. 10 people on this ship have died, and ~45 more were in critical condition at some point. If we take 712 as the true number of infections, we get a 1.4% death rate and 7.7% winding up critical or dead. The population was skewed elderly though which worsens the numbers somewhat, but that can’t change much more than a factor of two or three at the outside I think compared to a regular population. Thus an expected regular population’s death rate would be 0.5% to 1%.
20% of these people never had symptoms. Other data from other sources in China and Korea suggests the true number of asymptomatics could be above 30%, and Iceland is suggesting 50% but a lot of those people have not had followups so it could decrease. It is possible that more people on the cruise ship were infected but never tested positive, but it’s probably not huge compared to the positive tests.
The most extreme we could go is assuming that 100% of the ship was infected and four fifths of them were missed. This did not happen. But it’s an absolute upper bound. Then we get a 1.4% critical condition number and about a 0.3% death rate. Hospitalization would be higher. The parameters needed to have more than one or two percent of the British population have been exposed to the virus by now given reported deaths and hospitalizations require parameters for the fraction of people vulnerable to severe disease to be much lower than any plausible numbers here. Even with this extremely optimistic interpretation, you would’ve expected more than a full heavy flu season’s deaths squeezed into the last 2 months.
I am compelled to point to a twitter thread from Professor Bergstrom ( https://twitter.com/CT_Bergstrom/status/1242611599405277184 )
There IS a morbid case for it being relatively widespread in the North of Italy. The total excess mortality noted in several northern Italian cities is ~3x the official COVID death figures, indicating that a lot of people are dying without test results and without being noted down. That indicates ~30,000 total deaths in Italy. If we assume a 1% death rate for a population less skewed than a cruise ship, and a delay from infection to death this would indicate significantly upwards of three million Italians infected in the Northern provinces, which could be well over 10% of the badly affected provinces.
Median age on DM is 62 (basically, 50 per cent above 62). Most of the deaths were in people above 70.
In 2017, about 16 percent of the American population was 65 years old or over.
So, any DM mortality should be divided on 3 to get US mortality, given medical care will be available. If it will not be available, almost all critical care patients will die.
45 passengers of DM who were critical is 1.2 per cent of all 3700 DM passengers (I take all passengers’ number, not only infected ones, as it gives some estimation of the attack rate, that is the proportion of infected to those who escape infection, either via natural immunity or self-isolation. In US there will be many people, who will escape infection, via strong immunity, isolation, luck or very short asymptomatic illness—so short that it can’t be find via PCR).
Divided on 3, it gives 0.4 per cent of US population will die, or around 1.3 million people.
Cruise Ship passenger are a non random sample with perhaps higher co-morbidities.
The cruise ships analysed are non-random sample: “at least 25 other cruise ships have confirmed COVID-19 cases”
Being on a cruise ship might increase your risk because of dose response https://twitter.com/robinhanson/status/1242655704663691264
Onboard IFR. as 1.2% (0.38-2.7%) https://www.medrxiv.org/content/10.1101/2020.03.05.20031773v2
Ioannidis: “A whole country is not a ship.”
Perhaps other on the ship had already cleared the virus and were asymptomatic. PCR only works for a week. Also there might have been false negatives. I disagree that the age and comorbidity structure can only lead to skewed results by a factor of two or three, because this assumes that there are few asymptomatic infections (I’m arguing here that the age tables are wrong).
In my post, I’ve argued why the data out of China might be wrong.
Iceland’s data might be wrong because it is based on PCR not serology, which means that many people might have already cleared the infection, and it is also not random.
That’s the Grand Princess, not the Diamond Princess.
Cheers- corrected.
The DP data is commonly misunderstood. Influenza and COVID-19 (probably) both have a strongly age dependent IFR curve. The “COVID-19 is similar to influenza” model predicts IFR in the 1% range for a retiree age distribution like on DP but 0.1% range on the US age distribution. To a first approximation almost all the deaths are in the 65+ age groups, which are a small fraction of US population but about 50% of DP—it was a geriatric cruise.
So the DP is fairly strong evidence for influenza like mortality. I have an analysis here with more details, and this post by Nic Lewis has a more detailed analysis which considers a few more factors.
FWIW I tried to do an age adjustment for the Diamond Princess myself and what I got was that the 1.4% IFR for the cruise demographics translates into a 0.3% IFR for US demographics (factoring out gender adjustments). I think you could argue that because women were underrepresented on the cruise ship, the adjustment should be greater, so 0.25% is plausible. That said, this doesn’t yet factor in that the people who are medically worst off probably don’t book cruises, so my best-estimate adjustment is maybe 0.4% with a lot of uncertainty. I agree that the people who use the Diamond Princess as evidence for an IFR around 0.9% or higher seem to be making a mistake. At the same time, I do think the Diamond Princess is at least weak to moderate evidence against the 0.125% figure Ioannidis arrived at, or the 0.1% figure that I’ve seen discussed elsewhere.
I don’t really know how this compares to flu mortality, but I found myself somewhat skeptical about the claim I quoted above. You seem to get a 10x update for your age adjustment, whereas my update was only about 5.7x (before factoring in harder-to-quantify assumptions that IMO reduce the factor a bit more even).
(I made a huge mess of my calculations and I don’t recommend clicking on the following link, but just so people see that I’m not just making this up, here’s some evidence that I did something with numbers. Could also be that I neglected some considerations. For factoring in how much overrepresentation of age bracket 70-79 changes things, I based the adjustment off of previous estimates on how strongly Covid19′s IFR is age skewed. I’d imagine that this adjustment was uncontroversial because whether you subscribe to the low IFR theory or not, probably there’s no reason to question whether the proportionalities of the attack rate are correctly reported?)
For a really rough analysis, the overall IFR on the DP was probably about 1% (10 deaths / 1000 infections) after adjusting slightly for false negatives / missed tests.
All those deaths are 70+ age with an in IFR in that group ~2%. About 10% of the US population is in the 70+ bracket, so the projected IFR is ~0.2%. However about half the deaths were in the 80+ age bracket, and if you do a more fine grained binning it’s probably more like 0.15%, but it’s not a high precision estimate.
Very good analysis.
I also thought your recent blog was excellent and think you should make it a top level post:
https://entersingularity.wordpress.com/2020/03/23/covid-19-vs-influenza/
Out of 645 tests done in Colorado on first responders and their families, there were zero positive results.
This seems pretty hard to evaluate because with a large number of published pre-prints on the outbreak, it’s not very surprising that there would be many suggesting higher-than-expected spread. The question is how that weighs up against the opposing evidence, and to evaluate that I’d have to look at all the opposing evidence, which I don’t want to do. That being said, broadly I am unconvinced. Notes on some of the dot points:
Presumably some of these people are hypochondriacs or have the flu? Also, I bet people with symptoms are more likely to use the app.
This isn’t very important but 0.12 is only 1.5 orders of magnitude smaler than 4.19, which I wouldn’t call “several”.
Couldn’t this be explained by those populations travelling more, shaking more hands, meeting more people, etc.?
Iceland has 2 deaths and 97 recoveries. I would say that isn’t good evidence for an IFR of under 0.3%. Admittedly the number of deaths so far is 0.2% of the total number of cases, but given exponential spread most of the cases will be new and won’t have had time to die yet, so the deaths to recoveries ratio seems more important (although upward-biased given who gets tested).
I’m particularly unimpressed by the dot points noting things that happened to very few people:
Dengue “popping up in unusual places”, makes me think that it’s more likely that massive Dengue outbreaks in Latin America might have a high proportion of C19.
This is just to lend credence to the paper that shows there had been 2 million infections in China in January.
I find it very unlikely on the face of it that China, or any country for that matter, managed to suppress completely a disease so contagious that it’s now on almost every country on earth.
No, this is different. I’m not just cherry picking the tail-end of a normal distribution of IFRs etc. The Gupta study in particular and some of the other studies suggest a fundamentally different theory of the pandemic.
Yes, but similarly there are many asymptomatic people who do not use the app. The King’s Professor seems to find this number convincing.
Tom Hanks, Prince Charles and Boris Johnson don’t talk meet more people everyday then your typical Uber driver cashier etc. There millions of people working in retail. We don’t see them all having it. My theory is that they’re tested often and not that “there’s a lot of C19 in Westminster”
Crucially depends on the asymptomatic rate, which might very well be very high.
The point remains: given that some people have such a different theory, it’s unclear how many supporting pieces of evidence your should expect to see, and it’s important to compare the evidence against the theory to the evidence for it.
With all due respect it’s not that hard to get data that you yourself find convincing, even if you’re a professor.
They do meet more different populations of people though. So if a small number of cities have relatively widespread infection, people who visit many cities are unusually likely to get infected.
Not likely. About 1% of Icelanders without symptoms test positive, and all the stats on which tested people are asymptomatic that I’ve seen (Iceland, Diamond Princess) give about 1⁄2 asymptomatic at time of testing (presumably many later get sick).
Yes, that’s what I’m trying to do here. I feel this is a neglected take and on the margin more people should think about whether this theory is true, given the stakes.
“”Our first analysis showed we’re picking up roughly that one in 10 have the classical symptoms,” he said. “So of the 650,000, we would expect to see 65,000 cases.
“Although you can have problems of self-selection and bias, when you’ve got big data like this you tend to trust it more. What we’re seeing is a lot of mild symptoms, so I think having this data should help people relax a bit more and stop seeing it as an all or nothing Black Death situation.
“Other symptoms are cropping up. Thousands of people are coming forward to say they have loss of taste, and we may start to see clusters of symptoms.”″
https://www.telegraph.co.uk/news/2020/03/25/monitoring-app-suggests-65-million-people-uk-may-already-have/
You’d expect to see people to many severe cases amongst people who travelled for business a lot in January and February.
I don’t quite understand what you’re saying here.
It looks more like you listed all the evidence you could find for the theory and didn’t do anything else.
I don’t think this is actually how selection effects work.
Those people are less famous so you wouldn’t necessarily hear about them.
That the asymptomatic rate isn’t all that high, and in at least one population where everybody could get a test, you don’t see a big fraction of the population testing positive.
That was precisely my ambition here—as highlighted in the title (“The case for c19 being widespread”). I did not claim that this was an even-handed take. I wanted to consider the evidence for a theory that only very few smart people believe. I think such an exercise can often be useful.
The professor acknowledges that there are problems with self-selection, but given that there are very specific symptoms (thousands of people with loss of smell), I don’t think that selection effects can describe all the the data. Then he just argues for the Central Limit Theorem.
There’s no random population wide testing antibody testing as of yet.
I don’t think these tests are similar. See here,
Cheers—have taken this point out.
Very first serology data is coming out. 164 close contacts tested by PCR and serology. 16⁄164 of contacts PCR+ & all PCR+ also serology+. Additional 7⁄164 were serology+ but PCR-. Overall 23⁄164 close contacts + in at least one test; 10⁄23 were asymptomatic. Only about half of people show no symptoms, the rest show the spectrum of reported symptoms as spoken of before. The hope for widespread transmission and flulike death rate is gone.
https://t.co/nPiD6UP1eY?amp=1
meta but why argue about this if there is high confidence that serological data will determine things one way or another over the next week or two? Is there decision relevant leverage?
If you can predict the result of the data ahead of time, that seems very important for making decisions (eg. predicting stock market moves).
I don’t know how this would work. Trading on info always seems great until it comes time to actually do it. eg you get actual inside knowledge that suggests a trade, but wait, how do you know whether this inside knowledge is sufficiently leaked that it is priced in?
Many LW people have now taken the view that the market is not efficient when it comes to black swan events.
I’m having trouble imagining a concrete scenario where it would be possible to use this information to gain an advantage.
Whether or not a published paper has a hypnothesis that’s true or not true is a known unknown and not a unknown unknown or a black swan.
The idea is that the context surrounding this pandemic is unique.
Will all the black swam ETFs (like taleb assistant one universa) make it more efficient in that direction?
One benefit is that it is a good chance for training. We have a complicated real world question were the answer and even the best way of approaching the answer are currently unknown, but we will know the answer soon. Making predictions and recording the reasoning will allow for retrospectives.
What is your personal point estimate or credible interval for IFR?
If the Gupta study is true, then a rough approximation (ignoring lag) would be that it’s:
IFR = Number of UK deaths (~750) / 36-68% of the UK population (66 million).
So 0.002% to 0.003%.
In Italy, with almost 10k deaths it would be 0.02%-0.04%
In the province of Lodi (part of Lombardy), 388 people were reported to have died of Covid-19 on 27 March. Lodi has a population of 230,000, meaning that 0.17% of _the population_ of Lodi has died. Given that everyone hardly has been infected, IFR must be higher.
The same source reports that in the province of Cremona (also part of Lombardy), 455 people had died of Covid-19 on 27 March. Cremona has a population of 360,000, meaning that 0.126% of the population of Cremona has died, according to official data.
Note also that there are reports of substantial under-reports of deaths in the Bergamo province. Some reports estimate that the true death rates in some areas may be as much as 1%. However, those reports are highly uncertain. And they may be outliers.
https://www.facebook.com/stefan.schubert.3954/posts/1369053463295040
These numbers support my suspicion that >10% of North Italy has already been infected, with a death rate of ~1%.
If the IFR is indeed .003% (the upper end of your range), then assuming the worst case scenario that 100% of the population of the UK gets infected eventually, only .003%*66.4 million = approx 2000 people will die total.
Would you consider the theory falsified if the death toll in the UK surpasses 2000?
No. My ambition here was a bit simpler. I have presented a rough qualitative argument here that infection is already widespread and only a toy model. There are some issues with this and I haven’t done formal modelling. For instance, this would be what would be called the “crude IFR” I think , but the time lag adjusted IFR (~30 days from infection to death) might increase the death toll.
Currently, also every death in Italy where coronavirus is detected is recorded as a C19 death.
FWIW, if UK death toll will surpass 10,000, then this wouldn’t fit very well with this hypothesis here.
The UK death toll currently stands at 10,612 according to:
https://www.worldometers.info/coronavirus/country/uk/
Boy there was a lot of desperate motivated cogniton around a few weeks ago...
@Hauke Hillebrandt
If this update works then I feel like just looking at how the numbers in Italy came together would change your mind about the low-IFR hypothesis.
Alternatively, if the Covid-19 deaths in NY state go above 3,333 in the first week of April, that seems like it would also falsify the hypothesis. (NY state has fewer than one third the population of the UK.) Unfortunately I think this is >80% to happen.
On April 4, the death toll in NY state surpassed 3,333. As of April 10, there are 7,844 deaths.
The death rate data coming in seems to be converging on a 0.5% to 0.7% death per infection rate. Multiple sources have estimated that weeks ago based on age-normalizing the Diamond Princess, and on testing evacuees from Wuhan.
Two serology surveys have now happened in Europe. One was in a hard-hit town in Germany, and one was in a hard-hit town in Italy at the epicenter of its outbreak. In both places, they got approximately a 15% seropositive rate. In Germany, we only have information on deaths with positive test rates and it comes to 0.35%. In Italy, total excess deaths over this time last year are about 2.5x the confirmed positive deaths and account for 0.1% of the population, giving an infection fatality rate of 0.7%. It is easy to imagine that some deaths did not get positive tests in Germany which along with a less-old population could make up for the difference.
From this, I estimate that at least 10% and possibly up to 20% of New York City has been infected, given the delay between infections and deaths. (100*8000 = 800,000, out of about 8 million)
It’s worth noting that the German serology study (it was in the town Gangelt) has been criticized for being poorly presented: https://www.sueddeutsche.de/wissen/heinsberg-studie-herdenimmunitaet-kritik-1.4873480?fbclid=IwAR1mpGCPj21bffeXBe1fGJVeEWc7UlO2DkEP9-XrSCi4sJeh2-Ri_Cahwrw
One point of criticism is that the renowned German experts who were asked to comment on the study say they are skeptical about the antibody tests. They argue that to their knowledge, the only antibody tests widely in use in Germany at the time of the study can’t distinguish between SARS-CoV-2 and other coronaviruses responsible for a third of common colds. Because we are 1 month past the peak of cold season, they argue that the 15% could be largely picking up on false positives for SARS-CoV-2.
Some non-serology blanket RNA tests coming out of Austria.
https://www.theguardian.com/world/2020/apr/10/less-than-1-of-austria-infected-with-coronavirus-new-study-shows
Given a 0.3% current acute infection rate and some epidemiological modeling they estimate 1% of their total population has been infected, with a death rate of 0.77%.
Everything seems to be converging...
Yup. 0.77% is also what I keep stumbling upon when I look into various data points about the IFR! It’s my best guess about where Iceland’s IFR will end up, and very close to my best guess for proper age adjustment for the Diamond Princess.
New, amazing data from New York, as of April 13. https://www.nejm.org/doi/full/10.1056/NEJMc2009316
Hardly an unbiased sample, but of 200+ pregnant women coming into a hospital to give birth that were blanket-tested, 15.3% tested positive.
Of this set of positive tests, only 12% of them were symptomatic on admission, and a further 10% developed symptoms over the course of their 2-day-long stays bringing it to a total of 22% symptomatic upon discharge or transfer. Presumably already-symptomatic women were more likely to be in the hospital already.
Doing a little armchair epidemiology. Let’s assume that half of the deaths of currently infected people have happened, due to the lockdown extending the doubling time from three days to more than a week. We get:
~8000 deaths * 2 / (15.3% of 8 million) = 1.3% infection to mortality rate.
If we assume that there were more symptomatic women who didn’t show up to normal birthing due to going to the hospital for COVID symptoms, we get a lower death rate. If 20% of the total population is infected, we get a 1% mortality rate. Could go lower if the doubling time has slowed less than my assumption, or if people who have recovered constitute a large enough actual segment of the population. Probably can’t account for more than a factor of two though, given known recovery times.
Compare this to what I wrote 21 hours ago, based on serology data from Italy and Germany:
‘From this, I estimate that at least 10% and possibly up to 20% of New York City has been infected, given the delay between infections and deaths. (100*8000 = 800,000, out of about 8 million) ’
I think what ignoranceprior was originally asking was, given all the information you know, what is your best estimate of the infection fatality rate? Best estimate in this case implies adjusting for ways that some research can be wrong, and taking into account the rebuttals you’ve read here.
This is indeed what I meant. Also I was thinking about once-the-dust-settles IFR, not “crude IFR”.
I’m confused why you assume that 36-68% of the population in the UK is infected. I thought, based on comments here, that those numbers were the output of a model that made highly optimistic assumptions about IFR, not an attempt at estimating the actual proportion of infections.
Do you think this is a realistic range for the proportion already infected in the UK?
I’m not impressed by the comment about this paper here on LW or the twitter link in it.
This paper was written by an international team of highly cited disease modellers who know about the Diamond Princess and have put their reputation on the line to make the case that this the hypothesis of high infections rate and low infection fatality might be true.
I think it is a realistic range that this many people are already infected and are asymptomatic. Above I’ve tried to summarize and review the relevant evidence that fits with this hypothesis.
But I’m not ruling out the more common theory (that we have maybe only 10x the 500k confirmed cases). I just find it less likely.
Yes, but when you actually read the paper (I read some parts), it says that their model is based on an assumption of low IFR, and in itself did not argue for low IFR (feel free to prove me wrong here).
That’s true and that’s what they were criticized for.
They argued that the current data we observe can be also be explained by low IFR and widespread infection. They called for widespread serological testing to see which hypothesis is correct.
If in the next few weeks we see high percentage of people with antibodies then it’s true.
In the meantime, I thought it might be interesting to see what other evidence there is for infection being widespread, which would suggest that IFR is low.
I really appreciate your attempt to summarize this literature. But it seems you still believe that the Oxford paper provides evidence in favor of very low IFR, when in fact others are claiming that this is merely an assumption of their model, and that this assumption was made not because the authors believe it is plausible but simply for exploratory purposes. If this is correct (I haven’t myself read the paper, so I can only defer to others), then the reputation or expertise of the authors is evidentially irrelevant, and shouldn’t cause you to update in the direction of the very low IFR. (Of course, there may be independent reasons for such an update.)
Thanks Pablo for your comment and helping to clarify this point. I’m sorry if I was being unclear.
I understand what you’re saying. However:
I realize that the Oxford study did not collect any new empirical data that in itself should cause us to update our views.
The authors make the assumption that the IFR is low and the virus is widespread and find that it fits the present data just as well as high IFR and low spread. But it does not mean that the model is merely theoretical: the authors do fit the data on the current epidemic.
This is not different from what the Imperial study does: the Imperial authors do not know the true IFR but just assuming a high one and see whether it fits the present data well.
But indeed, on a meta-level the Oxford study (not the modelling itself) is evidence in favor of low IFR. When experts believe something to be plausible then this too is evidence of a theory to be more likely to be true and we should update. An infinite number of models can explain any dataset and the authors only find these two plausible.
By coming out and suggesting that this is a plausible theory, especially by going to the media, the authors have gotten a lot of flag for this (“Irresponsible”—see twitter etc.). So they have indeed put their reputation on the line. This is despite the fact that the authors are prudent and saying that high IFR is also plausible and also fits the data.
There’s an Italian village where 0.1% of the population already died with a confirmed diagnosis of Covid-19. Inferring from typical monthly death rates it’s also estimated that the twice as many people died from Covid-19 in that village without an official diagnosis. There’s a bunch of uncertainty about those additional 0.2%, but it would put the fatality rate at 0.3% already. And those figures are from 4 days ago (edit: 6 days ago actually).
Edit: It’s a province and city(!), not a village.
I do not think that can be used as decisive evidence to falsify wide-spread.
This is a non-random village in Italy, so of course, some villages in Italy will show very high mortality just by chance.
That region of Italy has high smoking rates, very bad air pollution, and the highest age structure outside of Japan.
It’s extremely implausible that it would be 10x or 15x higher than what’s expected for the typical Italian village. Besides, other villages like Cremona or Bergamo also seem to be close to those numbers. Smoking or age structure or air pollution doesn’t give you a 10x update.
UPDATE: Wow, I was totally wrong about those being villages. As Stefan Schubert pointed out, those are cities and provinces with tens and hundreds of thousands of inhabitants!
Thanks, Lukas. I only saw this now. I made a more substantive comment elsewhere in this thread. Lodi is not a village, it’s a province with 230K inhabitants, as are Cremona (360K) and Bergamo (1.11M). (Though note that all these names are also names of the central town in these provinces.)
“Czech Researchers claim that Chinese do not work well ”
This seems to be missing a word ;)
Of the many problems of this theory...
Many places need ~10 PCR tests to find one infection while the group tested is often highly pre-selected, such as “symptomatic people with known contacts”. You should have much higher prior it is infected. Some of the numbers proposed in the “tip of the iceberg” framework would actually mean the prior probability of being infected in the “tested group” is lower than in the general population.
With this hypothesis its very hard to make sense of China. Outside of Hubei, China managed to contain the outbreak in large part by contact tracing & testing. However if you assume there is some very high number of cases you don’t know about, it is difficult to explain why contact tracing can influence anything.
How do you explain Korean data with this hypothesis ?
As of March 30, for 410k tests, they detected 9789 cases. They are testing more than 10k people everyday since late February, never got more than a thousand cases on one day and this is biased by contact tracing.
I read the Covid 19 vs Influenza blog post and age mortality from the korean CDC is closer to 10x flu than classic flu.
Overall, Korea is evidence for “low infectiousness, high infection fatality rate”. The fact that containment works in some countries seems to be evidence too.
https://english.alarabiya.net/en/features/2020/03/25/Coronavirus-Iceland-s-mass-testing-finds-half-of-carriers-show-no-symptoms says half the carriers show no symptoms.
Note: half of carriers don’t show symptoms at the time they tested positive, could well be that they show symptoms later.
On the Diamond Princess it was reported that 20% of positives never showed symptoms, and Korea is reporting 30%. So some of those asymptomatics will probably progress, but not all.