Ricochet is the best place on the internet to discuss the issues of the day, either through commenting on posts or writing your own for our active and dynamic community in a fully moderated environment. In addition, the Ricochet Audio Network offers over 50 original podcasts with new episodes released every day.
A New Hope?
How many people have been infected with SARS-CoV-2 without being detected?
As everyone knows, that is the most important unanswered question in this sordid pandemic to date. Infection with SARS-CoV-2 is often mild or even completely asymptomatic, and with the tremendous gaps in testing in the US it is self-apparent that many cases have gone undetected, both by the authorities and by patients themselves. But knowing this figure is crucial for two major reasons: first, anyone who has been infected with SARS-CoV-2 (and survived) is now presumably immune and can re-enter society with few worries. But second, the larger the percentage of undetected cases turns out to be, the less menacing the virus becomes. Yet despite well over 5,000 scientific papers on the novel coronavirus in the past four months, there has not been a single reliable “antibody study” to date. Until today.
As announced by Ricochet contributor Jay Bhattacharya on the Ricochet podcast several weeks ago, he and his Stanford Medicine colleague Eran Bendavid have been conducting an antibody study to determine what percentage of the population of Santa Clara County, CA is “seropositive” – meaning their blood contains antibodies indicating prior infection with and immunity to SARS-CoV-2. And today they announced their results, and they’re huge:
They report that the actual number of SARS-CoV-2 infections is 50- to 85-fold higher than the number of reported cases.
But let’s back up a second: how was this study performed?
To get this marquee number, the researchers first recruited 3,300 residents of Santa Clara County (one of the heaviest-hit counties in CA) through targeted Facebook ads. These volunteers then provided blood samples and filled out a questionnaire in early April. The blood samples were tested for the presence of antibodies to SARS-CoV-2 (which indicate prior infection and presumable immunity to the disease) using recently-developed, commercial, single-use rapid antibody tests.
Of the 3,300 tests performed, about 1.5% were positive for SARS-CoV-2 antibodies. However, there are two problems with this raw number: first, we don’t know how representative the study population was, and second, we don’t know how accurate the new (and unvalidated) antibody test was. Luckily, the team had both of these bases covered: the former using the questionnaires and the latter using an archive of samples from Covid-19-positive and -negative patients at Stanford Medicine.
Normalizing the data for these confounding factors yielded a range of 2.49-4.16% of the residents of Santa Clara County who have already been infected. Contrast that with the official rate (by PCR testing) of less than 0.05%, and the difference is formidable: namely, 50- to 85-fold greater.
So what does this mean? Very little and a whole lot.
As always, this is still a single study with the usual limitations. Because California has low testing rates but has had coronavirus cases since February, we would expect the gap between confirmed cases and actual cases to be greater in CA than in most other states. Furthermore, the results may also be skewed high since the study population was self-selecting, and people who think they might have had Covid-19 but never got diagnosed might logically have been more motivated to enroll.
But even taking those caveats into account, these results are huge. Even dropping their magnitude by half, they’re huge. For one, they clearly show how easily SARS-CoV-2 infection can go completely undetected. For another, the nominal infection fatality rate based on these figures is somewhere around 0.12-0.2%, which is obviously much lower than either the crude rates of 1% or more endlessly repeated by the media, or even rates like 0.66% calculated in a moment of reflection by Imperial College.
So does this mean we can drop the lockdowns and immediately return to normal? I’d say not too quickly. First off, even with these promising figures, Santa Clara County (and likely most of the West Coast) is still far from herd immunity. This data seems to lay the notion of a “good bug vs. bad bug” to rest. And perhaps most importantly, there has also been another study taking place in New York City with a lot more than 3,300 people: and the result is that nearly 0.14% of the total population seems to have died from the coronavirus within a month. The fact that the absolute fatality rate in NYC was higher than the coronavirus fatality rate suggested by Bhattacharya and Bendavid suggests we should still proceed with caution.
Why is there such a difference between NYC and Santa Clara county? The low fatality rate in Santa Clara County might be a reflection of a much healthier population than NYC. Perhaps there’s some other aspect that makes the virus more deadly we’re not yet aware of, like the route of transmission or infectious dose. Maybe the virus just hasn’t spread enough yet in Santa Clara County to hit the (few) nursing homes and assisted living facilities.
But even with those grains of salt, we can make several confident conclusions:
a) perhaps we shouldn’t take the numbers in this study literally, but we should definitely take them seriously: under-ascertainment is likely rampant throughout the US, even if to differing degrees.
b) the virus is much more complicated than just “this is the deadliest thing since sliced heads!” While it has shown some teeth, it has also now given us multiple signs of not being the monolithic killer it has been portrayed to be.
c) we desperately need similar studies to confirm these results in other regions – and soon. The CDC and several other institutions have already started doing so, but on timelines lasting months. This is too slow in any case, but especially if it turns out that the virus is considerably less dangerous than originally believed. The present study demonstrates that speed is not incompatible with scientific quality if done right.
d) the results of this study need to be broadcast throughout the country – intelligently. Simply proclaiming “it turns out the coronavirus is no more dangerous than the flu after all!” is ignorant and counterproductive. The last 3 weeks in NYC have shown that statement to be untrue. But we can still vociferously point to this study as a sign that lockdowns need to be reconsidered.
d) every policymaker in the US needs to start drawing up much more aggressive plans to re-open their jurisdiction pending other indications that the virus may be less of a threat than previously believed.
If you’ve made it this far, thank you for your perseverance!
Published in General
I liked that line too. Most mass media presents this virus as though it has a near 100% fatality rate.
Yes, the analysis of hospital beds should have included distinct geographies, as opposed to the entire country at large. Perhaps the number of hospital beds within some distance of the regular census statistical regions.
Although, since health insurance is restricted to the state level, counties would probably do as a measure of where people are going to recieve care. So X number of hospital beds/ICU by county analysis would better serve the country.
No one is going to medivac a Covid patient from NYC to say, Eu Claire Wisconsin. Which is what the national bed analysis implicitly assumes.
Thanks for sharing that, I hadn’t been aware.
There’s far too little information provided in the article to put much confidence in the raw numbers cited. As this entire post shows, just reporting the raw number of positive antibody tests is not even half the battle.
All that being said, I’ll throw scientific caution to the wind and say: this finding seems much more in line with what I would expect than the numbers provided by Bendavid and Bhattacharya.
If 1/3 of the population of Chelsea was already seropositive, that would mean they have made it about halfway to herd immunity while having lost about 1 in 500 (80 in 40,000) residents to the virus within a month. That’s a crude infection fatality rate of 0.3%, which is more than twice as high as the upper bound of the crude IFR in the Bendavid/Bhattacharya study yet obviously much lower than many of the rates we’ve seen published elsewhere.
In general, it conforms with a degree of pathogenicity that’s well above the hysteria promulgated by much of the media, yet still much higher than risk suggested by Jay Bhattacharya and many other prominent dissenters (mostly on the political right).
I do fear that we’re being unnecessarily hampered in our ability to respond to this virus because our collective psychology is unable to conceive of a risk category between “just a severe case of seasonal flu” and “ZOMG it’s 1918 all over again!”.
“Pathogenicity.” What a cool word!
Wrong thread
I think I found the answer on the CDC website. The CDC does not count asymptomatic cases of flu.
This is also reflected in a chart on the CDC website which shows symptomatic cases as the foundation for its estimates. I can’t tell from the website whether all of the annual estimates from 2010 on are based on just symptomatic illness but, if so, it potentially changes the CFR if you are using it to compare with all COVID cases including those that are symptomatic.
For instance, the year with the highest number of flu cases (2017-18) shows a CFR of 0.137 which would be closer to 0.080 if asymptomatic cases were included. The year with the highest CFR (2014-2015) would go from 0.170 to about 0.097.
I agree it’s not a show stopper. But what is going to be quoted is the 50-85 number and then people are going to run with it and it’s likely high.