Claim: The true infection-to-fatality ratio is definitely about 0.5% to 1%, and most probably around 0.7%, with significant long term morbidity in at least several percent of survivors. Notions that this disease is already widespread or that it has flulike mortality and morbidity or most people are asymptomatic are definitively disproven.
This has been independently estimated in this range before, based on normalizing data from the Diamond Princess and areas where testing was thorough
Given a 0.3% current acute infection rate and some epidemiological modeling they estimate 1% of their total population has been infected at some point, with a death rate of 0.77%. Maybe a few false negative PCRs, which would lower that number.
2 - 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 results 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.
Hardly an unbiased sample, but of 200+ pregnant women coming into a hospital to give birth that were blanket-RNA-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 very-pregnant 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, or that there is a good stock of people who have recovered in the city, we get a lower death rate. If 20% of the total population was ever infected, we get a 1% mortality rate. 30% ever infected, 0.67%.
EDIT: 4 - Apparently there is a similar maternity ward study in Stockholm, revealing 7% positive. There have been 550 deaths there, and a population of 2.3 million. If we again assume half of current cases that will die has died, we get a infection to fatality ratio of 0.68% without further corrections. I suspect they haven’t crushed the doubling time as much as NYC, raising this number, which then can get lowered down again as I did above.
EDIT: 5, a meta analysis of a whole bunch of research comes to exactly my original conclusion, 0.5% to 1% with a central tendency of 0.8%.
There can be false negatives, but at this positive level false positives are less of an issue than at low levels. Also, they apparently specifically grabbed people out and about at grocery stores, so very sick people may have been excluded, pushing levels down. On the other hand, people shopping might be more likely to pick it up.
EDIT: Apparently this test also only detect IgG, which is the type of antibody that rises last and can take two weeks or more to be detectable in some people after symptoms develop.
That would be IFR when the sick can be appropriate treated, right? I think it can be >>1% when hospitals are overwhelmed. It also obviously depends on demographics, prevalence of diabetes, etc.
Indeed! Anti-inflammatories, oxygen therapy, anticoagulants, blood pressure management, eventually invasive ventillation (though that seems less effective than was previously thought).
I suspect the United States will have a substantially higher IFR than Europe due to all the obesity and metabolic disease, and that when the ICUs pop it also rises.
New information. The Italian town of Vo was blanket-RNA-tested twice in late February and early March. A total of 3% of the town tested positive, and they were able to lock down this subset and shut down transmission from continuing in the town indicating they caught enough of the asymptomatic-but-transmissive carriers.
Here we have a detailed analysis of these positive-testing people.
43% asymptomatic all the way through.
~20% hospitalized. This means almost 40% hospitalization of people who were symptomatic. Critera for hospitalization is an interesting question, as is the age breakdown of the town.
One death, in a town in which a total of 82 people tested positive. That one death was what alerted authorities to the outbreak in the town, so we need to take it as a given rather than taking it as even weak evidence of an over 1% fatality rate.
No cases among hundreds of children, even in houses with symptomatic family members. Extra cases among the elderly (as in 1% of 20 year olds versus 6% of 70 year olds). Small numbers but significant. Unclear if that means they are not getting infected or are not producing long-lasting infection that is detectable or if social structure has something to do with it.
No obvious difference in symptomatic versus asymptomatic across the age distribution, subject to sample size.
More men took more than 2 weeks to clear the virus than women.
Definite confirmed asymptomatic people passing it on, and presymptomatic people passing it on.
3% of the town testing positive via PCR in the beginning of March is compatible with 15% of other towns in the area testing positive via serology a month later, as that would be less than 3 doublings. They find a ‘serial interval’ of only about 7 days in their contact-tracing data from before the lockdown, and 10 days after the lockdown, and a replication number before the lockdown of about 3 and 0.14 during the lockdown. That’s a doubling time of well under a week before the lockdown. The rest of Italy’s lockdown didn’t include such good contact tracing and thus still was probably doubling a few times during that period.
Yeah, it’s been clear for some time that the IFR is about 0.5% and that about half the cases are asymptomatic. Give or take 30% on each. The reported variations are mainly due to testing or age/health condition bias.
Have you got a source for that ‘about half the cases are asymptomatic’? I was under the impression that far more cases show symptoms eventually, and that the studies showing half of the infections are asymptomatic add the disclaimer ‘so far’, which means very little if the spread is growing exponentially with a doubling time of several days.
The Diamond Princess was 50% asymptomatic after 2 weeks of [in cabin] quarantine. It would be nice to have more follow-up on it, but that’s already much better than most measures of asymptomatic cases.
The end data is that about 20% of the Diamond Princess was asymptomatic all the way through. They were particularly old, though.
In Iceland 50% are asymptomatic upon first test, then some progress. The data in Korea suggests 30% asymptomatic.
Small number statistics of government officials supports this too. 7 total congressmen have tested positive or been presumed positive. One or two were hospitalized, some said “it hit me hard” or “my case is mild”, and 2 out of 7 (Rand Paul and Joe Cunningham) reported either zero symptoms or only loss of the sense of smell.
Where do you get your data on the Diamond Princess? As far as I know, there are no updates on symptoms. Perhaps you get it from this, which is not data, but an inference?
Is there reason to believe the raw numbers are more accurate estimate of the rate than the model prediction? Also, what are the type-1 and type-2 errors of the tests used on the Diamond Princess? I heard some early reports that both of these might be significant, but then never heard anything about them again.
I checked that link above and followed their references to find other datasets, but two of them are in Japanese, one only deals with self-selected patients who showed symptoms, and the last two have small sample size (12 patients, two papers cover the same event).
Update: I have found https://www.eurosurveillance.org/content/10.2807/1560-7917.ES.2020.25.3.2000045, which benchmarks the real-time reverse transcription polymerase chain reaction (RT-PCT) tests. They state zero false positives in a trial with 297 non-COVID-19 samples, although they do retest 4 samples that showed “weak initial reactivity”. Since the non real-time version of RT-PCT is supposed to be even more reliable, this means false positives are presumably not a big deal (even at a pessimistic 4/297 false positive this still means only 41 false positives out of 3063 tests done on the Diamond Princess).
I don’t have a good model to give me any predictions on what reasonable numbers of asymptomatic cases would be, or how truncation influences these numbers. Could you explain why the inference is idiotic, and perhaps give a more reasonable one?
I am trying to find the Japanese government webpage with frequent updates as to the state of patients that were evacuated, still updating the ones that are still in the hospital (! Morbidities...)
Is there a good source for many things we know from the Diamond Princess data? Or even just the numbers so far from DP? I’m not sure how to find that data.
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.
The effects of measures on the spread take weeks to show up in the data.
If the doubling time hadn’t cratered, the hospitalization rate would’ve remained exponential. At the time of posting it was comparatively flat, and I estimated.
The half came from the fact that it usually takes ~3 weeks to die, that the exponential spread had only stopped a few weeks earlier, and a drawing of a triangle and square representing a rise and flat that I drew a vertical line through.
Claim: The true infection-to-fatality ratio is definitely about 0.5% to 1%, and most probably around 0.7%, with significant long term morbidity in at least several percent of survivors. Notions that this disease is already widespread or that it has flulike mortality and morbidity or most people are asymptomatic are definitively disproven.
This has been independently estimated in this range before, based on normalizing data from the Diamond Princess and areas where testing was thorough
https://www.thelancet.com/journals/laninf/article/PIIS1473-3099(20)30243-7/fulltext
https://www.medrxiv.org/content/10.1101/2020.03.05.20031773v2
There are a few robust new pieces of data supporting this now.
1 - Blanket RNA testing in 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 at some point, with a death rate of 0.77%. Maybe a few false negative PCRs, which would lower that number.
2 - 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 results 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.
3 - New test data coming out of NYC.
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-RNA-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 very-pregnant 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, or that there is a good stock of people who have recovered in the city, we get a lower death rate. If 20% of the total population was ever infected, we get a 1% mortality rate. 30% ever infected, 0.67%.
EDIT: 4 - Apparently there is a similar maternity ward study in Stockholm, revealing 7% positive. There have been 550 deaths there, and a population of 2.3 million. If we again assume half of current cases that will die has died, we get a infection to fatality ratio of 0.68% without further corrections. I suspect they haven’t crushed the doubling time as much as NYC, raising this number, which then can get lowered down again as I did above.
EDIT: 5, a meta analysis of a whole bunch of research comes to exactly my original conclusion, 0.5% to 1% with a central tendency of 0.8%.
https://t.co/51b3bJYg3e?amp=1
UPDATE as of 4/23/2020.
I’m so sick of being right.
Seroprevalence in NYC reported as 21%.
https://twitter.com/NYGovCuomo/status/1253353516803993600
There can be false negatives, but at this positive level false positives are less of an issue than at low levels. Also, they apparently specifically grabbed people out and about at grocery stores, so very sick people may have been excluded, pushing levels down. On the other hand, people shopping might be more likely to pick it up.
Pretty much right on the nose...
https://twitter.com/trvrb/status/1253398329766973441
″ If we then take deaths as of today as 17,200 based on excess deaths (https://nytimes.com/interactive/2020/04/21/world/coronavirus-missing-deaths.html), we’d get an infection-to-fatality ratio of ~1%. ” I suspect the true seropositivity is higher than the measured due to selection effects on the net, which would push this down a bit.
EDIT: Apparently this test also only detect IgG, which is the type of antibody that rises last and can take two weeks or more to be detectable in some people after symptoms develop.
That would be IFR when the sick can be appropriate treated, right? I think it can be >>1% when hospitals are overwhelmed. It also obviously depends on demographics, prevalence of diabetes, etc.
Indeed! Anti-inflammatories, oxygen therapy, anticoagulants, blood pressure management, eventually invasive ventillation (though that seems less effective than was previously thought).
I suspect the United States will have a substantially higher IFR than Europe due to all the obesity and metabolic disease, and that when the ICUs pop it also rises.
New information. The Italian town of Vo was blanket-RNA-tested twice in late February and early March. A total of 3% of the town tested positive, and they were able to lock down this subset and shut down transmission from continuing in the town indicating they caught enough of the asymptomatic-but-transmissive carriers.
https://www.medrxiv.org/content/10.1101/2020.04.17.20053157v1
Here we have a detailed analysis of these positive-testing people.
43% asymptomatic all the way through.
~20% hospitalized. This means almost 40% hospitalization of people who were symptomatic. Critera for hospitalization is an interesting question, as is the age breakdown of the town.
One death, in a town in which a total of 82 people tested positive. That one death was what alerted authorities to the outbreak in the town, so we need to take it as a given rather than taking it as even weak evidence of an over 1% fatality rate.
No cases among hundreds of children, even in houses with symptomatic family members. Extra cases among the elderly (as in 1% of 20 year olds versus 6% of 70 year olds). Small numbers but significant. Unclear if that means they are not getting infected or are not producing long-lasting infection that is detectable or if social structure has something to do with it.
No obvious difference in symptomatic versus asymptomatic across the age distribution, subject to sample size.
More men took more than 2 weeks to clear the virus than women.
Definite confirmed asymptomatic people passing it on, and presymptomatic people passing it on.
3% of the town testing positive via PCR in the beginning of March is compatible with 15% of other towns in the area testing positive via serology a month later, as that would be less than 3 doublings. They find a ‘serial interval’ of only about 7 days in their contact-tracing data from before the lockdown, and 10 days after the lockdown, and a replication number before the lockdown of about 3 and 0.14 during the lockdown. That’s a doubling time of well under a week before the lockdown. The rest of Italy’s lockdown didn’t include such good contact tracing and thus still was probably doubling a few times during that period.
UPDATE as of 4/28/2019.
Others coming to this exact same distribution more rigorously.
https://t.co/51b3bJYg3e?amp=1
Compiling rigorous data, the compatible range is circa 0.5% to 1% with a central tendency of 0.8%.
Yeah, it’s been clear for some time that the IFR is about 0.5% and that about half the cases are asymptomatic. Give or take 30% on each. The reported variations are mainly due to testing or age/health condition bias.
Have you got a source for that ‘about half the cases are asymptomatic’? I was under the impression that far more cases show symptoms eventually, and that the studies showing half of the infections are asymptomatic add the disclaimer ‘so far’, which means very little if the spread is growing exponentially with a doubling time of several days.
The Diamond Princess was 50% asymptomatic after 2 weeks of [in cabin] quarantine. It would be nice to have more follow-up on it, but that’s already much better than most measures of asymptomatic cases.
The end data is that about 20% of the Diamond Princess was asymptomatic all the way through. They were particularly old, though.
In Iceland 50% are asymptomatic upon first test, then some progress. The data in Korea suggests 30% asymptomatic.
Small number statistics of government officials supports this too. 7 total congressmen have tested positive or been presumed positive. One or two were hospitalized, some said “it hit me hard” or “my case is mild”, and 2 out of 7 (Rand Paul and Joe Cunningham) reported either zero symptoms or only loss of the sense of smell.
Where do you get your data on the Diamond Princess? As far as I know, there are no updates on symptoms. Perhaps you get it from this, which is not data, but an inference?
Is there reason to believe the raw numbers are more accurate estimate of the rate than the model prediction? Also, what are the type-1 and type-2 errors of the tests used on the Diamond Princess? I heard some early reports that both of these might be significant, but then never heard anything about them again.
I checked that link above and followed their references to find other datasets, but two of them are in Japanese, one only deals with self-selected patients who showed symptoms, and the last two have small sample size (12 patients, two papers cover the same event).
Update: I have found https://www.eurosurveillance.org/content/10.2807/1560-7917.ES.2020.25.3.2000045, which benchmarks the real-time reverse transcription polymerase chain reaction (RT-PCT) tests. They state zero false positives in a trial with 297 non-COVID-19 samples, although they do retest 4 samples that showed “weak initial reactivity”. Since the non real-time version of RT-PCT is supposed to be even more reliable, this means false positives are presumably not a big deal (even at a pessimistic 4/297 false positive this still means only 41 false positives out of 3063 tests done on the Diamond Princess).
First of all, it is very important to distinguish data from inferences.
Second, the inference is idiotic. It’s probably a calculation error, but it’s just not worth reading to determine what went wrong.
I don’t have a good model to give me any predictions on what reasonable numbers of asymptomatic cases would be, or how truncation influences these numbers. Could you explain why the inference is idiotic, and perhaps give a more reasonable one?
Here are some quotes from the paper. What is the simplest model you can make from them? Forget the word “model”; what conclusions can you draw?
Check all the references from https://www.nature.com/articles/d41586-020-00885-w for some data, as well as worldometer.
I am trying to find the Japanese government webpage with frequent updates as to the state of patients that were evacuated, still updating the ones that are still in the hospital (! Morbidities...)
So, yes, it is simply misquoting the source that I cited.
Is there a good source for many things we know from the Diamond Princess data? Or even just the numbers so far from DP? I’m not sure how to find that data.
I’ve looked into this a lot and I agree strongly with this being a good range.
How do you draw that conclusion?
The effects of measures on the spread take weeks to show up in the data.
If the doubling time hadn’t cratered, the hospitalization rate would’ve remained exponential. At the time of posting it was comparatively flat, and I estimated.
The half came from the fact that it usually takes ~3 weeks to die, that the exponential spread had only stopped a few weeks earlier, and a drawing of a triangle and square representing a rise and flat that I drew a vertical line through.