Friday, April 24, 2020

COVID-19: No, it is not time to "reopen the economy".

In response to this article: This article is pretty bad - it's full of really misleading misinterpretations of statistics, either through malice or ignorance.

The only path forward is through testing and contact tracing/isolation. Trying to isolate only the vulnerable population won't work.

Here it is bit-by-bit:

Five key facts are being ignored by those calling for continuing the near-total lockdown.
No, these key facts are being misinterpreted by those calling for "opening the economy".
Fact 1: The overwhelming majority of people do not have any significant risk of dying from COVID-19.
Everybody already knows this. The problem isn't that we are worried about the overwhelming majority of people dying, the problem is a bunch of people dying in a short time period overwhelming the health care system, and driving up the fatality rate. Plenty of people need hospitalization to survive, and if they don't get it they will die.
The recent Stanford University antibody study now estimates that the fatality rate if infected is likely 0.1 to 0.2 percent, a risk far lower than previous World Health Organization estimates that were 20 to 30 times higher and that motivated isolation policies.
Here's the link to the study.

The population consisted of facebook users who saw an ad, so if it was biased because, say, people who had an illness a month ago wanted to get the test, it would inflate the denominator and collapse the CFR. Says it right in the report.

From the Stanford report:
"This study had several limitations. First, our sampling strategy selected for members of Santa Clara County with access to Facebook and a car to attend drive-through testing sites. … Other biases, such as bias favoring individuals in good health capable of attending our testing sites, or bias favoring those with prior COVID-like illnesses seeking antibody confirmation are also possible. The overall effect of such biases is hard to ascertain."

The CFR reported by this report is literally impossible given the situation in New York City. As of today (April 23), NYC is reporting 10,000 confirmed COVID-19 deaths, and another 5,000 "probable". If you take the 15,000 deaths in NYC, and divide it by the reported CFR from the report, you get a range of infected people in NYC from 7.5 to 15 million people.

The total population of NYC is 8.4 million. If 90 or more percent of people in NYC had been infected, the rates of transmission would have slowed down long ago, so it cannot possibly be true that the CFR is 0.1 to 0.2%.

In New York City, an epicenter of the pandemic with more than one-third of all U.S. deaths, the rate of death for people 18 to 45 years old is 0.01 percent, or 11 per 100,000 in the population. On the other hand, people aged 75 and over have a death rate 80 times that. For people under 18 years old, the rate of death is zero per 100,000.
This is the rate of death across the entire population, not across the population infected. What fraction of NYC has been infected? Here's one possible study.

"The results differed across the state with the largest concentration of positive antibody tests found in New York City at 21.2%."

If that's the real fraction of NYC residents that have been infected with COVID-19, then you should multiply those death rates from above by 5 to get the real risk of death, so for 18 to 45 year olds you'd be at 0.05%, a 1 in 2000 chance of dying.

And, if that test is subject to similar selection bias or false positive results as the Stanford test, it could be EVEN WORSE.

Of all fatal cases in New York state, two-thirds were in patients over 70 years of age; more than 95 percent were over 50 years of age; and about 90 percent of all fatal cases had an underlying illness.
In 4 months I'll be over 50. Both of my brothers and my parents are over 50. Lots of people I care about are over 50. If you are talking about isolating the entire population over 50 that's a huge fraction of the population to isolate, and it's a huge fraction of the population you want to remove from isolation.
Of 6,570 confirmed COVID-19 deaths fully investigated for underlying conditions to date, 6,520, or 99.2 percent, had an underlying illness. If you do not already have an underlying chronic condition, your chances of dying are small, regardless of age.
This is actually a pretty reasonable statement, except for one big caveat. Here's the original data referred to here.

This table shows that out of 8811 deaths, 6250 had underlying conditions, 2241 had "underlying conditions unknown", and only 50 had "no underlying conditions". I do not know where the authors of this article drew the conclusion that unknown underlying conditions hadn't been fully investigated, but they use that fig leaf to throw out a large fraction of the data, in favor of their preselected view that the virus is No Big Deal.

The real fraction, 6250/8811, is 71% of deaths that had an underlying condition.

The best data I could find suggest that about 21% of adults under age 65 have an underlying condition, and including older adults it puts the total fraction of the population at risk at 38%.

If you take the NYC table and look at adults under age 65 you get 85% of deaths that have underlying conditions, where random chance would get you 21% of deaths with those underlying conditions. I played around with this in a spreadsheet, and it looks like this means that those with underlying conditions are about 20 times more likely to die than those without or with "unknown" conditions, under age 65.

So for those under age age 65, if you do not have an underlying condition, your chance of dying is about 1/20th that of someone who does have an underlying condition.

However, here's the big caveat: If you have an underlying condition and don't know it, you are at risk.

If we "reopen the economy" will we give doctor's notes to those who are vulnerable? If we don't, will they get laid off, or will they risk their health to afford groceries? Will they get relief payments or will those stop because the economy is "reopened"?

And young adults and children in normal health have almost no risk of any serious illness from COVID-19.
This, however, doesn't follow. Risk of deaths are not the same as risk of serious illness, see the next section for how bad this gets.
Fact 2: Protecting older, at-risk people eliminates hospital overcrowding.

We can learn about hospital utilization from data from New York City, the hotbed of COVID-19 with more than 34,600 hospitalizations to date. For those under 18 years of age, hospitalization from the virus is 0.01 percent per 100,000 people; for those 18 to 44 years old, hospitalization is 0.1 percent per 100,000. Even for people ages 65 to 74, only 1.7 percent were hospitalized. Of 4,103 confirmed COVID-19 patients with symptoms bad enough to seek medical care, Dr. Leora Horwitz of NYU Medical Center concluded "age is far and away the strongest risk factor for hospitalization." Even early WHO reports noted that 80 percent of all cases were mild, and more recent studies show a far more widespread rate of infection and lower rate of serious illness. Half of all people testing positive for infection have no symptoms at all. The vast majority of younger, otherwise healthy people do not need significant medical care if they catch this infection.

This is pretty bad. What do you think they mean when they say "For those under 18 years of age, hospitalization from the virus is 0.01 percent per 100,000 people; for those 18 to 44 years old, hospitalization is 0.1 percent per 100,000." ?

What is a percent per 100,000?

Never mind, they link to the data so we can reverse engineer this:

The rate of hospitalization for under 18 is 14 per 100,000, which is just .014%.
The rate of hospitlization for 18 to 44 is 154 per 100,000, which is just .154%

That sounds great! Except… what do you think the denominator is? This paragraph is worded to imply this is per 100,000 cases, but it's not. It's per 100,000 residents of NYC.

In order to determine your actual risk from getting the virus, you need to know what fraction of NYC has been infected. See above for evidence that this is AT MAXIMUM 21%, and now you know you have to multiply these numbers by at least 5. Adults under 44 have about 0.75% of being hospitalized if they get this disease, according to those rates. And if NYC residents have been exposed LESS, this number gets even worse. Children have a 2 in 3000 chance of being hospitalized if they get infected, or worse if the 21% exposure rate in NYC is off.

Now I need you to think about this - if the real rate of hospitalization from getting this virus among 18-44 year olds is 0.75%, this means that if 100% of those people were infected you'd have 7.5 hospitalizations per 1000 people. The US has less than 3 hospital beds per 1000 people, so if 40% of the population were sick at the same time it would require more hospital beds than we have. If half the hospital beds are already full from people with other problems, it would only take 20% infection to fill every hospital bed.

If the hospital beds fill up, those who "need hospitalization" now have a much higher death rate.

Fact 3: Vital population immunity is prevented by total isolation policies, prolonging the problem.

We know from decades of medical science that infection itself allows people to generate an immune response — antibodies — so that the infection is controlled throughout the population by “herd immunity.” Indeed, that is the main purpose of widespread immunization in other viral diseases — to assist with population immunity. In this virus, we know that medical care is not even necessary for the vast majority of people who are infected. It is so mild that half of infected people are asymptomatic, shown in early data from the Diamond Princess ship, and then in Iceland and Italy. That has been falsely portrayed as a problem requiring mass isolation. In fact, infected people without severe illness are the immediately available vehicle for establishing widespread immunity. By transmitting the virus to others in the low-risk group who then generate antibodies, they block the network of pathways toward the most vulnerable people, ultimately ending the threat. Extending whole-population isolation would directly prevent that widespread immunity from developing.

It is usually true that once enough people have been infected there will be immunity and transmission rates will slow. However, this is a new virus and we don't have a lot of information about it - does our immunity last for a month or a year or 10 years? Does the virus mutate like a flu virus to the point where it can reinfect someone who previously had it?

Even if we had confidence that naturally acquired immunity is present and lasts long enough to be effective, we still are advocating sending many people to the hospital or to the morgue to acheive this end.The argument that this prolongs the situation is based on many assumptions that aren't known to be true - will people self isolate anyway, and prolong the spread voluntarily? Will there be an antiviral therapy, or a vaccine introduced?

It is still quite possible we could reduce the area under the curve significantly even just by flattening the curve as we are doing now.

And it may never be necessary to gain natural herd immunity if we can apply testing and contact tracing to reduce cases until a vaccine or therapy is available, or even if we can eliminate it entirely through contact tracing. See for example South Korea.

Fact 4: People are dying because other medical care is not getting done due to hypothetical projections. Critical health care for millions of Americans is being ignored and people are dying to accommodate “potential” COVID-19 patients and for fear of spreading the disease. Most states and many hospitals abruptly stopped “nonessential” procedures and surgery. That prevented diagnoses of life-threatening diseases, like cancer screening, biopsies of tumors now undiscovered and potentially deadly brain aneurysms. Treatments, including emergency care, for the most serious illnesses were also missed. Cancer patients deferred chemotherapy. An estimated 80 percent of brain surgery cases were skipped. Acute stroke and heart attack patients missed their only chances for treatment, some dying and many now facing permanent disability.
All of these are arguments for reducing the total time under isolation, but that doesn't mean that opening up now is the right answer - if we cases explode we'll be back in isolation anyway, either through government action or just through people choosing to self isolate in order to avoid perceived danger.

Where is the data on how many people are impacted by this vs. the known death toll of COVID-19? Where is the data on how many people are impacted by this vs. the hospitalization toll of COVID-19?

Fact 5: We have a clearly defined population at risk who can be protected with targeted measures. The overwhelming evidence all over the world consistently shows that a clearly defined group — older people and others with underlying conditions — is more likely to have a serious illness requiring hospitalization and more likely to die from COVID-19. Knowing that, it is a commonsense, achievable goal to target isolation policy to that group, including strictly monitoring those who interact with them. Nursing home residents, the highest risk, should be the most straightforward to systematically protect from infected people, given that they already live in confined places with highly restricted entry.
If it's so easy to protect people in nursing homes, why are so many of them dying? We've locked down society in general, and we've instituted restrictions on nursing homes, and they are still dying in large numbers.
The appropriate policy, based on fundamental biology and the evidence already in hand, is to institute a more focused strategy like some outlined in the first place: Strictly protect the known vulnerable, self-isolate the mildly sick and open most workplaces and small businesses with some prudent large-group precautions.
Dividing the population into groups like this doesn't work. If you have one population with an exponential growth rate, and one population with exponential decay, the population with exponential growth will overwhelm the one with exponential decay. The people you are trying to protect will be facing a boundary with the outside world that is far more dangerous.
This would allow the essential socializing to generate immunity among those with minimal risk of serious consequence, while saving lives, preventing overcrowding of hospitals and limiting the enormous harms compounded by continued total isolation. Let’s stop underemphasizing empirical evidence while instead doubling down on hypothetical models. Facts matter.
Facts do matter, but "hypothetical models" are how you make rational decisions. This entire article is an attempt to build a hypothetical model of how you can open up society for a fraction of the population. I guess we can stop using hypothetical models soon, as Georgia looks like it's going to try to open up. In 3-6 weeks, we'll see how that works out.

From where I stand, the only solution that makes any sense is to have wide spread testing, both in the surveillance sense as well as in the contact tracing sense. Until we do, we're going to be in a world of hurt.