rallison

I'm not sure what California has to do with the point I was making?

Anyway, it's problematic because around half of the total recorded deaths for New York state, for the entire pandemic up to now, likely resulted from infections that happened during the no/minimal restrictions period in March of 2020. Presumably the original poster wanted a "strong restrictions" vs "light restrictions" comparison w.r.t. to deaths. And that's fine; it's just that New York is going to give you a poor starting point for that comparison unless you dive deeper.

New York ends up being a problematic comparison point, because they were at zero restrictions until mid March 2020, so most of the infections through near the end of March were effectively a "no restrictions" test case. I.e. people testing positive late March were the highly symptomatic cases (because of limited test availability), so these were generally people infected 10+ days before, which means they were before any restrictions. Restrictions on mass gatherings happened on March 12th, but the stay-at-home order didn't happen until March 22nd.

What I'm getting at is that, if you want to use New York as a test comparison for loose vs tight restrictions, you're going to have a much deeper analysis to do, and a large portion of the deaths in New York were effectively resulting from a period when there were zero to minimal restrictions.

If the downsides weren't so significant, then sure, the pro side makes sense.

But the downsides are incredibly significant - that you have huge numbers of people who spend time in jail or spend money for a bail bond, who are indeed innocent, is a sign of a fundamentally broken system.

> It should be relative to your ability to pay

Sure, this would help, but I suspect it would still end up often unjustly impacting those who are poorer and innocent.

If people don't show up for trial, charge them for that then.

But pre trial, while presumed innocent, aside from unique high flight risk scenarios? The bail system is terrible, and fundamentally unfair.

Edit: to be clear, making it relative to ability to pay, and eliminating commercial bail bondsmen would be a big step in the right direction. I still think that's rife with issues, but it would be significantly better than what we have now.

This is an odd resource, at least with respect to IFR. I didn't look to see if the CFR section was more reasonable. They don't date the commentary in the IFR section, but it seems like it's mostly March/early April? If so, we obviously know a lot more since then.

> We could make a simple estimation of the IFR as 0.35%, based on halving the lowest boundary of the CFR prediction interval in Europe.

I mean, sure, maybe an ok guess way at the beginning.

> In Swine flu, the IFR ended up as 0.02%, fivefold less than the lowest estimate during the outbreak (the lowest estimate was 0.1% in the 1st ten weeks of the outbreak).

We simply weren't doing the surveillance needed for the 2009 H1N1 pandemic during the pandemic to have reasonable estimates for quite a while. We are doing much more of that for Covid - we have serology studies, and we have fairly accurate death numbers (at least in some countries), the key pieces to put together more robust IFR estimates. We didn't have those for H1N1. I also find it a little odd this page just generically references it as Swine flu, vs being specific about it being the 2009 iteration of it. I'd expect that from a news outlet, not an Oxford associated group.

> In Iceland, where the most testing per capita has occurred, the IFR lies somewhere between 0.03% and 0.28%.

Iceland contained their outbreak (yay!), and had very few cases overall, and just 10 deaths. Their CFR is 0.28%. Given their massive amounts of testing, it's hard to understand where the 0.03% comes from (and, obviously, is wildly low). But mostly, there were so few cases in Iceland that's it's hard to draw many fatality rate conclusions.

But mostly:

> Antibody testing will provide an accurate understanding of how many people have been infected so far, and permit a more accurate estimate of the IFR.

> *Estimating CFR and IFR in the early stage of outbreaks is subject to considerable uncertainties, the estimates are likely to change as more data emerges.

Seems to be consistent with this having been written up in March/early April. Since they haven't updated it, and the reasoning even at the time was fairly poor, this doesn't seem a useful resource (at least for IFR).

It's worth noting that Sweden is actually more strict than a number of states in the US - 50 person limit on any gatherings, and that limit applies to restaurants as well. They were also better positioned than almost any country to potentially tackle the pandemic well - wealthy, very high percentage of single person households, high percentage of jobs amenable to working from home, and yet their deaths per million rate is close to that of the US. That rate is also 5-10x that of their Nordic neighbors.

What they did help show was that certain mixes of mitigation (vs suppression) measures could potentially keep R below 1 in certain places. But even there, because of a limited testing strategy, it was hard to confirm this for quite a while in Sweden. It would have been a lot more helpful if there'd been a lot more testing surveillance in Sweden, but alas.

I think this is misunderstanding the point. It's not that we can't protect the most vulnerable, it's that we can't protect them particularly well IF community spread is high.

We are partially effective in protecting the most vulnerable, e.g. through restrictions on nursing homes. But, infections in these communities still correlate with the level of community spread - e.g. if you halve the new case rate in nursing homes with restrictions, but you double the new case rate in the community, you may be back to where you started.

The point being, as you allow significantly more spread in lower risk populations, you'll have a really hard time keeping higher risk groups at their current level of transmission risk (which, in the US, is already a fairly high level of transmission risk).

> Any number of different things

Yes, it can mean lots of different things, potentially, which is why it is useful to look at factors that may have impacted implied IFRs. E.g. is the population you are studying older than the national average (NY: about average for percent above age 65). Is the population unhealthier than average, e.g. looking at obesity rates (NY: significantly lower obesity than the national average). You want to look at how many deaths were in nursing homes vs not, to see if policy decisions meant disproportionate impact in high risk populations (NY: one of states with the lowest percentage of deaths linked to nursing homes). So many factors were either neutral for NY, or suggest that NY's IFR would be lower than the national average. But, one also considers that NY was stretched thin, which no doubt meant non-ideal care for many people, so that likely countered some of the advantages NY had had.

You also then want to compare with other serology studies, and the implied IFRs from those studies - are we looking at outlier numbers, or does it generally align? Outliers can help uncover factors that were missed, or flaws in the study itself. And that's the thing, all of the NY numbers have suggested an IFR above 0.5%, potentially upwards of 1%, which is broadly in line with other IFR calculations from other places.

> Point being, the Santa Clara study was a lot closer to right than wrong, and dismissing it, and the people involved, because of politics and methodological mistakes is a cheap shot.

> When it came out, the media and “experts” were routinely claiming that this virus had a fatality rate as high as 3-5%. So I don’t particularly care if they were off by a factor of 3 (~0.5%); they were much closer to correct than anyone else at the time.

This is just flatly wrong. You've conflated case fatality rates with infection fatality rates. The general consensus around the time of the Santa Clara study was that the IFR for the US was most likely in the 0.5% to 1% range. The 3-5% numbers were referring to case fatality rates.

In April, I wrote this - https://news.ycombinator.com/item?id=22969994 -

  "The NY study, if it holds up, suggests an IFR in the 0.8-1.0% range for NYC (depending on whether or not you include the additional excess deaths), which is in the range most experts have been assuming (0.5%-1.0% has been a common range that's been tossed around)."

You'll note April is also when the Santa Clara antibody study was released.

> they were much closer to correct than anyone else at the time

If the best science at the time of the Santa Clara study was estimating an IFR of 0.5-1%, and the Santa Clara study estimated 0.12-0.2%, and the current CDC estimate is 0.65%, who was closer to correct?

Given this, does that change your stance?

> Downvoting and brigading don’t change them.

Just to be clear, I did not downvote, nor brigade.

> No, it doesn’t. You’re cherry-picking old data, and ignoring the links I gave you, which tell you the updated numbers:

I think you may not have understood the numbers in your link.

> ages 0-19: .003% ages 20-49: .02% ages 50-69: .5% ages 70+: 5.4%

Calculate the overall IFR from these numbers given the age breakdown of the population of the US, and you'll end up a bit above 0.6%, which is what I said (when referencing 0.65%). Your link explicitly confirms what I said.

> The old reported numbers in New York are not a counterargument: we now know that number of cases was drastically under-counted, and the inferred IFR, if you use the right denominator, was a fraction of a percent.

You misunderstood this. When I referred to NYC boroughs and countries, I was referring to population fatality rates. Those are independent of case rates - it's a measure of what percentage of the entire population of an area has died, independent of infection spread. They help set baselines.

> You clearly know this, because you’re trying to talk about the rates in “boroughs” which are higher. But sample heterogeneity matters: I’m sure if you compute the IFR in Maimonedes ICU, it’s really high, too. The fact is, the fatality rate estimated by the Santa Clara study has been shown to be correct, and you’re simply spreading misinformation, because these new facts aren’t as scary.

That's fine. We can look at entire states. 0.18% of the entire population of New Jersey has died from Covid. 0.17% of the entire population of New York has died from Covid. And does anyone think nearly the entire population of NJ or NY has been infected already? No. Even in NYC, we've only seen a couple of boroughs that have shown around 50% antibody positives, with most in the 20-30% range. And, of course, incidence rates in the rest of New York state are lower than NYC. What does this mean? With 0.17% of the entire population of NY state dead from Covid, with incidence rates generally at 1/5 to 1/3 of NYC, with incidence rates lower in NY outside of NYC, the implied IFR for NY state is north of 0.5%.

With all that said.. you sourced a CDC resource that puts the IFR of Covid-19 above 0.6%. You used it to support a claim of Covid having a similar IFR to flu. Given that your source actually confirmed the IFR I mentioned, does that change your stance?

> 2 - First image in this collection: https://swprs.org/covid-the-big-picture-in-7-charts/

This site seems entirely untethered to reality? E.g. "2) Covid mortality vs. flu mortality" section appears to put IFR for covid of those at age 80 at 0.16%? Yet, we have entire countries that aren't far off from that number for population fatality rates, let alone infection fatality rate.

Wait.. I was just assuming this had to be an IFR comparison, because for something like this, anything else would be disingenuous. I should assume less. It looks like it is comparing annual mortality of flu with the current mortality of Covid-19.

1. We're still far from a year into notable spread of Covid-19. 2. Normal flu seasons don't have lockdowns, masks, bans on large events, etc. 3. In digging a bit more, this uses the 2017 flu season as a baseline, which was one of the worst.

This site is actually a great example of Covid-19 misinformation. It has a veneer of authority - Swiss Policy Research as a name sounds like it might be credible. And, it presents what, on first glance, appears to be claims grounded in science (and hey, nice graphs and all!).

But, one starts digging into the claims, and it starts to become clear the flaws.

jasonv: there are a lot of pseudoscientific resources out there with respect to Covid-19. This looks to be one of them. It's more essential than ever to vet what you're linking to these days.

> Another interesting question is, what is the economic impact of those high risk people not being able to leave their homes? What happens to the schools and workplaces where they normally serve?

Exactly. This is something I've pondered in thinking through some of these proposals - if one wanted to pursue these policies and actually protect high risk folks, that alone would likely have significant impacts.

E.g. KFF estimated that a quarter of workers are at high risk of severe illness with Covid - https://www.kff.org/coronavirus-covid-19/issue-brief/almost-... - of course, some of these folks will be in jobs amenable to work from home. But, for the rest of them? That's a lot of the workforce either a) temporarily leaving the workforce (oh hello, economic consequences) or b) staying in, and accepting the increased risk from community spread (oh hello, more death).

> That is a rather coy statement. They don't even give us a clue how that could actually work.

Yep, lots of hand waving in that one. It's hard to take this declaration with any seriousness given the lack of substance.

For sure. Massachusetts seems especially high as a percentage re nursing homes. I believe the national average is ~40%. So, absolutely, a priority should be to focus on nursing homes.

And that's also part of my point - nursing homes were known to be extremely high risk from the very beginning - remember that outbreak in Washington. Despite that early warning, we still, by and large, failed to protect these smaller, super high risk communities. If we can't even do that well, what chance do we have in protecting the much larger "high risk but not super high risk" populations?

> You skipped the part about how it was right. Current IFR estimates are on par with the flu:

For some reason, I assumed that once we got to hundreds of thousands of dead in the US, that people would move on from this.

Even the CDC puts the IFR around 0.65%. And, for example, you've got multiple boroughs in NYC with hundreds of deaths each that have exceeded a population fatality rate of 0.5%. You've got entire countries that have passed 0.1% population fatality rates. And, of course, none of these places have actually hit 100% of their population being infected.

> Moreover, Michael Ryan of the WHO estimates 750M infected globally, which would put the IFR at around 0.1%

This is only useful if you also estimate deaths, otherwise you are comparing confirmed deaths against estimated cases. Given the level of missed covid deaths, this only works to downplay the fatality rate.

> I live in Contra Costa County (east bay, SF area), with a lot of retired people. 1.1 million residents. We’ve had 224 deaths with COVID so far.

> Does that sound like a pandemic?

It sounds like a region that has handled the pandemic well.

> In the US, death rates are pretty flat (not fluctuating much with cases)

Daily deaths are tracking case rates (given, one is best off looking at a combo of case rates, positivity and total testing). Of course, deaths lag a fair bit.

> of the 198k deaths in the US, only 16K were younger than 54 years of age.

And, we're only around 10% total incidence rate in the US, based on more recent widespread antibody studies. 16K for under 54 is still a fairly large number given that we're only at ~10%. Add in 55-64 and we're at 42K deaths. Let's ramp up to an incidence rate of 50% - super basic napkin math and we'd be at over 200K deaths for the under 65 crowd - as in, entirely people of below retirement age. That's huge.

> Is it reasonable to shutter most of our economies for those numbers?

We haven't though. Not to downplay the economic impact, but most of the economy is still running.

> People are suffering economically and emotionally because of stress and isolation.

Yeah, this aspect sucks. A few comments:

* Let's imagine we lifted all restrictions. How much do you think that would improve the economy? There's at least some evidence that economies that have failed in mitigation, such that the virus runs pretty close to free, have had the sharpest GDP declines - https://ourworldindata.org/covid-health-economy - granted, it is hard to separate out factors, as countries that have seen huge spread often also will, at least temporarily, lock down more. But, my general comment here is that the virus itself is a significant portion of the economic impact.

* The federal government should have done more to keep people afloat during this period. This sort of crisis is exactly when we should have committed to sustained support, e.g. UBI until the pandemic was over.

* Beyond that, re isolation, most places in the country allow for gatherings, for you to visit family and friends, for you to go out and about. Sure, big events and bars are still restricted in most places, but if one wants to socialize, they still can.

This seems incredibly hand-wavey in actually accomplishing "focused protection."

> Adopting measures to protect the vulnerable should be the central aim of public health responses to COVID-19. By way of example, nursing homes should use staff with acquired immunity and perform frequent PCR testing of other staff and all visitors. Staff rotation should be minimized. Retired people living at home should have groceries and other essentials delivered to their home. When possible, they should meet family members outside rather than inside. A comprehensive and detailed list of measures, including approaches to multi-generational households, can be implemented, and is well within the scope and capability of public health professionals.

The fundamental problem is that pretty much every country that has not been able to contain general community spread has had a hard time attempting to protect even just the super high risk populations (e.g. nursing homes).

But, in a country like the US, the high risk population is much, much larger than the nursing home population. Age or health conditions means that probably at least 100 million people in the country would be high risk. I've still not seen any comprehensive proposal for how you can protect the high risk populations while letting the virus run free in the low risk populations. In general, we're way too interconnected and aren't willing to invest the (likely massive) resources to even attempt this. Think of how many high risk people still have to work in non-WFH jobs. Think how many multi-family households there are. Think of how many low risk people live with high risk people. Think of how many parents of school kids and teens are high risk.

The more general community spread there is, the more it impacts high risk folks who don't have the luxury of fully isolating from community risk.

Finally, it doesn't help that Jay Bhattacharya is one of the authors - a Hoover Institution associated fellow who's pretty infamous in the epi community for the fundamentally flawed Santa Clara antibody study - significant statistical issues, a highly flawed recruitment strategy, a side story of his wife recruiting people through a non-approved side channel, questionable funding sources like the founder of JetBlue (with records that showed said founder might have influenced the research), and, and, and. Many of these flaws biased the study into showing a high infection incidence rate, and thus a much lower fatality rate, in the same ballpark as the flu. Given that he's continued to downplay the seriousness ever since then, his name being attached to any sort of declaration should give one at least pause.

Florida as a whole enacted mask recommendations in late June.

Miami Dade county enacted a mandatory mask mandate on July 2nd: https://www.miamidade.gov/global/initiatives/coronavirus/eme...

Palm Beach county enacted a mandatory mask mandate on June 25th: https://discover.pbcgov.org/PDF/COVID19/PBC-EO-12.pdf

Hillsborough county enacted a mandatory mask mandate on June 22nd (if I'm reading it right): https://web.archive.org/web/20200706143937/https://www.hills...

Broward county enacted a mandatory mask mandate on July 8th: https://www.broward.org/CoronaVirus/Documents/EmergencyOrder...

Orange county enacted a mandatory mask mandate on June 20th: https://www.orangecountyfl.net/portals/0/library/Emergency-S...

Pinellas county enacted a mandatory mask mandate on June 24th: https://covid19.pinellascounty.org/wp-content/uploads/2020/0...

In just this list, we're up to mask mandates that cover counties that account for 10 million people, 47% of Florida's population, and accounting for the densest population centers. A number of additional counties also enacted mask mandates, but it got a bit tedious listing them all out.

In other words, in the June-July period, mask mandates were enacted for a majority of the population of Florida.

In another comment, I critiqued this piece as being a shallow analysis. Your comment here is an example of that.

Not the original commenter, but no, I'd hope not, in its current state. It's a shallow analysis, and it does a poor job of analyzing the data to try to understand causality in explaining case trends.

For example:

> Why would Floridas numbers start to decrease if not for herd immunity? We know they aren’t being more careful. Disney land reopened and there is a police sheriff who literally made a rule that police officers and people going into the station were not allowed to wear masks.

As I mentioned in another comment:

> For example, bars opened in early June in Florida. They closed on June 26th after cases started spiking.

To miss a policy change this significant in such a piece is pretty egregious if you are claiming that this is publishable quality.

I'll assume this was a good faith effort, but the piece has many flaws and draws conclusions that aren't well supported.

> They aren't being more careful now than they were in early June.

Source? This is a significant claim. Early June was a low point in cases, hospitalizations and deaths. Lots of states were at peak reopening around then, as things seemed to be on the right track.

For example, bars opened in early June in Florida. They closed on June 26th after cases started spiking.

Louisiana was similar. The state started reopening May 15th. Later in June, bars were then closed back down again, and a mask mandate went into effect.

In Arizona, the story is similar, with gyms, bars and movie theaters shutting back down in July after having opened in May.

What sure seems to be the pattern is: things open up, cases rise, things close back down, cases drop. I'm not so sure that your "they aren't being more careful now than they were in early June" statement is accurate, and I think this is causing you to misattribute the cause of the drops in cases.

That dataset is super useful and interesting, but it has a huge caveat, assuming we're talking about the same dataset (https://www.nytimes.com/2020/08/19/nyregion/new-york-city-an...).

Oh, I see where the 68% comes from - https://www.nytimes.com/2020/07/09/nyregion/nyc-coronavirus-...

That's from one clinic, for people seeking an antibody test.

Anyway, the first link I mention is the broader dataset from all the clinics like this, so it helps smooth out anomalies. But, the big caveat in all of this is that these suffer from selection bias - it's not a randomized sampling of people in each borough. It's still useful data, but it's far from being able to tell one the incidence rate to a reasonable degree of precision. Still, many reasonable inferences one can make.

I think using herd immunity to apply to temporary network states is going to be more confusing than helpful. There will be some fuzziness at the edges - some of the changes and adaptations we make during this pandemic will probably persist longer term.

The fundamental issue I see in the piece is a conflation of what herd immunity means. Normally, it's used in the context of normal activity levels - as in, if everyone went back to normal pre-pandemic routines, would we see cases go up or down? But, you're using it in the context of current activity levels, which are still significantly different than pre-pandemic activity levels.

All you are effectively saying is that, given the current mitigation measures and given the current level of population immunity, R is less than 1 in these places. But, that's not really a revelation.

What one would expect is likely to be the case is that some areas have had enough exposure that it has a non-zero impact on their current R, as in NY and NJ. This means there may be more buffer to relax mitigation measures without seeing R rise above 1, but it doesn't necessarily mean they are at herd immunity.

Great read. And scary, and frustrating.

What might be entirely coincidence, but I found interesting, was that this was a JetBlue flight from JFK. The one time I triggered the explosives detection machine was a JetBlue flight from JFK. I had almost nothing special - a carryon I've used many times before, with clothes and toiletries. I also had two small fossils I bought from the Evolution store in NYC, which the TSA agent seemed especially interested in retesting.

In my case though, I just had to deal with a thorough search of the carryon, confusion from other TSA agents (something like "oh, you searched the carryon? That wasn't necessary" from one agent to another) and listening to the main TSA agent wax poetic about how evolution is a lie and how climate change is a hoax.

It's also possible many of these colleagues had concerns, but weren't previously airing them publicly.

Just searching academics I follow, the one thread that popped up from a year ago was this https://twitter.com/Graham_Coop/status/1132105378668326912

In which it seems many of these criticisms were well known at least a year ago. I hadn't previously followed this, but it seems like this was probably a case of the dam finally bursting.

Nobody likes that these protests are likely to lead to increased transmission. It's a truly terrible choice. People are making that choice because these protests get to such a fundamental problem in our society that they are still worth it, despite the cost. And, as has been mentioned, mask usage is high, and these protests are outside. It doesn't eliminate the risk, but many, many people are still trying to minimize the public health risk.

So, yes, context matters. I would absolutely argue the value in the being part of the strongest push against police brutality in decades, that just might result in systemic change, is incredibly more worthy than virtually any other activity. It does not mean there is no public health cost. It does mean it is a tragic choice to have to make.

I think your broad point is accurate. It's going to be a spectrum. Many people will want that social interaction in some form that is not just virtual, but the form it takes will vary a lot.

Sure, some people will go right back to indoor dining, bars, nightclubs, whatever is allowed to open back up. But, a lot of people aren't going to go back to pre-Covid behaviors just because things are open. Many will seek to resume social interaction, but in safer ways, and your example of the pub is a good one. Time for more beer gardens!

Just thinking about myself, I probably won't dine inside a restaurant for a long while. But outdoor dining set up in a reasonably safe way? That's possible.

For better or worse, we're moving into reopening, so there should be a big focus on harm reduction: https://www.theatlantic.com/ideas/archive/2020/05/quarantine...

> What does harm reduction look like for the coronavirus? First, policy makers and health experts can help the public differentiate between lower-risk and higher-risk activities; these authorities can also offer support for the lower-risk ones when sustained abstinence isn’t an option. Scientists still have a lot to learn about this new virus, but early epidemiological studies suggest that not all activities or settings confer an equal risk for coronavirus transmission. Enclosed and crowded settings, especially with prolonged and close contact, have the highest risk of transmission, while casual interaction in outdoor settings seems to be much lower risk. A sustainable anti-coronavirus strategy would still advise against house parties. But it could also involve redesigning outdoor and indoor spaces to reduce crowding, increase ventilation, and promote physical distancing, thereby allowing people to live their lives while mitigating—but not eliminating—risk.

> Ascribing all excess deaths to coronavirus is a mistake.

I worded my statement intentionally.

> It's also most likely still an undercount

Some of the excess deaths will be other things. It will take more careful analysis to come up with reasonable breakdowns, and what percentage is likely covid-19 deaths.

Basically, as this article mentions - https://www.washingtonpost.com/investigations/2020/05/02/exc...

> “I think people need to be aware that the data they’re seeing on deaths is very incomplete,” said Dan Weinberger, a Yale professor of epidemiology who led the analysis for The Post.

> Those excess deaths — the number beyond what would normally be expected for that time of year — are not necessarily attributable directly to covid-19, the disease caused by the coronavirus. They could include people with unrelated maladies who avoided hospitals for fear of being exposed or who couldn’t get the care they needed from overwhelmed health systems, as well as some number of deaths that are part of the ordinary variation in the death rate. The number is affected by increases or decreases in other categories of deaths, such as traffic fatalities and homicides.

> But excess deaths are a starting point for scientists to assess the overall impact of the pandemic.

The 2009 H1N1 Pandemic killed in the ballpark of 10k people in the US. I don't recall ever seeing mainstream news outlets reporting in the ballpark of 100k recorded deaths.

It's been 7 weeks since the Imperial College model that predicted up to 2.2 million deaths in the US, that used an infection fatality rate of 0.9% as an assumption. The best science still has us in the 0.5% to 1% range for IFR, still putting us in the 1-2 million range if let ourselves get to herd immunity levels without a vaccine.

I think we'll avoid getting to 1-2 million, but it'll be because we take actions along the way to push some/most infections to post-vaccine or post-effective treatments.

> Try to reach it without a vaccine, and millions will die.

https://www.nytimes.com/2020/05/01/opinion/sunday/coronaviru...

1) The healthcare system barely handled the load, and that was with heroic efforts to reposition ventilators and build up temporary capacity. We barely handled it because most of the country was not nearly as hard hit, so we could still shift resources around. If everywhere was hit this hard at the same time, that would be a different story. It would need to be spread out more than you imply if we wanted to have all of the country trying to stay just below capacity.

2) New York City has lost .2% of their entire population to get to that ~20% number. Extrapolated out, that's once again back to the 1-2 million deaths range for the country to hit herd immunity.

That's not how that works.

It's also most likely still an undercount. Confirmed plus probable cases are thousands short of total excess deaths for that period for New York City.

> Further, any variance of susceptibility or contact density means you don't need to reach 60% for herd immunity. Some estimates say that the real threshold may likely be 25% or less: e.g. https://www.medrxiv.org/content/10.1101/2020.04.27.20081893v.... Further, lighter weight social distancing measures can further reduce Rt and attain herd immunity at a lower percentage.

Not surprisingly, that paper has had some discussion: https://twitter.com/CT_Bergstrom/status/1257452758376091648

> It is true that if you could somehow pick and choose who became immune — the way you can in principle with vaccination — you might be able to do pretty well, vaccinating the high-contact individuals only and reaching herd immunity at low levels of vaccination.

> But natural infections don't spread like this. High-contact individuals are connected to low contact individuals. An epidemic spreads organically through the population, infecting who it happens to, not who you want it to.

> IF high contact individuals were only connected to other high contact individuals with high probability, you still might make some headway with nature disease progression. This sort of thing can matter for STD transmission.

> Respiratory viruses tend to be different. You have household transmission, cohort-based (school/work) transmission, and incidental transmission (subway/supermarket). These render the transmission network quite well-connected.

Additionally, if we're talking natural infection to get to herd immunity, you have the issue of overshoot, so unless you're very precise on measures to tamp down on transmission before you get to herd immunity levels, you easily overshoot the herd immunity threshold by quite a bit.

Edit: more Carl Bergstrom commentary on IHME:

> (That's not to let IHME off the hook; once they started promoting the ability for their model to do anything else, they became responsible for its predictions at other stages of the pandemic.)

https://twitter.com/CT_Bergstrom/status/1255350829382602754

Also, the model he references in the thread - https://twitter.com/CT_Bergstrom/status/1255343846445195266 - seems a lot more plausible for predicting the down-slope.