Two studies approaching publication indicate that the number of people infected with 2019 novel coronavirus may be much greater than anyone had previously thought—as much as 85 times greater. If the studies (one by researchers out of Stanford University School of Medicine, the other by the University of Southern California and the Los Angeles County Department of Public Health) are accurate, there is an almost unending list of good news. First, it appears that the novel coronavirus sparks serious disease in a much smaller percentage of the population than previously thought. Likewise, the case fatality rate would be hugely lower—something like 0.2%. Finally, this would also suggest that a reasonable percentage of the population, maybe as much as 4-5%, has already been exposed to the virus. That’s not the kind of number needed for true herd immunity, but it’s much, much greater than the confirmed case count would suggest.
This looks like such fantastic news that there’s a very good reason the pre-press articles have been widely circulated, linked, and promoted. This isn’t just the kind of thing that everyone has hoped for from the beginning: It’s everything hoped for plus the backing of some very significant players in medical research. This is genuine, unadulterated sunshine.
And now, let me rain on that parade—starting with the funeral of a janitor in Georgia.
Both of the new studies are based on data collected using blood antibody tests, which are in a way the gold standard when trying to find the extent of a disease outbreak. Unlike some other tests, an antibody test should detect not just those with an active case of disease, but those who have had it in the past. In this case, the Santa Clara County survey was conducted on a pool of 3,330 volunteers and indicated that somewhere between 2.4% and 4.2% of the population there had been exposed to SARS-CoV-2. That’s between 50 and 85 times the official number.
The Los Angeles County survey involved 863 people. Based on the results, USC and the county health department calculated that between 2.2% and 4.1% of the county’s adult population has been exposed to SARS-CoV-2. Those numbers are about 28 to 55 times the official number of confirmed cases in Los Angeles County. Based on the existing number of deaths, these results would move the case fatality rate of COVID-19 from 5%, down to a more flu-like 0.2%.
These are two apparently well-conducted tests by two different groups of medical professionals. The number of samples in the Los Angeles test may seem a little small to model such a large county, but the health department appears to have taken actions to match the distribution of samples to the population. Also, the two outcomes appear similar enough that the authors are confident enough to provide a more or less authoritative conclusion: Many more people have been exposed to SARS-CoV-2 than official testing would suggest, meaning that SARS-CoV-2 is much less dangerous than the official case counts would suggest, meaning that a much larger outbreak would not generate the kind of huge, multimillion death count that earlier models (like mine) have suggested.
And now, here is why all these very smart people are very likely very wrong ...
On March 12, a well-attended funeral was held in Albany, Georgia, for retired janitor Andrew Jerome Mitchell. By all accounts, it was a warm and loving sendoff for a man with a large family and many friends. As The New York Times detailed last month, there were all the homemade casseroles and sweet desserts that you would expect from a big southern funeral in an African American community. The only problem is that one of the visitors to this funeral, which was attended by about 200 people, was carrying COVID-19. This was still some time before social distancing guidelines were introduced in the area. Within a month after the funeral, two dozen of those who attended were seriously ill. Six were dead.
What does one thing have to do with the other? The Georgia funeral is just one of several such cluster cases in which the victims of an outbreak can be traced back to an original source. The spread of the disease from an initial infection in a South Korean church to 5,080 confirmed cases was tracked case-by-case by officials in that country. A birthday party in Connecticut spread a record of infections, hospitalizations, and deaths that spanned multiple continents. A similar scenario was at the heart of the original outbreak in Italy.
And here’s why that matters: If COVID-19 were really 28 times, or 55 times, or 85 times less threatening than the official case count suggests, none of these events would have happened.
The news from Tuesday evening that at least two deaths in Santa Clara County, California happened weeks earlier than expected might be taken as another sign that the virus had been circulating longer, and more widely than expected. However, this takes back the date of known exposures only a couple of weeks. Rather than showing that the disease spread widely and silently, it shows that the disease announced itself with undiagnosed deaths within a couple of weeks of the first known case in the United States. The realization that COVID-19 was circulating in the United States came with the diagnoses of a community-spread infection at the end of February, also in California. The deaths on February 6 and 17 indicate that, even when the pool of infected people was small, deaths were already occurring.
Imagine the funeral scenario with two sets of numbers: one in which COVID-19 causes illness in about half of those infected, and death in somewhere around 4%. Then imagine the same situation, only with a disease that causes a serious illness about 1% of the time and kills at a rate less than 0.2%. One of these scenarios produces an outcome that looks like the actual results in Albany, Georgia. The other simply does not.
In the Stanford paper, there is a mention of a similar study that was done in the town of Robbio, Italy, in which approximately 10% of the population was found to have had exposure to SARS CoV-2. This is true. Robbio (population 5,861) was tested in full and indications were that over 500 people had exposure, which was about 10 times the number that had been confirmed by previous testing in that town. However, testing in Italy has been as constrained as that in the United States, with most of those tested restricted to those displaying severe symptoms. In many cases, even those hospitalized haven’t been tested. In Robbio, the number of previous positive tests was essentially the same as the number who had become seriously ill. Which means the 500 cases overall is exactly what might have been expected in the town.
The paper fails to mention another Italian town, Vo, where the 3,000 citizens were all given a test by March 10. That study was done by placing the entire town in quarantine and double-testing every individual. What resulted was 89 positive cases, about half of which were asymptomatic at the time of testing. Over the course of this study, researchers found that about 70% of those infected remained asymptomatic or had only very light symptoms over the course of the infection. The other 30% developed more serious cases of the disease.
There’s another good example that was famous, or notorious, just a month ago before it disappeared from the headlines—that of the Diamond Princess. The Diamond Princess luxury cruise ship sailed from Japan on January 20. On January 25, a single passenger with COVID-19 symptoms boarded the ship in Hong Kong. That passenger was diagnosed in transit, and passengers were quarantined in their cabins. Over the next month, passengers who completed a 14-day quarantine were repatriated to their home countries. Every single person aboard the ship, including 3,711 passengers and crew, were ultimately tested for COVID-19. Many were tested multiple times. Ultimately, 712 had positive results. Of these 712, just under half were asymptomatic. However, 9.7% inquired intensive care and 1.3% died.
The numbers in Vo and the Diamond Princess mesh with results elsewhere and certainly seem to mesh with the data from South Korea, Singapore, and other nations that have practiced widespread testing and effective case management. The incidents that we’ve seen of COVID-19 introduced into a population, whether that’s the Diamond Princess, Italy, Connecticut, or that Georgia funeral, simply can’t be squared with a disease that’s as relatively mild as the one suggested by the Stanford and Los Angeles data.
And there’s another, somewhat larger example. If the virus were really generating the rate of hospitalizations and deaths that the results from the two California studies suggest, then the number of people hospitalized in New York City would require an infection rate of roughly … 100%. Taking mid-range numbers from either of the two California studies gives a case fatality rate of around 0.15%. Projecting that against just the number of deaths that have happened so far in New York gives over 13 million cases. The entire city of New York has a population under 9 million. Again, the numbers from the California studies fail to describe the disease in the real world.
Also, reaching the kind of infection rate that these numbers suggest would have taken months. Months during which there would have been millions of undiagnosed cases. In fact, to get the kind of numbers under discussion here, would require not only a disease that had been perking around longer than any known data suggests, but one with a higher rate of transmission. One more like measles than SARS.
In short: You cannot take the data generated from either of the two California studies and use it to model the results demonstrated by COVID-19 around the world when it comes to testing results, hospitalization, or deaths. The disease those studies describe, does not explain the bodies in the morgues or the thousands of people in hospitals.
However, you can take the results of those California studies and can generate results that are similar when it comes to some other test results. In the largest attempt to sample an entire population, Iceland has so far conducted 43,000 tests—about 13% of their total population. What they’ve uncovered is 1,778 identified cases, which works out to about 4% of the tested populace. It’s not a perfect fit, as about half of those sampled reported having symptoms (though there was something of a bias in early samples, as people were told to come in for testing only if not sick). In any case, the value generated seems awfully close to that in the California studies.
Which raises a singular possibility: false positives. The Stanford researchers indicated that they had conducted 60 tests of their system before going into the field, using 30 known positive and 30 known negative samples. But even a small number of false positives could replicate the values that all the blood-based tests are seeing, and the strangely consistent low-level numbers suggest that the tests may be sensitive to some factor not seen in the samples used for calibration.
In the case of the Stanford study at least, there’s also the possibility that the researchers unconsciously selected for people more likely to have been ill, because they solicited their subjects on Facebook by telling them they would be getting tested for COVID-19. People who were sick might have been more interested in participating. People who thought they had practiced good social distancing might have been reluctant to come in. It’s hard to know how these may have affected a volunteer pool.
What can be said for sure is:
- Blood sample tests looking at antibodies in Santa Clara and Los Angeles Counties, California, describe a disease where the great majority—something like 90% or more—have little to no symptoms. They also seem to demonstrate a hospitalization rate less than 3% and a case fatality rate of less than 0.2%.
- Widespread testing across multiple countries, case tracing, cluster investigation, and case studies of isolated communities describe a wholly different disease in which about half of cases have little or no symptoms, around 15-20% of patients develop serious respiratory illness, and case fatality is around 4%. That case fatality can be as low as 1-2% with early detection and good care. It can come close to 20% when healthcare systems are overrun.
- The testing system in the United States is wholly inadequate to effectively describe the true scale of infection. However, the results of the two studies cannot be squared with the actual on-the-ground outcomes in New York City or with clusters of cases confirmed through case-tracing.
One other thing that definitely needs to be restated here: I’m not a doctor, or a medical researcher, or qualified in any way to critique either the methods of these studies or their results, except as they are subject to basic statistical analysis. I’m trying to skate something of an edge here, because I want to believe these results. But I can’t see any way to make the numbers generated from these studies work in comparison to what’s happened everywhere that COVID-19 has become epidemic.
New York has its own antibody test in the field at the moment, which should generate some interesting results for comparison in the next couple of weeks. The New York test is newly developed and unlike the test employed by Stanford it has gone through an FDA approval process. Other tests are still seeking approval.
And now that I’ve cast such a cloud over these sunny outcomes, let me offer an alternative that’s even brighter, albeit one that’s completely speculative. Let me repeat that: this is completely speculative. As in, I pulled it out of my … gumdrops and rainbows dispenser.
One way to square the relatively upbeat results of the California studies and the grim march of deaths elsewhere, would be if they’re not describing the same disease at all. Maybe the blood tests are not catching COVID-19, but COVID-19’s less damaging cousin—let’s call it COVID-Mellow. If there’s another disease out there that shares enough genetic similarity to tick the flag on a COVID-19 test, but which is spreading through the population without generating a notable uptick in deaths and hospitalizations, it could harmonize the results of the two data sets. In fact, if there really is a second version of SARS-CoV-2 on the loose, it would surely be even more benign than the Santa Clara or Los Angeles data suggests. It would have to be pretty damn near completely harmless.
How sunny could this scenario be? Get out your shades. Should these two viruses be similar enough that catching one provides protection against the other, then 4% of the population may have already bought a free ticket to post-COVID land, and millions more could be right behind them. In fact, if COVID-Mellow exists, it would be essentially a ready made, self-administering vaccine against COVID-19.
And that, folks, is the end of your purely speculative entry from someone known to write science fiction novels. Until my COVID-Mellow theory is confirmed … please be sure you don’t catch the other one.