Most scientific theories and discoveries arise from the scientific method, with next steps based on the results. There are epiphanies (“Eureka!”*) that bypass this process, but those are rare. If a question is developed into a hypothesis, tested, reviewed, discussed, and yields an answer that furthers knowledge, then the scientific method has been applied; the next steps can be moved onto with confidence. Unfortunately, sometimes wishful thinking is applied instead. Wishful thinking may occasionally have a clear testable hypothesis, but when the test results are ignored or never even generated, then the scientific method has not been followed and the conclusion is baseless.

There is always a lot of wishful thinking, with or without a testable hypothesis. When wishful thinking contains a testable hypothesis, it may seem like the scientific method is being applied. It is when tests are performed that this type of wishful thinking can be differentiated from the scientific method. Any time science is discussed by the public, some people will resort to using wishful thinking to make their point, even when data resulting from the scientific method make wishful thinking appear as pure fantasy.

The Testable Hypothesis

The hypothesis that a significant portion of people in an area have had COVID-19 and are subsequently protected from infection by the novel coronavirus (SARS-CoV-2) is testable. Seroprevalence, which is a measurement of how many people have antibodies, can be determined by screening a sample of the target population for antibodies against the novel coronavirus. Fortunately, this is something that has been tested, although few large-scale studies are available. There was one study done in China in March, another in Spain in April, and a third in the United States in July.

For any study, it is important to recognize the limitations:

1. It may take two to three weeks after the onset of symptoms for antibodies to develop, so newly exposed individuals will still be seronegative. When reviewing PCR studies, this timeline is limited to a 4-7 day window after exposure for a positive result.
2. Specificity and sensitivity of the test is important, meaning they should have minimal detection of other antibodies and be able to detect low levels of the target antibody. The methods state 99.8% and 100% specificity with slightly lower sensitivity.
3. Another limitation is how the sample population was selected and how it relates to the general population. If a study only uses patients from a particular clinic with a particular ailment, that can be extrapolated to the general population, but the level of error is increased.
4. Another significant consideration in a rapidly evolving pandemic is time. Samples collected in February from the United States mean something very different than samples collected at the same time from Italy or China.

China

The three studies referenced above (also linked after this post) all took place at different times which can be compared to the known case numbers and deaths in each region or country to extrapolate the true number of cases and death rates. The first seroprevalence study in China focused on healthcare workers and the people they may have exposed, such as hotel workers at hotels set up for healthcare workers. Among healthcare workers in Wuhan (the capital of Hubei province and site of the first outbreak), seroprevalence was found to be close to 4% and as low as 1% when considering the surrounding areas. The authors acknowledged the limitation of sampling and the fact that newly exposed individuals may not yet be seropositive.

Spain

The study in Spain used a sample population with the same distribution of ages, genders, and other factors as the Spanish population. It took place in late April, so we can work with ~236,000 known cases on April 20 and a little over 25,000 deaths two weeks later. We’re using the deaths from two weeks later to account for the time until people die after being infected. They found a seroprevalence of ~5%; this would correspond to ~2.5 million cases and a fatality rate of just over 1%. This was the first population-based study to try and understand how prevalent COVID-19 was and measure how many people had antibodies against the virus that causes it.

USA

The most recent study samples patients in the US in July. This study was limited in that it used samples from dialysis patients, which has a very different distribution than the population as a whole. The study authors did attempt to normalize these values to the general population, but this is still limited because dialysis patients may not have the same behavioural patterns as other people. While this may add significant variability to the data overall, the impact to regional trends is likely negligible. They identified a nearly 10% seroprevalence, with significant regional variability of <4% in parts of the western US and close to 30% in some parts of the northeastern US. Given that this is from July, we know there were roughly 3.5 million confirmed cases and over 152,000 confirmed deaths in the US two weeks later. If 10% of the population has had COVID-19, that would be ~33 million people, meaning only 10% of all US cases were detected with a death rate just over 0.4%. A key difference here is that the US study also took place during a time of rapidly increasing case numbers, while cases in Spain were leveling out. As far as effects of sampling, dialysis patients are probably going to be more careful about being exposed, potentially reducing their infection rate; they are also probably at a higher risk of death from COVID-19 due to reduced kidney function, but may be at a lower risk of death since they are already undergoing regular medical care.

What do these studies tell us?

From these studies, it seems clear that most cases of COVID-19 are missed by testing, possibly up to 90% of them. This is unsurprising since testing was in its early stages in Spain and was simply never fully implemented in the US. It also puts the death rate, when case numbers are low enough to prevent hospitals from being overwhelmed, at closer to 0.5%.

This is simply what can be gleaned from the studies that have been performed. The scientific method forces us to consider the hypothesis (we can estimate how many people have been exposed to SARS-CoV-2), test the hypothesis (seroprevalence), interpret the results, draw conclusions, and communicate the findings.

Wishful thinking does not require such studies or interpretation of the data; we can pretend we’ve reached herd immunity or that the death rate is whatever we want it to be and everything is fine. The scientific method tells us that we should not make such assumptions when we have the ability to test a hypothesis. If the scientific method is being applied, then all of these steps are performed; with wishful thinking, we can just make an assumption and ignore the work being done to answer these questions.

Making something out of thin air is for trees at high altitudes.

If you use wishful thinking to claim far more people have been exposed that we have data for, you may be right, but you are also making a strong case against herd immunity. Seroprevalence only works if antibodies persist long enough. If the antibodies don’t persist long enough, then we need to be even more careful than the experts initially warned, because there are a lot of people that are at risk of getting this disease repeatedly.

Study on prevalence in China: https://www.nature.com/articles/s41591-020-0949-6

Study on prevalence in Spain: https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(20)31483-5/fulltext

Study on prevalence in USA: https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(20)32009-2/fulltext

Numbers of confirmed cases and deaths: https://www.worldometers.info/coronavirus/

*”Eureka!” here is a reference to Archimedes** and not the state motto of California nor the amazing TV show of the same name.

**Archimedes here refers to the Greek polymath and not the owl (nor the dove) with the same name.