What we knew we should have done about Covid, and we didn't

Why did it take so long to figure out whether to wear masks?

3 months into lockdown, and more into the pandemic, the front page article in the San Jose Merc described 8 things we got wrong at first Some of the list is right, and a few of them are dangerously wrong, but I want to focus on something else, namely the things we got wrong even though we (or at least many of us) knew better. Things we have little excuse for getting wrong, particularly because others got them right.

Not putting immense effort into testing

The most obvious and deadly error was not putting all effort into producing tests and distributing them as widely as possible as quickly as possible. Other countries cut off their outbreaks quickly doing this. Some of the things that went wrong were:

  • At first, all testing had to be done at the CDC. Almost no tests were done.
  • Even after that ended, tests were rare enough that only people who met a very high bar -- contact with people from Wuhan, strong symptoms got tests.
  • The lack of testing caused both under-reporting of extant cases, and a poor estimate of the case fatality rate.
  • Even later, as better tests were developed that did not require an unpleasant nasal swab, such as a saliva test, the FDA got in the way of their deployment.

Not putting immense effort into study of the means of transmission

We've learned things about the means of transmission very slowly. We're even still uncertain about a few of these:

  • What fraction of cases never show symptoms and can they transmit it?
  • When are presymptomatic people infectious?
  • What is transmission risk of various outdoor behaviours vs. indoor ones?
  • How much do masks reduce transmission, and which types of mask?
  • What are the transmission rates for housemates, strangers and near contacts?
  • How much transmission is via superspreaders? (Some research suggest it's a lot)
  • What are the risks for indoor office work or factory work? Protests marches? Outdoor dining? Indoor dining? Eating food cooked by an infected chef or served by an infected waiter?
  • How much transmission is there from surfaces and what surfaces?
  • Do children (who are less affected) also transmit it less?
  • Do seniors (who are more affected) also transmit it more?
  • What are the specific comorbidities? What drugs and supplements reduce acquiring infection? (There is good research on which ones reduce case severity but could be more.) Do any substances reduce transmission?

We're slowly learning some of these things, but slowly. At first, before we knew much at all, strong lockdown was our only clear path to be sure we were going to reverse the exponential growth of the virus. But lockdowns are very expensive, and once you know more about transmission, you can start relaxing them, relaxing the things you learn are lower risk. Which means every day sooner we learn those facts is a day sooner you can open those activities and the result can be billions of dollars of economic activity.

If we knew these things we could make scientific decisions about what to do about protest marches or outdoor dining. We could know what types of offices can open. We would know what can open if you wear a mask, but not open if you don't.

Which means there should have been as much money as needed, and more. to research these questions around the world. And we knew we wanted it, and some researchers went to work on it, but they were never given the tools they need to do it fast. Tools would include:

  • Huge budgets and access to researchers
  • Laws to compel disclosure of certain types of useful data
  • Contact-tracing apps, not just to track possible infected parties, but to be able to know what types of contact resulted in what rates of transmission. These apps could have been given to volunteers, with minimal privacy concerns. We need privacy protection in general contract tracing apps, and we've been spending a lot of time trying to get that right -- but we don't need to do that for a volunteer data gathering app.
  • Global efforts to quickly study the different effects of all the many lockdown rules -- and de facto lockdown practices and non-observances -- in all the cities of the world. Without planning to, the cities have created a giant laboratory of experiments which we failed to exploit at anything but a slow scale.

The key to a lockdown is to get the virus spread rate (the base of the exponential equation ) known as R0, down well below 1. For example, if you get it to 0.5, and the virus runs that in 5 days, then in 25 days you can get the virus down to 1/30th the size, after which you can inch the R0 up towards just below one, where it is present but not growing. Or you can keep up the strong lockdown and kill it even more, but either way, you find the fastest path to re-opening. You can't let it get much above 1.0 or it starts growing again. At least until you have a treatment or vaccine.

Research with willing test subjects

Normally, researchers are trained not to do research which puts experimental subjects at risk. However, it is OK to study what happens to people who voluntarily place themselves at risk.

Pretty early on it was clear that there was a large population who did not want to lock down. They eventually came to defying the lockdown, even protesting in the streets for the right to defy it. There is no question they wish to expose themselves to risk. The problem, of course, is that the lockdown is not to protect them, it's to stop them from being vectors who breed and transmit the disease to the vulnerable.

An opportunity was missed to make use of this enthusiastic consent. A lockdown is still workable when some people are given an exception to it. You just can't give the exception to everybody or you get exponential growth again. The lockdown rules always gave an exception to the so called "essential workers" who provided necessary services. It would also be reasonable to not lock down a group of people who demand to go about normal lives and constantly wear data gathering devices and be subject to regular testing. While it is not ethical to do an experiment which pushes people to take risks, it is very acceptable to ask subjects to wear recording devices and yield some privacy in exchange for a reward. The reward being to be treated like an essential worker. (Actual essential workers could also consent to this data tracking.)

To be clear, encouraging people to take risks for science would not be ethical. Asking them to collect data when they take the risks they keenly want to take is not. It is the lockdown law which holds them back from what they want, not being outside the test group.

There would have to be some caveats on participation:

  • All members of the household would need to consent to a person's participation (if not an essential worker already,) particularly older members. With consent, the entire household would probably become subjects together.
  • Companies which wish to re-open in this fashion would need full consent of any employees joining them, and we need some way to assure that the pressure of "not getting laid off" is not used to push employees into taking risks they don't want to. For low-skilled jobs, it will be necessary to use other volunteers who don't work at an open business and train them. This works fine for things like restaurants, shops etc. but not for specialized work. However, it is not necessary to study every type of work.
  • Typically, only a small subset of any type of business would open, because only a small set of customers will be able to come. If 10,000 people are study subjects, you can only open enough restaurants to serve those 10,000 under normal circumstances.
  • Businesses which deal with the public would have to verify that all customers are in the program, probably using the tracking app they are required to carry at all times.
  • At-risk people and their housemates would not qualify. This would reduce information on whether reception of infection is different for the at-risk, that must be learned another way.
  • Chances are this would only be done in a few towns, the towns where the volunteers are most insistent at volunteering. It would also be done in different countries, particularly places with differing lockdown rules.
  • All volunteers would be given extensive education on the risks to assure consent is informed as well as enthusiastic.

With tracking devices (bluetooth, GPS log and body cam) almost every person to person transmission would be detected. It would be harder to detect transmission via surfaces or ambient air, but the body cam would reveal what the subject touched (outside their home) during the likely time of infection.

Such precise data on all transmissions would give us much of the data we need to fine tune the lockdowns for maximum virus prevention at minimal economic cost. It would save huge numbers of lives, and untold billions of dollars. It's hard to think of anything with bigger payback.

As noted, all subjects would get regular testing. If infection were found, they would naturally go into isolation. There would be superb contact tracing both to find the source of their infection, and to warn and isolate any contacts during their infectious period. With saliva based testing and group testing, testing could be done twice daily to result in minimal transmission. That's actually not ideal from a research perspective -- this group will be much safer than the public due to this testing -- but it is of course necessary from an ethical perspective.


We actually had reasonable data that masks are effective early on. Masks are simple and cause no economic hardship, so it is unclear why they were not required early -- other than of course, that there were shortages of them due to other mistakes.

Not clearing the senior homes

It was immediately clear back in January that older people were at the most risk. It was obvious that the fastest path to saving lives would have been to get people out of the senior homes if we could, or to improve procedures in them. A large fraction of the 110,000 deaths so far in the USA came to seniors, and we've spent untold billions trying to reduce that total -- but we could have reduced it a lot with much less than billions.

Not already being practiced at all of this.

We've known some pandemic was coming for decades. Bill Gates and many others warned of it and didn't listen. I warned about how tech like videoconferencing could make it more tolerable in 2005 but nobody prepared. We could have prepared, and been ready to reduce the economic impact a lot.

Not realizing the public just won't get non-linear thinking

Sadly, too many of our intuitions are linear. People looked at low, but exponentially growing case numbers and figured there was little to worry about.

What we should have known

There are some things that were less clear and perhaps it is hindsight that makes us realize we should have known them

The public will only tolerate lockdown for so long

It was not clear when lockdowns began that there are things which, even if they are the right medical choice, the public, and even the government officials, will not do. Fatigue sets in, and they will start taking reckless risks with their own lives, and the lives of others.

This may mean that it's better to do a hard lockdown for a shorter period, to reduce fatigue, than to have a lesser lockdown for a longer period, even if the former choice is worse for other reasons. The long lockdown is simply not on the table short of a police state. At least in the USA.

It will become a political issue

Fatigue over lockdowns is universal, but it and many other factors became political issues, and people's views on scientific and medical questions have now become driven by their political biases. Those biases are also affecting what research they read and believe, so they (I) feel that they are making all their conclusions only through rational scientific analysis. But nobody is.

It became a tool of international conflict

Just as Russia interfered with U.S. elections in its efforts to disrupt its rival, it is likely they are also interfering via social media with the virus situation, to increase the polarization and politicization and drive people towards more chaos and more wrong decisions. Even though we were shocked when it happened in elections and vowed never to be surprised again... we seem to have been surprised again.


Overestimated the hospitalization and fatality rate, underestimated the infection rate, told people not to wear masks, imposed lockdowns that were unsustainable, closed colleges, encouraged unemployment with a $600/week bonus, imposed price controls and other regulations that have caused massive shortages (especially N95 masks), thrown massive amounts of money at Wall Street causing an enormous bubble, wasted enormous amounts of time and money on ventilators and hospital beds that weren't needed (largely because of the first reason).

Maybe I'll think of more later.

A few of those are in the Mercury article, but I didn't list some of these because you expect those sorts of errors. I did not cover economic and stimulus decisions. My focus is on decisions on mitigating the virus (and the lockdown) which we actually had enough information on at the start to do something about, but didn't. Of course errors were made in studying these rates.

However, there was not an error in the fatality rate as far as I can see, because we didn't know the fatality rate very well at all then. Back then we only knew case fatality rates which were really fatality rates among those who qualified to get a test. Since whether you got a test was so unpredictable, this was not very good data. An error would be somebody who said they knew the IFR back then.

Now, we seem to have a converging handle on the IFR, around 1%.

The masks I thought I listed. While we did not have solid data on mask effectiveness then, we had a lot of clues that they would be a wise idea.

My list was meant to be independent of yours, not a list of things you missed.

The fatality rate could have been extrapolated from the fatality rate on the cruise ships, adjusted for the ages of the cruise ship occupants (it was an isolated population that was 100% tested).

I did that back in late March and came up with about 1%.

The number of actual cases could then be backed into from there, using the growth rate, the number of deaths, the days from infection to death, the fatality rate, etc. There was even an article I remember seeing doing the calculation.

By then it was already too late to get much out of increased testing, though. Testing was fairly useless as a means of stopping the spread by the time we had community spread.

I'm not sure you can separate economic decisions from virus mitigation. Interfering with the economy causes deadweight losses. Maybe the Wall Street bubble didn't have much effect on virus propagation (the price for that may not be paid for years), but the other issues I pointed out all did. (I'm not even sure which ones you're saying are economic.)

We had good data that masks probably worked. The CDC and WHO intentionally lied about it in an effort to stop shortages. Medical professionals lied about masks so they could hoard them for themselves. https://www.nytimes.com/2020/03/17/opinion/coronavirus-face-masks.html

The cruise ship data was useful, but the sample sizes were very small. In addition, on one hand the population skews older, but on the other hand, the seniors doing Asian cruises are ambulatory seniors of above average fitness. But the ships put some constraints on things. However, very soon there was Lombardy, and of course there was Wuhan, though data from it could not be trusted.

That's why for me the most useful number has been excess deaths. There's no faking that, not even in China. There's no interpretation. There is of course the issue of some other hidden source of excess death, but the numbers got so large as to preclude most chance of that. Testing has always been too erratic.

Next useful number is antibody presence done in a well sampled population group with one of the better tests. (Or eventually with two tests, one highly sensative, the other highly specific.) False positive rates are too low for those in areas with low prevalence, but over 15% the results are reasonable.

As long as the uncertainty bars on the virus included catastrophic death, it was necessary to consider primarily how to avoid that and not the economic consequences. Once your narrow those bars -- or learn how to fine tune the techniques you use because you have more data on transmission -- you can start putting in economic factors.

I agree that we had data that masks were good early on, and for some reason did not act on it. California only required them statewide today.

Sample sizes don't have to be very large. I think it was clear from the cruise ships that 3% was way too high and that we were missing a lot of cases. The antibody tests confirmed this, but they came much much later. When they did come, I wasn't surprised by that they showed that cases was much higher than so many people suspected.

I don't think excess deaths tells you much once we imposed lockdowns. There are too many confounding factors. Lives were saved, and lives were lost, for reasons that have nothing to do with the virus except that our normal routine was disrupted.

Again I don't think you can separate the economic consequences from everything else, and I'm not sure what (other than the stock market) you're even referring to as purely economic.

Are you saying that the $600 extra for unemployment didn't affect the number of people infected. Surely it did. More people stayed home because of it, and most people who were infected were people who stayed home. The massive unemployment, the closing of schools and colleges, the lockdowns of businesses, the economic devastation, it all led to people coalescing in their homes, often with multiple households living together, and this helped spread the virus quite a bit. And the lockdowns weren't sustainable, so now people are forced to go into risky situations just to start to earn back what they lost from the disruption, and to try to get their sanity back. Cases are going up exponentially in many places now. Even if the lockdowns worked for a while, it's looking like the virus is going to still run its natural course over the next 6 months. If we had a large portion of under-35 individuals recovered and immune that would slow the spread much more effectively than trying to lock down everyone indefinitely, and the death rate under 35 is extremely low. From NYC antibody tests it looks like as low as a 25% immunity rate might be enough to get the R0 for unavoidable activities well below 1.

Maybe you were talking about the "enormous amounts of time and money on ventilators and hospital beds that weren't needed"? The point of that is that the time and money we spent on that could have been spent on something more useful. More masks. More treatments. More studies of exactly how this thing spreads.

Mask laws don't work. You have to convince people to wear them. You have to make people feel as socially obligated to wear them as they are socially obligated to wear pants. The biggest unnecessary problem with the masks is that the CDC and our president lied to us about them, and now a large portion of the population is confused about them.

Excess deaths are the best thing we have, and they are much more useful than you suggest. Sure, there are some excess deaths caused by the lockdowns -- but a tiny, tiny, tiny portion of those caused by Covid unless you have some strong evidence to the contrary, and in fact for every argument I have seen about suicides or other problems there is a stronger argument that other deaths will be reduced by lockdown even more.

More to the point, they match. in NYC, we see about 24K excess deaths and 17,000 confirmed covid deaths, and a lot of people who died at home or in nursing homes with covid like symptoms but untested and thus not confirmed from the early days.

Deaths can't be hidden, nor greatly miscounted.

Now I make no comment here on the bail-outs, the unemployment benefits.

I fail to see how having a body of under 35s who are immune slows the virus more than having those same people sheltering at home slows the virus.

Mask laws only add a little compared to social pressure, but they start social pressure. One mistake was people thought of masks as protection for themselves, and thus their choice, their freedom to decide. If it were more clear they were to protect others, it would be less common to think it is about your freedom to decide. It is more clear that if you're not wearing one, you're a dangerous prick, not a bold risk taker.

Excess deaths are the best thing we have for what?

They're not the best thing we have for measuring the number of people who die as a direct result of COVID-19.

They're maybe a way to estimate how many more people are dead compared to a hypothetical world where COVID-19 had never existed. But that's a much much different thing, and I'm not sure what the point is of even measuring that.

Sure, there are some excess deaths caused by the lockdowns -- but a tiny, tiny, tiny portion of those caused by Covid unless you have some strong evidence to the contrary

How do you define "deaths caused by lockdowns" and "deaths caused by COVID"? By one definition, all deaths caused by lockdowns are also deaths caused by COVID, since COVID was a but-for cause of the lockdowns.

So I'm not sure what it is you're even trying to measure, let alone why.

George Floyd was unemployed because of the lockdowns. Would he be alive today if there weren't any lockdowns? Probably. So do we count him as a death caused by the lockdowns? Do we count him as a death caused by COVID-19?

I fail to see how having a body of under 35s who are immune slows the virus more than having those same people sheltering at home slows the virus.

Sheltering in place perfectly for a year to year and a half, sure. But that's not going to happen. That's not realistic. People aren't going to adhere to that. It isn't even the law. There are always exceptions.

When I say that lockdowns don't work, I don't mean that lockdowns wouldn't work if everyone followed them perfectly. What I mean is that they don't work in reality. They cause lots and lots of suffering, and the virus just keeps going anyway. The virus can withstand the lockdowns longer than we can.

I don't know, maybe you think we're going to lock down to the levels we were at a month ago, now that the virus is growing exponentially again. I doubt it.

If the virus is going to keep going until we have herd immunity, and we aren't willing to lock down enough to get a vaccine before we hit natural herd immunity, then we should reach that natural herd immunity with the lowest risk population.

It's a big if, and just a few weeks ago I though we were going to be able to keep the virus contained through the time when we get a vaccine. I don't think so any more. Neither does the IHME.

Mask laws only add a little compared to social pressure, but they start social pressure.

Definitely not. The social pressure came before the laws in every jurisdiction I'm aware of. In fact, in most jurisdictions there are no mask laws, but there is quite a bit of social pressure (and mandates from business owners).

Mask laws that aren't supported by the vast majority of the population won't be enforceable. What are you going to call the police every time someone isn't wearing a mask? It's not going to work, especially in this anti-police environment.

If it were more clear they were to protect others, it would be less common to think it is about your freedom to decide. It is more clear that if you're not wearing one, you're a dangerous prick, not a bold risk taker.

I'm not sure that's true. I think most of the people not wearing them are dangerous pricks.

I'm very much in favor of people wearing masks, but it's primarily business owners that are going to need to enforce this, not the police. With some things you can have the police enforce a rule requiring the business owners to enforce something (for instance drinking age). Even with things that somewhat works for (like drinking age), a lot of lack of compliance slips through. Especially in lower income, higher crime, areas, which is where the COVID problem is worst. Underage drinking is rampant in these places, despite there being laws against it. But I don't think mask enforcement can even reach that level. Not in places where people aren't currently wearing masks.

By the way, I don't oppose laws requiring people to wear masks in places open to the public. I think such laws are justifiable (or at least acceptable). I just think they're useless.

Lockdowns, on the other hand, I don't think are justifiable or acceptable.

We want to learn the IFR. The IFR lets us calculate the cost, in lives, of allowing the virus to spread. Of attempting the path to herd immunity. Excess deaths are the only number with low error. The only error in them is the fraction that are not Covid deaths, and while that error is present, it is probably not much more than 20%. On the other hand, estimates of the CFR, due to testing sampling error, have varied by hundreds of percents. The error on the excess death numbers is low enough that we can make decisions on it.

A Covid death is a death from the disease. You can argue the fine points of people who were going to die in 2 months and instead died today, but that's again what is good about excess deaths -- it accounts for that, over time.

I am sure you can list 100, even 1,000 anecdotes about deaths which can be loosely attributed to the lockdowns, even a few which are surely attributed to them. You can't list 100,000 of them. You can't even come close, I predict. Millions are unemployed, only an almost undetectable fraction are dead because of it. However, when a million are sick with the virus, we now know roughly 10,000 to 14,000 will end up dead.

We aren't going to shelter in place for a year. If we do this right, we will figure what things carry the most risk of transmission. We may have to stop those for a year, but can allow others to continue, after the virus has been nearly wiped out by stronger lockdowns.

Lockdowns don't have to be followed perfectly -- in fact inherently we allow the "essential" workers to bypass them. The lockdowns are strong to allow that margin. No one person's activities are on their own very dangerous, but they multiply. Since you can't let everybody do risky activity, you must constrain who does so, and do it in a fair and just way. The way currently chosen is to allow the essential jobs to not be locked down, because that's useful.

The goal is not to get to herd immunity. The current lockdown strategy is not the right strategy for that. Bergamo got to herd immunity, the very, very hard way. We don't want that.

Yes, business owners should enforce the mask rules. Best to avoid police, unless somebody is a dick and won't obey the owner of a business they are entering.

The difference between the IFR and the CFR is the denominator, so I'm not sure why you're even bringing it up.

To calculate the IFR or the CFR, we need to know the number of deaths directly due to the disease. Not the number of deaths due to people being too afraid to go to the doctor. Not the deaths from people who can't get their prescriptions filled, or who don't exercise as much, or who eat too much junk food, or due to any of the other indirect results of the pandemic, nor do we want to count deaths completely unrelated to COVID-19. We also don't want to subtract all the deaths that aren't occurring because of the pandemic. We don't want to undercount deaths from the virus because deaths from work-related stress are down, or deaths from car crashes are down.

You say the error from these factors is probably not much more than 20%. I think that's a ridiculous assertion without confining it to a particular geographic location. But even if you confine it to a particular location (say, NYC), the only reason that's more than a guess is because you've looked at things other than excess deaths.

I don't see how excess deaths helps you get the IFR, except that it can point you to the possibility that you're undercounting. If excess deaths are a lot more than deaths attributed to COVID-19, then you need to look more closely at them and come up with an explanation for them. That explanation might be that you're undercounting COVID deaths, or it might be something else.

We aren't going to shelter in place for a year.

We certainly aren't. So we need to come up with something that can be sustained for a year.

If we do this right, we will figure what things carry the most risk of transmission. We may have to stop those for a year, but can allow others to continue, after the virus has been nearly wiped out by stronger lockdowns.

We tried that. For three months. It didn't work. The virus isn't nearly wiped out. In fact it's more prevalent now than it was three months ago.

Time for a new plan.

Lockdowns don't have to be followed perfectly -- in fact inherently we allow the "essential" workers to bypass them.

Lockdowns have to be followed nearly perfectly if you want to nearly wipe out the virus. For this virus, at this point in time, anyway.

The goal is not to get to herd immunity.

Then the goal is not to stop the virus.

The goal has to be to get to herd immunity. The idea that this virus is just going to disappear some other way is ridiculous. Herd immunity, whether it be natural, via a vaccine, or some combination of the two, is our best hope.

Hopefully we can do that with as little pain and death as possible.

The current lockdown strategy is not the right strategy for that.

Which lockdown strategy? There are hundreds of them (counting states, counties, and cities).

Bergamo got to herd immunity, the very, very hard way. We don't want that.

No, we don't want that, and we're not seeing anything like that. Even in NYC, which probably experienced the worst that we'll see in the USA, hospital resources weren't completely overwhelmed.

Yes, business owners should enforce the mask rules. Best to avoid police, unless somebody is a dick and won't obey the owner of a business they are entering.

Well over 20% of the population consists of dicks. You can't call the police on them all.

We can't calculate the IFR exactly. CFR is not that interesting from the public policy standpoint, other than as a path to IFR. You make a calculation with error bars. So you want inputs with the smallest potential error. And right now, excess deaths has an order of magnitude smaller error than anything else. It's not even a contest.

I assert the low error because we see that excess deaths is within range of confirmed Covid deaths, and we know there are many Covid deaths that are not confirmed as Covid. For a long time we didn't have much idea at all about Covid confirmation because testing was so variable from place to place.

I understand the various sources of error you cite. The case rate is seductive. It is of course precise, but completely inaccurate. Excess deaths is not precise, but is vastly more accurate.

Many countries "tried that" and got success. Some wiped it out entirely. Yes, the USA seems to suck at managing this -- in large part I fear because the process has become politicised.

Lockdowns don't have to be followed anywhere near perfectly, and not just because we allow the essential work and travel to go on. The goal of any anti-virus strategy is to reduce the virus' multiplication number (R0 or whatever you prefer) below 1. At 1.0, the virus is stable -- the number of cases remains roughly constant. Below 1.0 it begins exponential decay. (Above 1.0 it has exponential growth.) Get it to 0.5 and do that for 10 generations of the virus (about 2 months) and you have the virus down to 1/1000th of what it was. More generations and you can reduce it all the way to extinction, but that's expensive. Once it is rare, you can then open up and bring R0 up to 0.99 and it will continue in tiny pockets as you hunt for vaccines and treatments. You can even go a little bit above 1 for a little while. You can't bring it back up to 3, as that will quickly explode, and 2 months later you have 60,000 times as much.

The lockdown strategy is proven to work. It has worked in many places around the world to get the virus load down or even to zero in a few cases. It has even worked in areas of the United States. But once it gets it down, you can't return to full open, but you can open up a bit. Until you get a better strategy.

Herd immunity means 2 million dead. That's 3 times the dead of all the wars in the history of the US going back to the Revolutionary war. That's your answer?

Add new comment