the Air Vent

Because the world needs another opinion


Posted by Jeff Id on April 18, 2020

Finally, we are getting some good statistics.   A very interesting study of total COVID infection rate has made its rounds through the news.   The methods appear reasonable and the conclusions are interesting in that infections in that county are 50 to 85 times more cases of infection than our reported.

COVID – 19 Antibody Seroprevalence in Santa Clara County, California

This data leads the authors to the conclusion that the virus is much less deadly than we’ve been led to believe.

A hundred deaths out of 48,000 – 81,000 infections corresponds to an infection fatality rate of 0.12 – 0.2%.

What’s very nice about this paper is that it is statistically simplistic making it easy for most to understand.

At StatNews site, an article I read some time ago reported similar results based on DiamondPrincess data.
Projecting the Diamond Princess mortality rate onto the age structure of the U.S. population, the death rate among people infected with Covid-19 would be 0.125%. But since this estimate is based on extremely thin data — there were just seven deaths among the 700 infected passengers and crew — the real death rate could stretch from five times lower (0.025%) to five times higher (0.625%).
For perspective, the common flu which is much better studied than a new virus has a mortality rate of about 0.08%.   Still, if we assume the us statistics were even 50 times more infections, somewhere around 3 million people are walking around with the virus antibodies already completely unaware of their problem, and we would have approximately a 0.1% death rate for those who have been exposed to the virus.  Several more of these studies will be out soon, the data should be very interesting.

22 Responses to “COVID Stats”

  1. RB said

    At a 0.12%-0.2% infection rate, given NYC deaths of 12,000+, that would mean that ~6-10 million of the city’s 8 million have been infected. This study is good for laughs.

    • Jeff Id said

      I would read how the data was collected more carefully. There could be a 2x error in it but you might consider why ny has such a differential instead of dismissing something so simple as this study. I can think of several reasons for the data, quality of hospital or friendships drawing the wealthy, 50!% higher death/infection than the min from this study, overwhelmed hospitals. Non use of chloroquine azythromycin !!!!!!! Lots of punctuation for the last bit. Most doctors are using it now and cuomo blocked it for a while.

      Whether any of that is true, the study seems quite solid.

  2. RB said

    There are plenty of holes identified in this study already – which includes that the positives in the study are fully within the range of false positives . Besides, there is the selection bias issue. This not only fails for NYC but for plenty of places in Europe including Lombardy. I’m sorry, this is wishful thinking.

    • Jeff Id said

      I agree with the selection bias issue but this study is still strong on its overall count. I also wasn’t blown away by the re-weighting strategy so the initial sampling was at 1.5% but they tweaked it and ended up with like 2.5% (from memory cause I don’t have time to look it up in the link above). I wouldn’t be the slightest bit surprised at a 0.2% mortality rate. New LA study bears this out, claiming 4% infection rate already. That’s pretty huge. And very different from the official numbers.

      The New York counter example isn’t terribly compelling to me as a lot of suburbia folks might seek out the better city hospitals for treatment. Also, I believe the genius governor banned use of hydroxychloroquine in many of them whereas other areas are using it with success.

  3. RB said

    LA study has the some of the same authors, uses the same antibody test. All antibody tests are going to be affected by the same issue of false positives dominating when prevalence rate from positive tests is so low. This is true of anywhere in the world where these tests have been conducted. It is going to be a while before we know whether antibody tests are useful for epidemic control when prevalence is low. NYC deaths are an undercount, there are potentially an additional 4000 deaths yet to be accounted for. I wouldn’t be surprised by a 0.5% mortality rate as a lower bound on the average mortality rate, there are some reanalysis of the Stanford study arriving at this number without taking the very significant recruitment bias into account. Coupled with an R0 that is>2X of flu and therefore higher spread for achieving herd immunity that is a nearly 10X the flu death count. Flu truthers don’t seem to acknowledge how abnormal the death count is from this virus has been. Hospital overload is a damning indictment.

    • RB said

      BTW, Gelman thinks that this study is so bad that the authors should issue a public apology.

    • Jeff Id said

      I haven’t read the LA study but if they are looking at a 1.5% false positive, these aren’t going to work very well for the authors. LA cleared that hurdle in percentage (from the news article). Also the potential for bias is pretty big as you have mentioned. I still won’t be surprised to find out 0.2% or even under based on the princess diamond data. It also depends a lot on how treatments go. The in-vitro studies for hydroxy-c zpack were pretty solid, and there is more and more anecdotal evidence daily. I am sworn to secrecy by a doctor who’s also studied these drugs impact on corona for several decades. He’s fairly positive it works quite well. There is some good data coming from a prophylactic standpoint as well. Watch for who is and is not being treated with it making a difference in the data.

  4. RB said

    LA may not clear that hurdle either if you look at the manufacturer’s product spec sheets for an independent evaluation by China CDC. The Santa Clara study by the same authors checks for presence of either of IgM or IgG antibodies. This has a net false positive of 5/150 (3.33%).

    Click to access COVID19_CDC_Evaluation_Report.pdf

    • Jeff Id said

      It misses positive cases almost equally with the false negative cases? Doesn’t seem great for a study which achieves a 1.5% positive rate. They claimed better test accuracy in the study I thought though. I would have to go back and look at the specific test kit to know. Have you done that work RB?

      • RB said

        They use this particular kit. For specificity (used to characterize accuracy of false positive in serology), they assume 99.5%. But their test is for either IgG or IgM. The manufacturer validation data (not China CDC) has 369/371 for IgG and 368/371 for IgM. This would give 5/371 false positives for a test where we do not know if they overlapped. It appears that they used only the IgG data along with their own 30 samples which were correctly evaluated. The independent China CDC test gave 4/150 for IgM and 1/150 for IgG. The one IgG case here overlapped with IgM, so perhaps you could assume false positives as 4/150. Either way, when prevalence is low, they are undoubtedly counting false positives – possibly 2 false positives for every true positive. I saw four different reanalyses all arriving at peak prevalence estimates neighboring 1% such as this one

  5. Jeff Id said

    In a sampling of 3,000 people taken at grocery stores across the state, the survey showed 13.9% of those had coronavirus antibodies, indicating they had previously been exposed but have overcome the disease.

    19.5 million people – 2.71 million in New York state have had the virus. 11000 died ~ 0.4% fatality rate which stinks as if the number is true, it may be higher due to lag in patient outcome.

  6. RB said

    I was just going to say – NYC had 21% infection rate which is consistent with the peak prevalence estimate of 0.5% mortality rate from Santa Clara I posted above estimated from noisier data.

  7. RB said

    NY state test is for IgG, I’m not an expert, but looks to be more accurate than IgM. But their website says 93-100% specificity . Not great to have such a wide band but it is what it is.

  8. Matthew W said

    I think it was “known” much earlier that the China Wuhan Virus wasn’t the impending apocalypse that it was advertised.
    It really sucks to live in Illinois or Michigan.

  9. Jeff Id said

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: