In the face of the current coronavirus pandemic, authorities are taking measures that are having massive negative economic impacts. Some say that these measures are wildly unwarranted. Others, that they are disastrously insufficient. Who is right?
Nobody knows. No one can know what the right course of action is, because the available data to provide a basis for these enormously consequential decisions is woefully inadequate. The most important statistics we need to know to develop a strategy to deal with the situation are the size of the pandemic and its lethality. Current data provides no basis for accurate estimation of these vital metrics.
For example, as of the morning of March 30, Johns Hopkins reports that 143,532 Americans have tested positive for COVID-19, with 2,572 deaths. If taken at face value, this would suggest that the infection rate among Americans is 0.044 percent, with a 1.8 percent death rate among those infected.
But we know that such a conclusion would be inaccurate, because millions of people feeling symptoms are being turned away by testers, and there must be millions more, with and without symptoms, who are choosing not to volunteer for testing. So, without doubt, the size of the infected population is being underreported, perhaps grossly so.
The shortage of testing has led to a situation where only VIPs can be sure of getting tested if they suspect illness. Thus we hear about Tom Hanks, Prince Charles, and Boris Johnson testing positive, but how can any such assemblage provide a statistical base? Well, there is the U.S. Congress, where four House members and one U.S. senator have tested positive for COVID-19, for an infection rate of 1 percent and a death rate of zero. But Congress, while arguably politically representative of America at large, is hardly an optimal sample for medical polling purposes. Plausible arguments can readily be made why congressmen should have either higher (they attend lots of meetings) or lower (they don’t take mass transit) infection rates than the general population.
To get the right answer, we need to random-test the public. We don’t need to test all 327 million citizens — although, when feasible, that would be very desirable for the purpose of identifying immune individuals and putting them back to work. For the purpose of getting a rough estimate of the size of the infected population, we need to random-test only about 1,000 people nationwide. That is the approximate size of most election polls, and, provided that reasonable care is taken to ensure that the sample is representative of the electorate, such polls can generally predict the outcome within plus or minus 3 percent. It’s true that occasionally a 45 percent–polling underdog can pull off an upset victory, but never a 20 percenter, let alone a 1 percenter. Limited polls might not always predict the winner, but they inevitably show who is in the competitive range. For purposes of quantifying the epidemic, that would be an enormously valuable correction to our current state of ignorance.
It is also probable that deaths by coronavirus are being underreported, or overreported, given opportunistic infectious diseases, notably pneumonia. There is reason to suspect this is the case for Germany, which reports 560 deaths in 63,929 cases (compared with 2,612 deaths in 40,751 cases in France), and certainly Russia, which claims nine deaths in 1,836 cases. So all deaths from disease should be tested to see if coronavirus played a role. But clearly, the great unknown that needs to be measured is the prevalence of COVID-19 among the population that is outside the medical system altogether.
Let’s say that the poll tests show that the general infection rate is, in fact, the same 1 percent as seen so far with Congress. That would mean that instead of there being 120,000 infected Americans, there are over 3 million. It would also imply a much lower lethality rate than 1.8 percent, although care must be exercised in drawing conclusions in this regard, since the deaths occurring today need to be compared against not today’s infected population but what its size was in the recent past, as there is a time lag between infections, illness, and death. Once that average time lag is assessed, the true lethality rate can be computed. This is very important, because if the lethality rate really were 1 percent, then, lacking alternatives, it could make sense to allow risky treatments that kill 0.1 percent of patients, whereas this would not hold true if lethality were only 0.01 percent.
If we can increase the nationwide random-testing rate to 5,000 per day (about twelve daily in each congressional district), then once the time lag is assessed, repeated poll testing of the population would provide predictive information about what to expect in the near future, giving advance intelligence to our defenders about how hard the enemy virus will hit us and when and where such blows will strike.
We need that intelligence if we are going to win this war. Random poll testing of the general public should begin without delay.