539 karma · joined February 29, 2020
I should have made it clear but I wasn't attempting to construct an exhaustive list of risk factors, but more give an example of what that might look like
In short, the important distinction is not that it's embedded into the founding documents, but that we do not place the limitations on free speech that other countries do.
i.e. places like UK, germany, france - while they may have the phrase "free speech" in their documents, the limitations put on it are extreme enough that they're basically referring to a totally different concept.
So, it would be more accurate to say that other countries do not share our concept of free speech, in that ours has _very_ few limits put on it.
Public policy wrt Covid basically occurs at the intersection of epidemiology and economics. The epidemiologists can help tell you what will happen under different scenarios, and the economists can tell you what will happen economically speaking in those same scenarios.
We're basically balancing the economic impact directly attributable to sick days and to COVID-19 deaths, with the impact of the shutdown itself.
As you already know, my conclusion is that the latter is far more severe. But that balancing act is of course precisely what all the debate is about.
I think the Diamond Princess is the perfect case study to use as an upper bound, since we conducted extensive testing and the population is heavily skewed towards the elderly (and therefore higher risk than general population).
See https://www.medrxiv.org/content/10.1101/2020.03.05.20031773v...
> Adjusting for delay from confirmation-to-death, we estimated case and infection fatality ratios (CFR, IFR) for COVID-19 on the Diamond Princess ship as 2.3% (0.75%-5.3%) and 1.2% (0.38-2.7%).
Simply put, this argument is invalid. Why? Because my writeup went into the IFR. Therefore we can calculate how many deaths might occur in the "do nothing" scenario, and compare that to lockdown-induced deaths.
Also we should note that social distancing and lockdown are different things, and it _appears_ that the former is much more responsible for the slowed rate of transmission. My proposal does not involve getting rid of social distancing, although it does encourage it to be voluntary as opposed to compelled.
> TLDR: We're undercounting deaths due to covid-19 by ~60% nationally, and 40% in NYC.
What it does say is this:
> “Thus I would probably add about 15% or so to the known death toll,” she wrote to two city officials. “However — no city or state will be factoring this in or reporting it, so I don’t think we should either. We should just assume that the deaths are about 15% more than we can count, but not include them in official modeling, because we will never really know.”
It does pay some lip service to the notion that some of these at-home deaths could be precisely because of the "elective" (meaning, scheduled) surgeries that we had postponed.
Interestingly per the https://www.cdc.gov/heartdisease/facts.htm, we see something like 647,000 heart disease deaths per year. So, I haven't done the napkin math, but could the 20%-odd spike of cardiac arrest events they mentioned have more to do with postponed surgeries?
As you can see in your quote, I was talking about IFR. In general - and this is a theme of the writeup BTW - we should be looking at IFR and not CFR. With IFR we can reasonably extrapolate how many deaths we'll have before new infections halt, whereas CFR is borderline impossible to use because we'd need to know how many cases we would expect per infection. At which point, why not just calculate the IFR anyway, right?
See https://www.medrxiv.org/content/10.1101/2020.03.05.20031773v...
> Adjusting for delay from confirmation-to-death, we estimated case and infection fatality ratios (CFR, IFR) for COVID-19 on the Diamond Princess ship as 2.3% (0.75%-5.3%) and 1.2% (0.38-2.7%). Comparing deaths onboard with expected deaths based on naive CFR estimates using China data, we estimate IFR and CFR in China to be 0.5% (95% CI: 0.2-1.2%) and 1.1% (95% CI: 0.3-2.4%) respectively.
You might also want to look at https://www.eurosurveillance.org/content/10.2807/1560-7917.E...
Basically, users flagged it. I think the explanation Dang gave me in his e-mail reply is almost certainly correct, and it doesn't contain anything sensitive so I'll just share it here:
> All I can tell you, though, is that users flagged it. We don't know why they flag things, and can only guess. My guess in this case is that it's a combination of some users having covid-story-fatigue, and other users disagreeing with the view in the article. As you know, this question has become politicized and is getting more polarizing as we go forward.
The IMHE model, after re-estimating, _was_ too optimistic. My argument does not hinge on the IMHE model being too pessimistic; but rather simply that the IMHE model is basically useless and thus we shouldn't base policy on it.
I expect us to have well over 100,000 COVID-19 deaths - indeed probably 3x that but I admit I haven't sat down and done the age-bracket-weighted mortality napkin math. (Basically, take each age group's IFR and multiply by that age group's population in the US, after multiplying by some constant <1 which represnts the sum of herd immunity threshold plus overshoot)
So basically, I think that COVID-19 mortality is going to be way worse than the model predicts, but given the way I view the cost of lockdown, etc, I still think the approach I am proposing is superior to locking down.
Does that make sense?
Say R0 is 2, 1/(1 - R0) means that we already hit herd immunity before SARS-CoV-2 even existed!
Yes, I am a person that makes and is currently making mistakes.
> And you also think, that politicians are doing an economycal suicide just because they like it?
I analyzed the "game theory" of this a bit here: https://www.ryankemper.io/post/2020-04-29-the_case_for_endin...
> Hanlon’s Razor is very relevant here; we think it is not reasonable to assume that they were trying to usher in a techno-totalitarian state. Rather, they were trying to prevent mortality, but their incentive structure is such that it is much worse to have done nothing and later find out that the virus was worse than we thought, than what actually happened, which is to have undertaken actions that were self-destructive and then discovered that the virus was way less deadly than we thought.
An IFR of 1% seems too high. This doesn't seem to line up with data from the Diamond Princess (~1% IFR in a population biased towards elderly). We can also look as a lower bound at the Theodore Roosevelt (the navy ship), which had I believe 840 positives, 4 hospitalizations, 1 death (https://www.sandiegouniontribune.com/news/military/story/202...), and this is a population heavily biased towards the young and fit thus why it's a nice lower bound.
> 2. There is little evidence that we can reach herd immunity levels of infection with a Coronavirus like SARS-CoV-2 as both the unconstrained R0 is too high and the immunity too short lived.
The unconstrained R0 doesn't mean we can't reach immunity, it just means that almost every single person needs to get it, which to me still seems preferable to years-long lockdown.
But the more interesting point of yours is the part about short-lived immunity. I'm going to copy-paste a portion of https://news.ycombinator.com/item?id=23025880 to avoid repeating myself too much:
I think of immunity simplistically as two components:
(1) The presence of actively circulating antibodies in the bloodstream. This is what the (oddly controversial) serology studies are measuring. It is thought that having a significant quantity of these antibodies prevents infection - i.e., what most people envision when they talk about immunity.
(2) Even after the antibodies have faded, there are still Memory B Cells, which lay dormant up to decades, waiting for exposure to the characteristic antigen (in this case, an antigen telling them that they have encountered SARS-CoV-2), at which point they resume and rapidly scale up production of antibodies.
The thinking here is that reinfection is likely possible after a sufficient length of time - whether that's a couple months or a couple years isn't yet known - but when you do get infected, your immune system will respond sooner, more strongly, and thus you will achieve a far lower peak viral load meaning a less serious infection with reduced transmissibility.
So in short, I'd like to see what makes you think the reinfection immunity window is short, but regardless of the window length, we can hopefully agree that the immunological memory cells will stick around for a while.
> 3. The IFR has a threshold where it dramatically increases when the number of new cases overwhelms the hospital capacity. We have seen this to a certain extend in Wuhan, Northern Italy and NYC.
Agreed. My only caveat is this effect is probably not _as_ dramatic as it initially looked, but it is definitely dramatic.
> As treatments improves the IFR should fall. This mean if we can shift an infection to the future we should see less people die. Flattening the curve is more than just avoiding overwhelming the hospital system, it is also about shift cases to a time when we have improved treatments.
Also agreed. I thought about paying lip service to this notion but decided to leave it out of the first draft. Basically my assumption (and it is an assumption) is that the treatments we discover will do a little bit but not a lot. So probably what we've learned thus far about proning, managing oxygen levels etc doesn't leave a ton of room for improvement without a miracle therapeutic. I think antivirals and other treatments will help a small amount but not nearly enough to justify postponing infection.
But yes, at this point I think we can say that in a scenario of successful full containment+vaccine, we would have less area under curve as far as the "overshoot" is concerned and also to a lesser extent the improvement in treatments.
> Sweden has shown that it doesn’t matter if you have an official lockdowns or not, most people will self isolate on their own. There is little extra economic damage from imposing official lockdown policies since all the damage is done by people avoiding infection on their own.
I have to strongly disagree with this. In fact I view this as a strong point in favor of "my" side: we can get the majority of the benefits of the lockdown without actually locking down.
The magnitude of economic damage is much more than just the demand dropoff. We've forced closures of businesses that actually would have been able to stay open. And particularly small businesses which are not well equipped to weather systemic shocks.
But to be clear, I absolutely agree that even without a lockdown we would see very sizeable demand drop-off and thus there would still be some sizeable amount of furloughing, layoffs etc.
--
Thanks so much for the points. In particular I'd really like to hear more from you about (2), since that is the point I disagree most with, and after that, point (1). 3-5 I have very few objections as discussed
The most important point of criticism is this section: https://www.lesswrong.com/posts/QuzAwSTND6N4k7yNj/seemingly-...
(BTW, something that occurred to me is that the fact that infants are largely spared from COVID-19 is even more of a blessing than it already seems. Once this passes through our population, new infections will be in those who are not immune which will primarily be infants/young children, who we know thus far have had really good outcomes relative to adults. That means the enduring legacy of this will be much better than what we see with Influenza)
> One, saying something that is 300% larger than another is “similar” is stretching the definition of similarity.
I defined "ballpark" in the article as less than an order of magnitude. But, I think this is a thought exercise that might help:
Imagine that SARS-CoV-2 doesn't exist, but we learn that this flu season is going to be extra deadly. Instead of the ~50k of deaths last year, we know that there will be ~150k deaths.
Would we rationally take the same extreme of measures that we have taken for Covid? (Or, conversely, should we already have been taking these measures for past flu seasons)?
Also I think you're very aware of this, but just for any on-lookers: don't look at the flu comparisons as intending to minimize COVID-19; but rather comparing to flu gives us an existing mental model that most people are familiar with that helps us figure out what measures make sense and what don't.
The above being said, I think the new york numbers are a perfect upper bound for something close to what a worst case scenario looks like. Although we'd have to look at the age breakdown and obesity rates there; it's definitely very imaginable that there might be other cities in the US with relatively more at-risk individuals.
(BTW I didn't explain it thoroughly in the writeup but New York's data is why I gave a really broad range of .1-.7% net IFR - and it should be clear but this is still very napkin math-y on my part)
I'm also curious if NY serological data includes <18 year olds - if it doesn't include children we need to account for that when extrapolating in the 100% infection scenario.
Anyway, the above is all a really drawn out way of saying that I agree :P
> Additionally, if we give up on social distancing, then those people will die in a large wave, overwhelming our hospitals.
If we gave up everything, certainly. I think with mitigation targeted at the at-risk, that's less likely.
For me "ending lockdown" doesn't mean that everyone stops wearing masks and/or trying to maintain physical distance where possible. But I do want to avoid a scenario where we're mandating that people wear masks outside as opposed to just encouraging them to.
But I do agree that just because we haven't overwhelmed hospitals yet doesn't mean that we're in the clear. I was considering saying as much in the doc but wanted to keep the word count down for the first iteration.
However I do think that since pretty much every city has at least some low amount of infection, and that we will have some capacity to do some clever shuffling of ventilators, tents, etc, that hospital overrun isn't nearly as much of a risk at this point overall.
> 4. I would love to see a plan that keeps seniors safe and lets everyone else out. So far, that plan is not forthcoming. It’s hard to see how we keep them safe if their family living with them and their caregivers are all infected. So far, we’ve done a piss poor job of keeping them safe in senior living facilities, even with a lockdown.
Yeah, really the best thing would be voluntary exposure of people surrounding the at-risk individuals. But I don't foresee that actually happening. I really think nursing homes are the low-hanging fruit here, and as you indicated it seems like we've already kind of screwed that one up in a lot of places.
But essentially, in the alternative of full containment+waiting for vaccine, it seems to me much harder for the caregivers surrounding the extremely-at-risk to be able to avoid getting infected and passing it on to the at-risk, whereas in a scenario where they can get it over with then they don't have to worry at all going forward. But that requires being able to detect that you're sick before infecting the at-risk individual so that you can isolate while recovering, which obviously requires really good testing capacity.
Basically, it would be great if we could selectively redirect the bulk of our testing capacity for individuals in these situations. But there's not a very clear way to do that, the last thing we want is having a bureaucrat decide if you're eligible for a test or not. So perhaps the best we can do is - and this is a theme here - encouraging those who are surrounding at-risk individuals to get tested extremely frequently.
--
Lastly just wanted to call out this:
> I think you’ve gotten a lot right, and i think analyses like this get down to the point where we can have a rational discussion.
That means a lot. I really want us as a society to get to a point where there is broad agreement on what the facts on the ground are, so that differences in policy come down to differences in philosophy/values rather than where it feels like we're at now as a society, which is basically two warring factions each with non-intersecting sets of facts ("the death rate is 5%!" versus "no the death rate is .03!)
(The irony in me having said it was "exactly" what Wikipedia says is not lost on me :P)
https://gitlab.com/ryankemper/writings/-/commit/093ade629447...
So then, it does seem that if we successfully contained the virus, and then avoided infection spread until a hypothetical future vaccine is developed, that we could avoid the number of infections required by some amount.
Just to tie it back to the broader lockdown vs not lockdown debate, we could probably say that a successful full containment strategy+vaccination could decrease the net area under the curve. As far as policy is concerned I still think the costs of locking down for that long (particularly given that the time is unbounded) is not worth it, but that is a very valid point that I will need to circle back and weave into the document.
> The premise that the United States, alone, recognizes freedom of speech as a legal right is simply absurd.
Hmm, maybe I accentuated too much the "embedded into their founding documents" part.
My understanding is that other countries do not have the same protections put upon speech. For example, in many places you can be arrested for racism or, say, holocaust denial.
That being said, I don't know much about France's laws.
I just found https://www.loc.gov/law/help/freedom-expression/france.php which is pretty enlightening.
So to start, I think the statement as I wrote it is false, since the distinction is really about the possible eroding of speech rights based on non-protected classes of speech. Here's an interesting excerpt from that link:
> Despite its foundational importance, freedom of speech was never intended to be absolute. In contrast to the First Amendment of the United States Constitution, the 1789 Declaration of Human and Civic Rights provided limits to freedom of expression in its very definition. Article 10 declares that “[n]o one may be disturbed on account of his opinions, even religious ones, as long as the manifestation of such opinions does not interfere with the established Law and Order.”[7] Article 11 provides that “[a]ny citizen may therefore speak, write and publish freely, except what is tantamount to the abuse of this liberty in the cases determined by Law.”[8]
And further:
> Similarly, the European Convention on Human Rights declares that
> the exercise of these freedoms, since it carries with it duties and responsibilities, may be subject to such formalities, conditions, restrictions or penalties as are prescribed by law and are necessary in a democratic society, in the interests of national security, territorial integrity or public safety, for the prevention of disorder or crime, for the protection of health or morals, for the protection of the reputation or rights of others, for preventing the disclosure of information received in confidence, or for maintaining the authority and impartiality of the judiciary.[9]
So, the interesting bit here is there are a _ton_ of categories that are valid grounds to restrict speech. So again I think as I stated it it is false, but were my statement more accurate about making the point around the restrictions on content, it would presumably be correct.
> I'd soften it a lot, because as far as I know, no one is being forcibly imprisoned in their homes under shelter-in-place orders in the US. At least in my case, the orders for my city don't prevent me from moving freely. I have to wear a mask in public, and my local supermarket opens later and closes earlier and has item limits to stop panic buying, and there is an 11pm curfew, but those are hardly what I would consider to be the signs of a totalitarian police state.
Yes, I would soften it a lot as well. "Slightly" is an understatement. I do take issue with this though:
> and there is an 11pm curfew
A curfew to me is exactly the restriction on freedom of movement/expression that I would consider highly problematic. I'm curious, is that curfew related to COVID-19? Or was it already in place? Because after 11pm seems like the worst time to restrict travel given that there's not as many humans around.
(Thanks for making those points!)
Additionally, the two companies that I had narrowed the search down to did not freeze hiring. Pretty lucky, since tons of companies did, although I constrained my search to remote jobs only so that's probably why.
I accepted a job offer and start soon, actually.
That being said, I do think that the general experience of being jobless during this has made me more sympathetic of those that truly lost their jobs than I otherwise would have. Which is not to say that I wouldn't have been sympathetic, I absolutely would have, but the fact that I had to deal with COBRA and other fun artifacts definitely made things a bit more tangible for me.
Excellent point/question.
Sorry to bombard you with replies, but I realized I never addressed this.
I think you might have a different idea of what "hero" means, basically all I meant is that we tend to elevate these figures (Fauci in this case) up as "heroes" and therefore act like they are infallible and thus you cannot express an opinion that contradicts them. That is what I was arguing against.
It's not about "hero" in the sense of, like, what a young child would imagine a hero is. So in the sense I'm using it, people totally do view the "nerdy scientist" (Fauci) as a hero. And I understand why they do.
> This seems like it should be a point in support of lockdown to me (tbh, it seems like it shouldn't be a point at all, it's just an attempt to discredit experts)
No, it's not discrediting them, it's just saying: evaluate policies and statements on their merits, not based on simply who said them.
I hope that makes sense.
Thanks again for all your thoughtful feedback. You clearly put time into it (and a good bit of frustration, understandably :P)
First, `1/1-R_0`. This is exactly what Wikipedia says: https://en.wikipedia.org/wiki/Herd_immunity#Mechanics
And it makes intuitive sense that once you infect <= 1.00 people, growth is no longer exponential. So I don't think that point is wrong at all.
> The idea that vaccination and epidemics need the same percent of immunity is completely 100% wrong. An active epidemic is going to infect way more people before dying out than the simple, naive herd immunity calculation will give you. In fact, while an r0 of 2 indicates that you get herd immunity with just 50% of the population, a typical active epidemic with the same r0 would infect 80+% before dying out.
This doesn't make any sense to me. Vaccination is just an artificial way to build immunity. So I don't understand your active epidemic point at all. Or put another way, as the proportion of % recovered goes up, we start approaching the eventual limit where `1/1-R_0`->1. That is to say, in your example we start with R_0 = 2, then as people recover that value slowly starts decreasing until it hits 1 at the herd immunity threshold. It's not clear to me what the discrepancy you're detecting is here.
Thanks for the feedback. I think the dog-whistle part is quite true. I will need to circle back and add citations where I can. Particularly given that one of the thematic elements is that the lockdown is not based upon evidence.
> There seems to be excessive use of bolding random segments of sentences. I don't know if this is just a stylistic thing, but ¯\_(ツ)_/¯.
I was worried about this being distracting. I think particularly bolding segments within a sentence is definitely distracting, so I will (eventually) try to clean that up.
> On suicides, it seems like most of the argument should be verifiable from data (i.e. by its claims, we should be able to see an actual increase in suicide deaths). Is that data not available yet?
As far as I know it's not, but I'm not entirely sure where to go looking for weekly suicide data. I've only been able to find data over much longer intervals. This is where being an actual expert would probably help! :P
> this is clearly a plea to the emotions of a specific subset of the population
Just to bite of a piece of this before I conclude here, I personally think that a shift to censorship on such a wide scale is absolutely something to raise a bit of moral panic about. The new usage of the term "disinformation" has hit a point of Orwellian double-speak.
Typo fixed! I think I've found at least 15 typos total now. Oops...
Right, but in this case it's leading to workers getting furloughed. So that is a direct lowering of capacity, although I suppose one could argue that those workers will be called back when shit hits the fan. But if that's the case...why not have them doing important preventative care in the meantime, since we know our hospitals are not overwhelmed yet? Basically anything that doesn't require people being in hospital beds for days since that would take away resources for future COVID-19 victims.
> Finally, those less-vulnerable members of the population you are imagining actually end up hospitalized at a rate between 2.5% and 7.4%.
This appears to be a hospitalization rate that is not accounting for all the hidden infections, which is why that number is so high. That's why the serological studies are so important. Otherwise, we need to estimate how many "cases" there will be (meaning, "serious cases" essentially), which is a lot more difficult than finding the true IFR / hospitalization rate and then multiplying by age-stratified population size. In particular, the CDC numbers are very at odds with the hospitalization rate of some of the serological studies I cited in my article. And it's not just a factor of 2x, it seems to be a massive difference. For example they show a 2.5% hospitalization rate for ages 18-49 whereas the number I've been using shows a .8% hospitalization rate for ages 40-49 at the high end, and then a .2% hospitalization rate for ages 20-29 at the low end.
I will need some time to dig into https://projects.propublica.org/graphics/covid-hospitals, but I'm having trouble finding where they give their estimated hospitalization rates. I did note that the article has the following quote:
> “You will have people on gurneys in hallways," Cuomo said at a press conference on Monday, later adding that New York would need up to 110,000 total hospital beds, around twice the number it currently has. “That is what is going to happen now if we do nothing."
And we know now that this was absolutely not the case. Recall that New York ended up not really needing the USS Comfort, and we now know that at least 15% of the population has been infected per the serology studies. So the fact that that prediction was wrong makes it feel like other predictions using the same model will be very wrong. I'll see if I can find the paper that the Harvard team published.
And then this quote seems to imply they're operating off of the bogus 20% hospitalization rate that I mentioned in the article:
> According to the Harvard scenario where 40% of adults in the country contract the disease, about half a million people in the San Francisco region may get infected, with more than 100,000 residents requiring hospitalization.
Anyway, moving on:
> The idea of cordoning off the vulnerable when those who are most vulnerable are essential workers in groceries, transit, delivery not to mention the non-essential service workers, hospitality staff, janitors, teachers, etc. What you are proposing can have no basis in reality.
It's not clear to me why this is the case. Those essential workers who are vulnerable would be quarantined (whereas right now I believe they are not quarantined which is a problem). So, the 70 year old working as a bank teller would stay at home, collect unemployment (I wish we had a better system of unemployment but that's another matter), and the bank would presumably try to fill the role with someone else who is available to work in the meantime.
--
Thanks for the discussion.
Have you seen any scientific literature alleging this possibility in those who are not seriously ill? Not to sound like the WHO, but I have not personally come across any evidence that this is at all true. And indeed it seems highly non-credible.
> Also, while the the goal of flattening the curve was focused on lowering the impact on health care, as seen in Australia and New Zealand it can also result in stopping nearly all local transmission.
This is true, but now they can't let anyone into their country until the whole island has been vaccinated. That could get very tricky.
Well, I suppose they could do what Taiwan and other countries have done and aggressively test&quarantine all entrants. That's certainly possible.
(For what it's worth, I agree that it doesn't make sense to participate in a discussion like this without at least having done some background reading)