Comment by robbiep
5 years ago
You have to call bullshit in those numbers.
Take a country with good surveillance and reliable death statistics. NZ and/or Australia will fit the bill.
The case fatality rate there is a touch above 3%, mostly in aged care deaths, but with enough in the lower age groups to give anyone who has a comorbidity and is above 40 a quickening of the pulse (854 deaths, 26,942 cases, positive test % has never breached 1% and generally been under half a % so incidence is likely to be close to reported cases).
US CFR at reported numbers isn’t far off 3% and if you take the Excess deaths figures that suggest there has already been more like 250,000 deaths, along with the missed cases, you’re probably closer to the mark. Even if spread is confined tightly to younger age groups with lower death rates, there can be no serious person who believes close to 60m Americans have either previously or Are currently Infected.
What do you think the numbers are?
Why do the CDC's seroprevalence studies [1] consistently show the estimated case counts are much higher than the reported cases (6x in NYC area, 5x in Philly, 9x higher in South Florida)?
Do you not think there is a strong correlation, everywhere around the world, in which people who have minor symptoms or people who have no symptoms do not get tested?
If the case fatality rate is actually 3%, how does a place like Singapore have >50k cases with fewer than 50 deaths? That seems impossible.
1: https://covid.cdc.gov/covid-data-tracker/?#serology-surveill...
I think that if we were to be able to bring up a magical data box and interrogate the world's actual data, the US incidence numbers will be an extrapolation of the CFR of nations that have done a reasonable job of keeping their infections under control (ie health systems not overwhelmed) and have consistently had low positive rates. I see no evidence that the US has been successful in quarantining aged care facilities and think that probably an accurate IFR is 2.5% if infections are normally distributed in the population. I therefore think it's possible 10m americans have been infected (using data from the CDC's excess mortality that suggests 250,000 excess deaths so far this year, and attributing the mismatch between COVID reported deaths and CDC excess mortaliity as missed cases)
I remain sceptical of seroprevalence data at this point in time given multiple other possible explanations such as lack of specificity & cross-reactivity to other CoVs and evidence of moderate seroprevalance in absolutely unexposed populations (0). There are even suggestions that, even should a reasonable proportion of the seroprevalance data be correct it may not be protective, and therefore might as well be ignored. I remain open to better data on this (My scepticism arises largely from my major in biochemistry and the time I spent understanding what and how these things are being tested)
>Do you not think there is a strong correlation, everywhere around the world, in which people who have minor symptoms or people who have no symptoms do not get tested?
No I don't. I think that there are strong political and financial reasons not to get tested in some places which can contribute to smouldering outbreaks, but I don't think it's possible to have an aggressive testing regime and have test positives in the <1% range, consistently. I've been tested once, for free and had my results back in under 24 hours which is pretty standard here in Australia. I think there this argument misses the corollary that should a signifiant percentage of asymptomatic people remain infected and circulating they will eventually lead to an outbreak that will be detected. We have evidence for this sort of outbreak in Melbourne and is the cause of 20+k of infections in Australia, and the corresponding high (3%ish) death rate, as I mentioned earlier largely attributable to aged care deaths.
With regard to the singapore data, I am unable to find good case demographics but in principle agree with the SCMP article (1) that suggests the outbreaks being confined to young and otherwise healthy migrant populations whilst missing the elderly and at-risk. Singapore is a high social contract nation and rates of mask-wearing are high anyway; therefore in Singapore you get a CFR that is reflective of infections isolated to young people. This article is also interesting (2).
(0) https://blogs.sciencemag.org/pipeline/archives/2020/07/15/ne...
(1) https://www.scmp.com/week-asia/health-environment/article/30...
(2) https://www.todayonline.com/singapore/why-singapores-covid-1...
Your link (0) doesn't rebut the seroprevalence studies at all. It is talking about T-cells, which is not what the CDC seroprevalence studies are measuring. The CDC is testing for SARS-CoV-2 IgG antibodies in blood. And look, if the CDC's seroprevalence studies can't convince you, then I'm not going to. They have data on the specificity and sensitivity of their antibody tests, which gives high confidence ranges for the data. They talk about their methods in detail [1]. And these tests are not about calculating herd immunity, not about calculating who might have prior immunity from a related disease, not about saying who can or cannot get COVID again. They are merely tests for very specific antibodies that are formed in a high percentage of people who had COVID. Thus they should be very accurate at determining how many people have had COVID if the proper statistical care is taken in interpreting the results (which it is!).
With ~7M confirmed cases in the US, and your estimate of 10M cases, that would be a multiplier of less than 1.5x from confirmed cases to estimated cases. That is well below any of the seroprevalence estimates, and remember many of these cases happened months ago when there was almost no testing. The positive test percentage in NY was at or above 20% until May! It was over 40% for the first half of April! Also the current positive test rate in the US is ~5%. There still is a lack of testing in many places.
Also using the death rate alone in trying to figure out the number of cases is highly prone to error, since the virus has death rates that can vary by over 10,000x based on age (85+ vs <18) [2]. And this is especially true if many of the people who don't die, also don't even get seriously ill and thus don't get tested. Combining seroprevalence studies, hospitalizations and deaths by age group is a much more accurate method, and there is no way you could reasonably look at the data points and come up with 10M total estimated cases.
1: https://www.cdc.gov/mmwr/volumes/69/wr/mm6932a4.htm
2: https://www.cdc.gov/coronavirus/2019-ncov/covid-data/investi...