He’s one of the few people who I consistently follow regarding COVID-19 reporting and his methods are typically very sound. He’s also always honest about the limitations of the data we currently have.
The most interesting thing to me is the Herd Immunity Threshold portion of his paper, which I hadn’t really thought through prior to reading this. Even a true prevalence (we don’t actually know this value, but he attempts to make good guesses) of 20% could have a huge impact if the 30% of the people that aren’t socially distancing are made up substantially of that population. He also cautions that simply opening the floodgates again could quickly revert the threshold back to normal.
I’m going to read it again a few times and digest it all, but I definitely appreciate his approach.
I need to read it again to see if I missed it, but I’d like to see a discussion of what reasons there would be – apart from just the obvious of who’s getting tested at what time – for for the virus to spread at different rates in different age groups.
I could imagine it being plausible that the elderly could have started ‘hunkering down’ after the early stages, but what (behavioral) explanation could there be for young people to have been less likely to be infected in the early stages and more likely later?
Well, I can confirm that the 60+ folks I know are quite definitely ‘hunkered down’ more than in March-April. The high incidence in nursing homes also skewed early numbers older. Also, it seems likely that younger cohorts were infected in the early stages but testing limitations and less serious disease meant that they were missed - but still infected.
Well, right, that’s what I’m getting at. I don’t think such a thing is included in this model. But if it were true, it seems that it could have big implications for the ‘bottom line’ sort of results he’s trying to arrive at.
One thing he neglected in his paper, but pointed out in a Tweet, was the higher IFR in older age groups.
I have been estimating a crude IFR by age in my state in a spreadsheet. By using his .25% estimate, the IFR looks like it could be .7% for 60-79 and 2.5% for 80+.
Of course, there can be selection bias on how different age groups are getting tested. One can argue that the reason there is a higher proportion of older people in March/April relative to June/July is because testing was limited, and hence older individuals were prioritized for testing. So one would expect that older age groups have a lower positivity rate than the younger age groups (since you are catching more cases). But if you look at the data, the opposite is true: in March/April, the older age groups actually had a higher positivity rate than the younger age groups. By our prevalence ratio calculation above, this indicates that the prevalence is actually even higher in the older age groups than the younger age groups. This trend was reversed starting in late April, and now younger age groups have a higher positivity rate than older age groups.
We can use our prevalence ratio formula from above to estimate the proportion of true infections by age group given the number of confirmed cases and test positivity rates:
In regard to the bolded above, I guess it’s the connection with the positive rates that bakes it in, but it still seems to be as though the analysis is anchored on the ‘number of confirmed cases’, which presumably would be more highly underestimated in the younger age group.
Epidemiology makes my head hurt (perhaps that is why I abandoned my Public Health PhD program for software engineering…)
Confirmed cases and test positivity by age groups still gets skewed by the percentage of each group being tested. So until we can test much more widely than we do now, even the fancy estimations will be guesses.
Still, I can imagine that test positivity by age could have reversed more recently simply because it is easier to get tested. Older individuals can get tests as a caution, even without major symptoms (lowering positivity) while younger individuals (who might not seek testing without symptoms) are able to get tested when symptomatic (raising positivity).
(In March, when two family members tested positive, it was very hard to get tested. The first, age 69, had all the symptoms but couldn’t get tested until he mentioned a 90+ relative in the same house.)
The ‘earlier’ part is pre-lockdown, with conditions even freer than when ‘states mostly opened back up’. What you’ve pointed out is exactly the reason why one would expect to see the opposite of what the data he is using shows.
I like the effort to come up with some sort of way to quantify the interactions between positivity rate and reported infection rate on true infection rate, but I don’t get where he comes up with what is the true prevalence rate to form fit into as he says he did here:
So you meant why were there less young adult cases back in March? When there were only 4,000 US cases before the lockdown vs now when there are 5,000,000 cases?
Any asymptomatic person who has not been tested is “an unreported case” because no one knows the person is positive for COVID-19, and the majority of people in this nation have not been tested. Even people who were tested can contract the disease afterwards and be asymptomatic and unreported. Given those circumstances, it is easy for me to see that we can have a much larger number than is officially reported.
You’re getting around to it, yes. Why – apart from just the number of confirmed cases being so comparatively low, as you point out – would older people have been more likely to be infected in the early stages than young people were? Why any differences at all, and assuming there are indeed differences, why would they run the opposite way to what you would expect? Meaning, younger people would seem more likely to out doing things that interact with more people, if we can make broad generalizations.
The earliest US big outbreaks were in nursing homes, first in WA and then NY, and the average age of nursing home staffers is 36, so young people weren’t typically exposed early on.
“To you” who has what training and education on the subject of immunology and communicable diseases?
The “endgame” is to reduce the number of infected as long as possible in order to “flatten the curve” and not overwhelm local health care systems.
Sorry, but what I see is a lot of “wishful thinking” that all the experts at the CDC, Johns Hopkins, etc, are wrong or somehow missed some nuance that a high school educated idiot gleaned online.
“Reduce infections for as long as possible” is the opposite of what “endgame” means. Sorry. Most western countries are in the next stage of “how quick can we open”.
Well the “endgame” is “life goes back to normal”. Or at least some sustainable “new normal”. I don’t really see that happening until we discover a vaccine or scientific minds come up with a more definitive definition of what “herd immunity looks like”. Because right now the “endgame” looks a lot like “let’s open everything up and hope everyone already has ‘herd immunity’”.
First note the difference between your proposed reduction, flattening the curve to not overwhelm systems, and the goal of suppression to near zero. I’d add to that flattening the curve also attempting to most avoid infections among the highest risk most vulnerable and to be very careful to avoid having a peak timed with flu demands.
The former has an implied endgame strategy: as the population with at least some degree of protection grows the severity of social distancing measure required to keep the effective R low decreases. Allow it to grow in a manner that causes the least possible harms and that does not come close to overwhelming systems, methodically, not pell mell, keeping effective R low by the feeling along moving edge of fraction now immune and gradually relaxing controls, listening to what the disease tells us as we go.
The latter is praying for an effective enough safe enough vaccine that is widely available and widely accepted to provide the exit deus ex machina. Or more cynically knowing that cooperation with suppression level activities will eventually fail and that a horrible flare might occur at the worst possible time. (Fauci now is setting expectations that a very effective vaccine may be unlikely.)
True fraction already infected, degree and durability of protection associate with antibody positivity and with T cell reactivity without antibody positivity, magnitude of protection from pre-exisiting cross-reactive T-cells, true degree that children under 11 function (or do not) as relative dead ends for transmission, all can help inform best guesses, but whatever the guesses the former still requires the cautious moving forward.
Well, I’ll certainly buy that the deaths were there, yes. But I find it really hard to imagine a scenario where the virus picked and chose, as it were, its hosts. I find it much easier to believe that in the early stages it was spreading exactly the same as it was in any later stages, when it comes to age groups. And younger people simply not being tested to the same extent as older people is, I find, a perfectly reasonable explanation.