We’re fast approaching flu season, which is especially harrowing this year.
We, as a people, have struggled to respond to this calamity. We have a lot of scientific data about Covid-19 now, but science is never value-neutral. The way we design experiments reflects our biases; the way we report our findings, even more so.
For example, many people know the history of Edward Jenner inventing the world’s first vaccine. Fewer are aware of the long history of inoculation in Africa (essentially, low-tech vaccination) that preceded Jenner’s work.
So it’s worthwhile taking a moment to consider the current data on Covid-19.
Data alone can’t tell us what to do – the course of action we choose will reflect our values as a society. But the data may surprise a lot of people – which is strange considering how much we all feel that we know about Covid-19.
Indeed, we may realize that our response so far goes against our professed values.
Spoiler: I think we shouldn’t close in-person school.
Since April, I’ve written several essays about Covid-19. In these, I’ve made a number of predictions. It’s worthwhile to consider how accurate these predictions have been.
This, after all, is what science is. We use data to make an informed prediction, and then we collect more data to evaluate how good our prediction was.
Without the second step – a reckoning with our success or failure – we’re just slinging bullshit.
I predicted that our PCR tests were missing most Covid-19 infections, that people’s immunity was likely to be short-lived (lasting for months, not years), and that Covid-19 was less dangerous than seasonal influenza for young people.
These predictions have turned out to be correct.
In my essays, I’ve tried to unpack the implications of each of these. From the vantage of the present, with much more data at our disposal, I still stand by what I’ve written.
But gloating’s no fun. So I’d rather start with what I got wrong.
My initial predictions about Covid-19 were terrible.
I didn’t articulate my beliefs at the time, but they can be inferred from my actions. In December, January, and February, I made absolutely no changes to my usual life. I didn’t recommend that travelers be quarantined. I didn’t care enough to even follow the news, aside from a cursory glance at the headlines.
While volunteering with the high school running team, I was jogging with a young man who was finishing up his EMT training.
“That new coronavirus is really scary,” he said. “There’s no immunity, and there’s no cure for it.”
I shrugged. I didn’t know anything about the new coronavirus. I talked with him about the 1918 influenza epidemic instead.
I didn’t make any change in my life until mid-March. And even then, what did I do?
I called my brother and talked to him about the pizza restaurant – he needed a plan in case there was no in-person dining for a few months.
My next set of predictions were off, but in the other direction – I estimated that Covid-19 was about four-fold more dangerous than seasonal influenza. The current best estimate from the CDC is that Covid-19 is about twice as dangerous, with an infection fatality ratio of 0.25%.
But seasonal influenza typically infects a tenth of our population, or less.
We’re unlikely to see a significant disruption in the transmission of Covid-19 (this is the concept of “herd immunity”) until about 50% of our population has immunity from it, whether from vaccination or recovery. Or possibly higher – in some densely populated areas, Covid-19 has spread until 70% (in NYC) or even 90% (in prisons) of people have contracted the disease.
Population density is hugely important for the dynamics of Covid-19’s spread, so it’s difficult to predict a nation-wide threshold for herd immunity. For a ballpark estimate, we could calculate what we’d see with a herd immunity threshold of about 40% in rural areas and 60% in urban areas.
Plugging in some numbers, 330 million people, 80% urban population, 0.25% IFR, 60% herd immunity threshold in urban areas, we’d anticipate 450,000 deaths.
That’s about half of what I predicted. And you know what? That’s awful.
Each of those 450,000 is a person. Someone with friends and family. And “slow the spread” doesn’t help them, it just stretches our grieving to encompass a whole year of tragedy instead of a horrific month of tragedy.
If we don’t have a safe, effective vaccine soon enough, the only way to save some of those 450,000 people is to shift the demographics of exposure.
Based on the initial data, I concluded that the age demographics for Covid-19 risk were skewed more heavily toward elderly people than influenza risk.
I may have been wrong.
It’s difficult to directly compare the dangers of influenza to the dangers of Covid-19. Both are deadly diseases. Both result in hospitalizations and death. Both are more dangerous for elderly or immunocompromised people, but both also kill young, healthy people.
Typically, we use an antigen test for influenza and a PCR-based test for Covid-19. The PCR test is significantly more sensitive, so it’s easier to determine whether Covid-19 is involved a person’s death. If there are any viral particles in a sample, PCR will detect them. Whereas antigen tests have a much higher “false negative” rate.
Instead of using data from these tests, I looked at the total set of pneumonia deaths. Many different viruses can cause pneumonia symptoms, but the biggest culprits are influenza and, in 2020, Covid-19.
So I used these data to ask a simple question – in 2020, are the people dying of pneumonia disproportionately more elderly than in other years?
I expected that they would be. That is, after all, the prediction from my claims about Covid-19 demographic risks.
I was wrong.
In a normal year (I used the data from 2013, 2014, and 2015, three years with “mild” seasonal influenza), 130,000 people die of flu-like symptoms.
In 2020 (at the time I checked), 330,000 people have died of flu-like symptoms. Almost three times as many people as in a “normal” year.
For people under the age of 18, we’ve seen the same number of deaths (or fewer) in 2020 as in other years. The introduction of Covid-19 appears to have caused no increased risk for these people.
But for people of all other ages, there have been almost three times as many people dying of these symptoms in 2020 compared to other years.
In most years, one thousand people aged 25-34 die of these symptoms; in 2020, three thousand have died. In most years, two thousand people aged 35-44 die of these symptoms; in 2020, six thousand have died. This same ratio holds for all ages above eighteen.
Younger people are at much less risk of harm from Covid-19 than older people are. But, aside from children under the age of eighteen, they don’t seem to be exceptionally protected.
Of course, my predictions about the age skew of risk might be less incorrect than I’m claiming here. If people’s dramatically altered behavior in 2020 has changed the demographics of exposure as compared to other years – which is what we should be doing to save the most lives – then we could see numbers like this even if Covid-19 had the risk skew that I initially predicted.
I predicted that four or more years would pass before we’d be able to vaccinate significant numbers of people against Covid-19.
I sure hope that I was wrong!
We now know that it should be relatively easy to confer immunity to Covid-19. Infection with other coronaviruses, including those that cause common colds, induce the production of protective antibodies. This may partly explain the low risk for children – because they get exposed to common-cold-causing coronaviruses so often, they may have high levels of protective antibodies all the time.
Several pharmaceutical companies have reported great results for their vaccine trials. Protection rates over 90%.
So the problem facing us now is manufacturing and distributing enough doses. But, honestly, that’s the sort of engineering problem that can easily be addressed by throwing money at it. Totally unlike the problem with HIV vaccines, which is that the basic science isn’t there – we just don’t know how to make a vaccine against HIV. No amount of money thrown at that problem would guarantee wide distribution of an effective vaccine.
We will still have to overcome the (unfortunately significant) hurdle of convincing people to be vaccinated.
For any individual, the risk of Covid-19 is about twice the risk of seasonal influenza. But huge numbers of people choose not to get a flu vaccine each year. In the past, the United States has had a vaccination rate of about 50%. Here’s hoping that this year will be different.
Covid-19 spreads so fast – and so silently, with many cases of infected people who feel fine but are still able to spread the virus – that it will almost certainly be a permanent resident of the world we live in. We’re unlikely to eradicate Covid-19.
Which means that elderly people will always be at risk of dying from Covid-19.
The only way to protect people whose bodies have gone through “age-related immunosenence” – the inevitable weakening of an immune system after a person passes the evolutionarily-determined natural human lifespan of about 75 years – will be to vaccinate everybody else.
Depending on how long vaccine-conferred immunity lasts, we may need to vaccinate people annually. I worry, though, that it will become increasingly difficult to persuade people to get a Covid-19 vaccine once the yearly death toll drops to influenza-like levels – 50,000 to 100,000 deaths per year in the United States.
I wrote, repeatedly, that immunity to Covid-19 is likely to be short-lived. Immunity to other coronaviruses fades within a few months.
(Note: you may have seen articles in the New York Times suggesting that we’ll have long-lasting protection. They’re addressing a different question — after recovery, or vaccination, are you likely to become severely ill with Covid-19? And the answer is “probably not,” although it’s possible. When I discuss immunity here, I mean “after recovery, or vaccination, are you likely to be able to spread the virus after re-infection?” And the answer is almost certainly “yes, within months.”)
And I wrote about the interplay between short-lived immunity and the transmission dynamics of an extremely virulent, air-born virus.
This is what the Harvard public health team got so wrong. When we slow transmission enough that a virus is still circulating after people’s immunity wanes, they can get sick again.
For this person, the consequences aren’t so dire – an individual is likely to get less sick with each subsequent infection by a virus. But the implications for those who have not yet been exposed are horrible. The virus circulates forever, and people with naive immune systems are always in danger.
It’s the same dynamics as when European voyagers traveled to the Americas. Because the European people’s ancestors lived in unsanitary conditions surrounded by farm animals, they’d cultivated a whole host of zoogenic pathogens (like influenza and this new coronavirus). The Europeans got sick from these viruses often – they’d cough and sneeze, have a runny nose, some inflammation, a headache.
In the Americas, there were fewer endemic diseases. Year by year, people wouldn’t spend much time sick. Which sounds great, honestly – I would love to go a whole year without headaches.
But then the disgusting Europeans reached the Americas. The Europeans coughed and sneezed. The Americans died.
And then the Europeans set about murdering anyone who recovered. Today, descendants of the few survivors are made to feel like second-class citizens in their ancestral homelands.
In a world with endemic diseases, people who have never been exposed will always be at risk.
That’s why predictions made in venues such as the August New York Times editorial claiming that a six- to eight-week lockdown would stop Covid-19 were so clearly false. They wrote:
Six to eight weeks. That’s how long some of the nation’s leading public health experts say it would take to finally get the United States’ coronavirus epidemic under control.
For proof, look at Germany. Or Thailand. Or France.
Obviously, this didn’t work – in the presence of an endemic pathogen, the lockdowns preserved a large pool of people with naive immune systems, and they allowed enough time to pass that people who’d been sick lost their initial immunity. After a few months of seeming calm, case numbers rose again. For proof, look at Germany. Or France.
Case numbers are currently low in Thailand, but a new outbreak could be seeded at any time.
And the same thing is currently happening in NYC. Seven months after the initial outbreak, immunity has waned; case numbers are rising; people with mild second infections might be spreading the virus to friends or neighbors who weren’t infected previously.
All of which is why I initially thought that universal mask orders were a bad idea.
We’ve known for over a hundred years that masks would slow the spread of a virus. The only question was whether slowing the spread of Covid-19 would cause more people to die of Covid-19.
And it would – if a vaccine was years away.
But we may have vaccines within a year. Which means that I may have been wrong. Again, the dynamics of Covid-19 transmission are still poorly understood – I’ll try to explain some of this below.
In any case, I’ve always complied with our mask orders. I wear a mask – in stores, at school pickup, any time I pass within six feet of people while jogging.
To address global problems like Covid-19 and climate change, we need global consensus. One renegade polluting wantonly, or spewing viral particles into the air, could endanger the whole world. This is precisely the sort of circumstance where personal freedom is less important than community consensus.
The transmission dynamics of Covid-19 are extremely sensitive to environment. Whether you’re indoors or outdoors. How fast the air is moving. The population density. How close people are standing. Whether they’re wearing masks. Whether they’re shouting or speaking quietly.
Because there are so many variable, we don’t have good data. My father attended a lecture and a colleague (whom he admires) said, “Covid-19 is three-fold more infectious than seasonal influenza.” Which is bullshit – the transmission dynamics are different, so the relative infectivity depends on our behaviors. You can’t make a claim like this.
It’s difficult to measure precisely how well masks are slowing the spread of this virus.
But here’s a good estimate: according to Hsiang et al., the number of cases of Covid-19, left unchecked, might have increased exponentially at a rate of about 34% per day in the United States.
That’s fast. If about 1% of the population was infected, it could spread to everyone within a week or two. In NYC, Covid-19 appear to spread to over 70% of the population within about a month.
(To estimate the number of infections in New York City, I’m looking at the number of people who died and dividing by 0.004 – this is much higher than the infection fatality rate eventually reported by the CDC, but early in the epidemic, we were treating people with hydroxychloraquine, an unhelpful poison, and rushing to put people on ventilators. We now know that ventilation is so dangerous that it should only be used as a last resort, and that a much more effective therapy is to ask people to lie on their stomachs – “proning” makes it easier to get enough oxygen even when the virus has weakened a person’s lungs.)
Masks dramatically slow the rate of transmission.
A study conducted at a military college – where full-time mask-wearing and social distancing were strictly enforced – showed that the number of cases increased from 1% to 3% of the population over the course of two weeks.
So, some math! Solve by taking ten to the power of (log 3)/14, which gives an exponential growth rate of 8% per day. Five-fold slower than without masks.
But 8% per day is still fast.
Even though we might be able to vaccinate large numbers of people by the end of next year, that’s not soon enough. Most of us will have been sick with this – at least once – before then.
I don’t mean to sound like a broken record, but the biggest benefit of wearing masks isn’t that we slow the rate of spread for everyone — exponential growth of 8% is still fast — but that we’re better able to protect the people who need to be protected. Covid-19 is deadly, and we really don’t want high-risk people to be infected with it.
I’ve tried to walk you through the reasoning here — the actual science behind mask policies — but also, in case it wasn’t absolutely clear: please comply with your local mask policy.
You should wear a mask around people who aren’t in your (small) network of close contacts.
I’m writing this essay the day after New York City announced the end of in-person classes for school children.
This policy is terrible.
A major problem with our response to Covid-19 is that there’s a time lag between our actions and the consequences. Human brains are bad at understanding laggy data. It’s not our fault. Our ancestors lived in a world where they’d throw a spear at an antelope, see the antelope die, and then eat it. Immediate cause and effect makes intuitive sense.
Delayed cause and effect is tricky.
If somebody hosts a party, there might be an increase in the number of people who get sick in the community over the next three weeks. Which causes an increase in the number of hospitalizations about two weeks after that. Which causes people to die about three weeks after that.
There’s a two-month gap between the party and the death. The connection is difficult for our brains to grasp.
As a direct consequence, we’ve got ass-hats and hypocrites attending parties for, say, their newly appointed Supreme Court justice.
But the problem with school closures is worse. There’s a thirty year gap between the school closure and the death. The connection is even more difficult to spot.
Even if you have relatively limited experience reading scientific research papers, I think you could make your way through this excellent article from Chistakis et al.
The authors link two sets of existing data: the correlation between school closures and low educational achievement, and the correlation between low educational achievement and premature death.
The public debate has pitted “school closures” against “lives saved,” or the education of children against the health of the community. Presenting the tradeoffs in this way obscures the very real health consequences of interrupted education.
These consequences are especially dire for young children.
The authors calculate that elementary school closures in the United States might have (already!) caused 5.5 million years of life lost.
Hsiang et al. found that school closures probably gave us no benefit in terms of reducing the number of Covid-19 cases, because children under 18 aren’t significant vectors for transmission (elementary-aged children even less so), but even if school closures had reduced the number of Covid-19 cases, closing schools would have caused more total years of life to be lost than saved.
The problem – from a political standpoint – is that Covid-19 kills older people, who vote, whereas school closures kill young people, who are intentionally disenfranchised.
And, personally, as someone with far-left political views, it’s sickening for me to see “my” political party adopt policies that are so destructive to children and disadvantaged people.
So, here’s what the scientific data can tell us so far:
- We will eventually have effective vaccines for Covid-19. Probably within a year.
- Covid-19 spreads even with social distancing and masks, but the spread is slower.
- You have no way of knowing the risk status of people in a stranger’s bubble. (Please, follow your local mask orders!)
- Schools – especially elementary schools – don’t contribute much to the spread of Covid-19.
- School closures shorten children’s lives (and that’s not even accounting for their quality of life over the coming decades).
- An individual case of Covid-19 is about twice as dangerous as a case of seasonal influenza (which is scary!).
- Underlying immunity (from prior disease and vaccination) to Covid-19 is much lower than for seasonal influenza, so there will be many more cases.
- Most people’s immunity to Covid-19 probably lasts several months, after which a person can be re-infected and spread the virus again.
So, those are some data. But data don’t tell us what to do. Only our values can do that.
Personally, I value the lives of children.
I wouldn’t close schools.