Will I get sick, will I die? Surprises in the COVID-19 hard data on infections, mortality, as the April 6th witching hour nears

Stunned viewers of CNN’s State of the Union program on Sunday morning heard Dr. Anthony Fauci, director of the National Institute of Allergy and Infectious Diseases say that “projections showed the (COVOD-19) disease would kill between 100,000 and 200,000 Americans and infect far more,” and predicted “millions of cases” while warning that the projections and number are fluid. The director also noted that an end to US economic and social shutdown is going ” to be a matter of weeks,” and depends on when exactly 15-minute COVID-19 tests are available at scale.

Yikes, that’s what they call “a variation in guidance”.

The coronavirus story is not usual subject at the Daily Digest, industrial biotechnology, which is being deployed in support of solutions, such as ethanol plants converting to the production of hand sanitizer: that’s our wheelhouse. As a result, we’ve been inundated with soft data and hard data. You may be interested to learn what the hard data is telling us.

The scary raw data

First of all, let’s look at this chart, which we drew up to illustrate why the heck we all feel so darn scared and confused about coronavirus and COVID-19, even when it comes to basic questions like “Am I going to get it?” and “Am I going to die?” The reports on infection and mortality are all over the map. 

Here, we’ve plotted the known data from Wuhan, the US, Italy, Germany and the UK, courtesy of Johns Hopkins, whose site is being bombarded with a billion visits a day at the moment, we’ve heard.

We can null out some instability in the data. We know that in China, Italy and the UK, testing was restricted to severe cases. Meanwhile, testing in the US has been lagging and doesn’t catch all the flu-like sufferers or any of the cold-like victims. Germany has done a good job of testing those with flu-like symptoms.

We’ve read in the New York Times 80 percent get mild symptoms, like a cold, about 15 percent show flu-like symptoms, and 5 percent generally have the more serious respiratory problems. So, we need to correct the “infected” numbers accordingly, to project where authorities did not test victims and underreported the infections number.

Looking at infection data

By doing so, we can see a much more stable data set for infections. We might speculate that Wuhan’s and Italy’s numbers spiked because the crisis hit early and response was slow.

Meanwhile, let’s add in a couple of other hard data points on infection rates. For one, NBA players were tested, we have the infection rate from the Diamond Princess “cruise ship from hell”, we have data from tests performed on US repatriates from Wuhan, and also there is a Reuters poll that came out, asking people if they had the virus. 

What you see is remarkable. The infection rate is around 2.5-3 percent for quite a number of these cases. We might speculate that the close quarters and slow response on the Diamond Princess made that a data outlier, and we may speculate that Germany, the US and the UK are early in their coronavirus outbreaks and the numbers will eventually inflate.

Would a 2.5 percent infection rate equate to “millions of cases”? Yes it, would. 2.5 percent of the US population is roughly 8.175 million people. But that’s illness.

Looking at fatality data

With our revised infection numbers, let’s put mortality rates under a microscope. 

First, we’re going to note that Italian cases have an average age of 63 (compared to 48 in Germany), and that’s spiking the mortality rate — we think that’s sample bias, clearly older people have higher rates of severe illness and mortality, but there’s no driver to suggest that young Italians are immune. When we look at German data we see Germany’s media age is 47 and the median for COVID-19 victims is 46, so there’s reason to adjust Italian numbers to null out the sample bias towards older patients. 

We find that, once we null out the presumed differences in testing resources & strategies, that the mortality rate stabilizes across different geographies. Italy and China are farther along in disease progression and that may explain the persistently higher mortality.

Bottom line, there’s mounting evidence, though not proof, that 2.5 percent infection rates and 0.3 percent mortality is common with this pandemic. 

For the US, that would project out to 8.2 million cases of COVID-19 and around 25,000 deaths. 

Double checking against previous experience

Let’s double check that range against the known history for epidemics. 

We’ll start with last year’s flu season, which took 34,200 lives based on 35.5 million infections, a 10.8 percent infection rate and a 0.1 percent mortality. In 2017, the rates were 13.7 percent for infection and a 0.13 percent fatality rate. 

Compared to flu, COVD-19 fatality is 3X the recent US experience, while infections appear to be lower. We can look at that through the lens of biology — generally speaking, viruses that are more virulent and dangerous have lower rates of infection — very sick hosts are less able to spread infection through mobility. So, it makes sense that more dangerous epidemics, in leading to societal shutdowns and social distancing, to give an example, are tougher to spread than milder illnesses. 

A visit with Farr’s Law

Another thing we can do to double check our data is to look at the mortality rate for epidemics in terms of time —  and compare the hard data with historical models.

We have a 180 year old Farr’s Law to begin with, which states that epidemics generally rise and fall in mortality with a standard distribution — think Bell curve. When people speak of “flattening the curve”, that’s what they mean, spreading it out over a longer period of time.

In a standard Bell Curve, regardless of flatness, 0.13 percent of the fatalities will be in the first time block, 2.14 percent will be in the second. 13.59 percent in the third and 33.4 percent in the fourth. These are the standard deviations you may have used in school or in your work.

As of today, the US has experienced 2479 identified deaths attributed to COVID-19, according to Johns Hopkins. We have plotted them here, on a day to day basis.

Clearly we’re seeing fast growth. Is this is the earliest stage of a Bell curve, or farther along? Remember, the farther along we are in our progression, the sooner it’s over and the fewer deaths.

To test that, we’ve plotted an normal distribution consistent with Farr’s Law based on a standard deviation of 12.5 days and a projected overall mortality in the United States of 30,000 people. That’s around a 2.5 percent infection rate and a 0.36 percent mortality rate. Here’s that chart.

So far, it’s a pretty good match. But we can’t be sure, yet — this is a plausible scenario, not a proof.  

April 6th – the Witching Hour

The good news is that we have a data coming up, right around April 6th (a week from now), where we can check this curve, because we can expect (sadly) some 4750 deaths to have accumulated in the US, that’s about 360 per day between now and then. If we are in that vicinity, there’s mounting evidence that this could be a 30,000 death season. That’s a terrible toll, but when you look at the scenarios that Dr. Fauci is looking at, you might find yourself uttering a “let’s hope so”.

If the fatalities are much higher, that suggests that we are in a much earlier stage of our progression, and that we are in for much more grief over a longer prior of time. Also, just because other countries have a 0.3 percent mortality rate doesn’t guarantee that result here. Our experience will be different, based on our public and private response. The US is a free-moving and mobile society, we’re behind on testing, there are multiple ports of entry and spreading, all of which are factors that can spike infection numbers.

April 6th will not make this scenario “a sure thing”, but other scenarios begin to become less likely. That model of a 30,000 mortality season is also consistent with a 100-day season of disease, which would end around June 10th with a peak of mortality around April 20th. That’s not dramatically different than the experience in China, where COVID-19 once raged but not if experiencing little in the way of new infections.

Just one day’s difference

To illustrate how sensitive Farr Law’s is, what if the fast expansionary period of mortality continue just one extra day, and we reach 15.86 percent of our total fatalities by April 7th (for math junkies, that’s a period of 13 days between standard deviations, instead of 12.5). In that scenario, the death toll in the United States could reach 58,000. That’s one of the reasons why early April is so important as we aim to halt this epidemic.

Are more dire outcomes possible, probable?

Let’s go back to the Goldman Sachs numbers for a second. What would the numbers look like for a 3 million death season in the US? In that case, we would be just 0.08 percent completed on our march through the Vale of Shadows, and given that we’re 30 days into the cycle, it would suggest a 6-month battle in the US, with the worst of the season hitting around the end of May.

What about the references by Dr. Anthony Fauci to 200,000 deaths, expressed over the weekend? In that scenario, we’d face a 3-1/2 month epidemic, hitting the peak around late April, and something like 60 million Americans would catch the cirus (detected or undetected).

Given the experience of China, there’s not much hard data yet to suggest a 3 million death scenario. 30,000 fatalities is no laughing matter, but many observers have begun to wonder at the wisdom of shutting down the economy to save 30,000 people. They would have multiplied 30,000 by the $10 million “value of a statistical life” , and come up with a cost of $300 billion as the maximum that Americans would have been expected to support.

But they would have forgotten that the lower death rate will have been, actually, because of the response, not in spite of it. And Americans who’ve been generally pretty terrified at the mortality rates being tossed around may well have given a lot more than $300 billion. Not because of the value of the lives lost, but the value of all the lives that felt threatened during this period where data was flying around all over the map, and leadership was hard to find.

What can we do?

The old rules apply. First, wash your hands, a good scrub.  If you develop flu-like symptoms, see a doctor. If there’s a test available, take it. Stay inside whenever possible. If exposed to the virus, self-quarantine for 14 days. 

Simple deeds are powerful. If the United States had insisted in January that air travelers to the US had current documentation of a negative coronavirus test, and tested all cruise passengers disembarking and quarantined those who tested positive, and built up a bank of testing kits early to test anyone with flu-like symptoms, we wouldn’t be in the mess we’re in.  

As Frank Borman once observed of the Apollo 1 pad fire tragedy, there was a “failure of imagination”.