Let me start by saying I’m very hopeful about treatments, which may come any day, reduce all these numbers, and quickly resolve this mess. I hope they do. But since we can’t count on that, we have to make decisions now based on the best numbers we have.
Everyone wants to know what will be the final mortality rate. This is a combination of the attack rate (how many will get it), and of those, the case fatality rate (how many will die).
All of these numbers are in flux but let’s try to tighten our understanding of each:
Attack Rate: The attack rate is doubling every three days. It’s still very low—with only Spain at over 0.25 percent—but it has been doubling unabated in all but a few countries like South Korea (a nation we should be emulating). The current low attack rate is of little solace, however, because in cases where we have had an isolated population, we see much higher numbers. Diamond Princess had an attack rate of 19 percent, the staff at a Washington state nursing home (average age 43.5) had a rate of 29 percent, and a 60-person choir in Washington state had an attack rate of 75 percent. Attack rates in the 20s are common for flu (despite 37 percent of people having been vaccinated each season) and for other coronaviruses like the common cold. We have seen no examples where natural immunity protected people at a better rate. Therefore we must start with 20 percent as a minimum assumption and just hope it doesn’t go to over 50 percent. The nominal case fatality rate (CFR) in the United States, using Worldometer’s numbers, is currently at 2.57 percent. This nominal number is very rough and we have to look at four factors to adjust it. One of these factors will dilute it down but three factors will push it up.
1) Down: Undetected cases, if found, would drive the percentage down. If we really have 10 times the number of cases than we think, our CFR is really only 0.255 percent. But do we? Well, the country with the fewest undetected cases is probably South Korea. They are testing like crazy with over 50 tests per detected case. Plus, they have contact-traced and have almost slowed down new infections so they aren’t hitting a moving target anymore. I would say that undetected cases largely have been baked into their numbers. Their CFR is 1.73 percent and rising (the rising part is explained by the next factors). Their denominator is solid and clearly points to a CFR over 17 times that of the flu.
There has been some talk that the virus has actually already spread undetected widely throughout society and that most people have already gotten over it. This question can and should be resolved over the next few weeks with a new test. The current swab test looks for the antigen (the virus which is present in your nose). A new blood test looks for the antibody (the cure that your immune system creates once it’s beaten the virus and which floats in your blood thereafter). If this conjecture were to prove true, it would be great news—although cases are still rising, the virus would be about to run out of new people to infect. The sick ones we see now would be about all we’d get.
This conjecture likely won’t hold up. Consider the Diamond Princess Cruise. Of 3,700 people, 19 percent got the virus, and eight died (and some are still sick). The conjecture says the virus is already widespread in society, so 19 percent detected on the Diamond Princess is not unusual.
But do similar populations all have 10 pneumonia deaths to match? We’d expect any group of 3,700 Americans with similar ages to have about 0.16 flu or pneumonia deaths per month. Instead the Princess got 10—60 times higher. The Diamond Princess is not typical. It only got 60 times the number of deaths because its 19 percent attack rate was 60 times higher than society’s. Therefore, society may have a long way to go before the virus runs out of victims.
2) Up: Falsely Detected Positive Cases and the “Bayes Theorem.” Many, and perhaps most, of the detected cases don’t actually have the virus and this drives the denominator back down. This is confusing math, but it’s well known among researchers and very common among all tests.
Let’s say you have 1,000 infected people and you mix them with another 100,000 who are not infected. So your universe is 101,000 people.
Let’s say you have a test that is 97 percent accurate. If you test positive, given that the test is 97 percent accurate, you are naturally going to assume there is a 97 percent chance you have the disease. But surprisingly, the odds that you have it are only about 25 percent.
How is that even possible? Here’s how:
- 100,000 are in fact negative: 97,000 test negative and 3,000 (incorrectly) test positive.
- 1,000 are in fact positive: 970 test positive and 30 (incorrectly) test negative.
So even though you only have 1000 infected people, you have about 4,000 positive tests (3,970 to be precise).
Thus if you tested positive you have a 25.2 percent chance (1000/3970 = 25.2) of really being infected. Some studies have indicated that the current test is allowing for about half of the positives to be wrong. So if we have 200,000 positive tests reported on Worldometers, we might have found 100,000 infected people. Of course, this plays off the first factor of untested people and so the number goes up again. So if we have 200,000 positive tests with 4,000 dead, what’s the CFR? Well, if only 100,000 of the 200,000 actually have the virus, then our “real” CFR is 4 percent. But if there are 400,000 infected out in the world (which might require 800,000 positive test results to find), then our real CFR is 1 percent. So these first two factors make the denominator hard to know, but just realize that while undetected cases are often cited, they are offset by false positives which are rarely mentioned.
3) Up: Unresolved Cases. Also known as counting chickens before they’ve hatched. Yes, in the United States, only 2.57 percent of the detected people have died, but another 2.2 percent are in serious condition, and 64 percent, were just diagnosed in the last week! They are hardly out of the woods and many will die—only 4.5 percent are listed as fully recovered. This disease kills 1-2 weeks after being detected. You can’t just hand everyone with a positive test at a drive thru an “I survived corona” t-shirt and dump them into the “didn’t die” bucket.
Worldometers distinguishes between active, serious and resolved cases, but people just want to jump to the deaths/cases number to get a quick and dirty nominal CFR. By this nominal method, South Korea was looking great for a while at 0.6 percent—that is, until they stopped getting many new cases to artificially dilute their CFR. Once the cases they already had were given time, the CFR began, predictably, to inch up. It’s now up to 1.73 percent and rising every day.
The same is true of Germany. They looked to be at 0.3 percent as recently as late March, and a Stanford professor claimed that by finding almost all undetected cases, Germany had revealed the “true” CFR of only 0.3 percent. But it was wishful thinking. Germany’s CFR had been diluted by prematurely counting unresolved cases. In just a week it more than quadrupled to 1.40 percent and is rising so rapidly that I had to adjust it up five times while editing this article. Similarly the United States has drifted up from 1.5 percent last week to 2.57 percent today, despite discovering tens of thousands of undetected cases, which otherwise would have brought those numbers down.
Even as I review these numbers, notice that the 0.1 percent CFR of the flu is less than the daily rounding errors. COVID-19 isn’t just a bad flu.
4) Up: Overwhelm—COVID-19 has one of the highest hospitalization rates, longest hospital stays, and requires very difficult PPE. It’s exhausting and nerve racking for medical personnel to suit up, disinfect, go home to the family and hope you aren’t contagious. Even getting a drink of water or going to the bathroom are big deals. This is a war for the medical profession. On April 1, 1,049 people died of COVID-19 in the United States, making it the third leading cause of death at the moment. Heart disease and cancer are at about 1,700 per day but those are nowhere near as taxing on the staff. Remember that 1,049 died from an attack rate at only about 0.1 percent, not the 20 percent or more we might see soon.
Right now we are within our medical capacities and we are growing those capacities. But if we were to let the attack rate continue repeatedly to double, which it surely would absent the shutdown, we would overrun our capacity. To be conservative, I’m actually downgrading overwhelm as a factor in the United States. Recent data is showing that the ventilators, while lifesavers for some patients, aren’t helping as many as hoped. Perhaps one-third of ventilated patients are saved. So, for every two that die today, a third patient might die for lack of a ventilator—a 50 percent jump for that subset. I think that we will be able to keep up with demand for other treatments.
Where does this leave us?
The nominal (not adjusting for any of the above factors) global CFR is 5.37 percent but is too muddled to be predictive of the United State’s final CFR. Our best guides appear to be South Korea and Germany. Having already extensively tested and contact-traced, newly detected cases are unlikely to bring their CFRs down further than their current 1.73 percent and 1.4 percent respectively. In contrast, the other three factors have been and will continue to drive up their numbers. Remember that Bayes Theorem means that we will likely never see a correct denominator and that those two CFRs would already calculate to well over 2 percent if they could be Bayes-corrected. But ignoring that factor (as everyone else will) and looking only at the nominal method that most will use, we will continue to see the South Korean and German CFRs trend higher towards 2-3 percent, primarily because active cases will become resolved.
So, this argues that if treated in modern hospitals and in manageable numbers a 2-3 percent CFR is the “natural” effect of the disease itself. Deviations from that are the result of how a country handles it in bulk.
If we were to fully reopen the economy, we would experience an unchecked attack rate and this threatens to overwhelm the system. The overwhelm factor is an escalating factor. That is, if you have a natural CFR of 2 percent, the overwhelm might jump that to 2.4 percent, but if the CFR is naturally at 3 percent, the overwhelm might kick in even more and jump the CFR to 4.5 percent.
First World countries such as Italy and Spain will probably see their numbers drift down (if they test) because undetected cases are likely a larger factor than unresolved cases. After weeks of quarantine their new infections are trending down, so further overwhelm may not occur, but some overwhelming damage is already done. Their final CFRs will probably be in the 4 percent range—if they figure out how to prevent new outbreaks, stave off civil unrest, and restart their economies (piece of cake).
Undeveloped countries will likely see the overwhelm factor dominate throughout the pandemic and see CFRs of 5-6 percent or more. We are unlikely ever to know their denominators and will eventually switch from CFRs to overall population mortality numbers and won’t allow us to distinguish the CFR and attack rate separately. Sadly, developing countries might be better off accepting a high attack rate, as any methods used to lower it may trigger a famine and an even higher death rate. There may simply be no solution for those countries other than to hope for treatments. For them it may be 1918.
If we were to lighten up our controls and let the attack rate rise to its natural biological level, we easily could end up with 20 percent or more being infected (65 million), with 2-3 percent CFR, plus an overwhelm factor. So 1.5 million to 2.5 million fatalities. To be honest, I’m still being conservative here because there’s no evidence that the attack rate couldn’t be 50 percent if unchecked. Then you are talking about 4 million to 6 million deaths. In the event of 200 million cases with 2 percent CFR overwhelmed to 3 percent and boom—you are at 6 million. That’s hard to think about but it’s in the realm of possibility. But even 1.5 million is too high for the public to accept, just to save the economy.
This is why there is a shutdown. Letting the attack rate hit its natural biological number without quarantine is just not acceptable to most people. I’m very hopeful that effective treatments can drive the CFR way down, but with no guarantee of that, I’m pointing out what I consider to be the most reliable current math, if we were to immediately resume the economy as normal. This is nothing like a bad flu. I notice a lot of shutdown opponents who are comparing predicted fatalities with the shutdown and saying that the shutdown isn’t worth it. But they are mixing apples and oranges. The 100,000-200,000 fatalities that National Institute of Allergy and Infectious Diseases Director Anthony Fauci is talking about will not hold if we remove the measures.
What I fear the “don’t panic” advocates are missing is that once people start dying, individuals aren’t going to want to go to Las Vegas or to ballgames, anyway. They won’t want to go out to eat or go to shopping malls. Either we shut down with a plan, or we half shut down by individual decision, ruin the economy anyway, but then don’t stop the spread.
To halt the attack rate now, we are using a full shutdown—a very blunt instrument. We might be able to implement a far more precise instrument and get a decent result with only a partial shutdown. The “flatten the curve” strategy actually assumes we arrive at the natural biological attack rate, just slowly, preventing only the overage deaths from overwhelm.
We are in this predicament because we let the horse get out of the barn. Painful as it is, we have to put it back inside by using this shutdown and testing. Then we can look at how we can partially operate while keeping R0 (the replication number) below it’s “chain reaction” number. That is, the attack rate will grow, but it will not be allowed to engage in regular doublings. We have to do that while we wait for a vaccine or treatments.
How can this be done? South Korea and Japan are already succeeding. What to do next will be the subject of a future piece.
Stay safe and healthy.