Re: Herd-ish brakes on outbreaks?

1

Where are you seeing the trend down in cases? The most recent few days of cases will always be a slight decrease from reality due to a lag in reporting. There might also be a holiday lag in reporting that makes the last week seem to be trending down that is not real.


Posted by: Bass | Link to this comment | 11-30-20 7:56 AM
horizontal rule
2

The NYT? Maybe so - you're saying fewer people got tested over the past 8-9 days due to the holiday?

That's a much less complicated theory than mine.


Posted by: heebie-geebie | Link to this comment | 11-30-20 8:00 AM
horizontal rule
3

Either / both of less people got tested and it takes longer to report results. Probably more so the latter. A lot of people were getting tested in advance of traveling for the holiday. But testing results aren't available instantly.


Posted by: Bass | Link to this comment | 11-30-20 8:07 AM
horizontal rule
4

You want to be cautious in interpreting the most recent case reporting. Here is an explanation https://www.cdc.gov/foodsafety/outbreaks/investigating-outbreaks/epi-curves.html


Posted by: Bass | Link to this comment | 11-30-20 8:10 AM
horizontal rule
5

This was Drum's comment:

Most of you know this already, but you should ignore the small downward dip in the US data. It's just an artificial reporting gap due to the Thanksgiving weekend. It will return to normal next week.


Posted by: ogged | Link to this comment | 11-30-20 8:10 AM
horizontal rule
6

I hope Drum's ok. He's blogged about very little other than photos and cats lately.


Posted by: One of Many | Link to this comment | 11-30-20 8:16 AM
horizontal rule
7

In my state they seem be be developing a backlog in their ability to report cases on a given day.


Posted by: Spike | Link to this comment | 11-30-20 8:23 AM
horizontal rule
8

5 was how I interpreted it. Almost every US state I checked had a weird decrease at the same time as the holiday weekend for both cases and deaths, which usually lags. On the other hand, UK deaths have been more-or-less constant while cases have fallen significantly. Throughout the pandemic the regions of the US have trended very differently from each other, so if all the regions are moving in unison my presumption is there's some common cause, and the holiday is right there.


Posted by: dalriata | Link to this comment | 11-30-20 8:30 AM
horizontal rule
9

Ok, so even if my theory doesn't apply here, it could still apply at other times?

OR

My theory could apply when rates reach 70%, and that's why the clever epidemiologists have arrived at that number as the lowest point at which you start to see some protection.

OR

my theory could be horseshit.


Posted by: heebie-geebie | Link to this comment | 11-30-20 8:33 AM
horizontal rule
10

Our 7 day averaged positive rate started decreasing on the 20th, so a few days before the holiday. I'm not sure if that means anything.


Posted by: AcademicLurker | Link to this comment | 11-30-20 8:35 AM
horizontal rule
11

In California, both Sundays and Mondays are typically about 70% of the weekly average. I'd be surprised if Thanksgiving didn't create a similar reporting effect.


Posted by: Minivet | Link to this comment | 11-30-20 8:59 AM
horizontal rule
12

After our big Halloween outbreak, we had a statewide mask mandate and no-gathering order, which lasted until just before Thanksgiving (so we're due for our next spike in about four days.) So I think some behaviors have changed a little.

But it's also the case that not everyone spreads the virus equally well. Something I read this morning that I can't find said that 71% of people don't really spread it at all, and 8% of people spread it to a lot. So it would seem that if there is an outbreak, it would naturally spread to some people who won't pass it on, and if containment measures can stop some of those 8% from spreading it to a lot of people, the rates will fall back.


Posted by: Cala | Link to this comment | 11-30-20 9:02 AM
horizontal rule
13

8% of people spread it to a lot

Fucking extroverts.


Posted by: ogged | Link to this comment | 11-30-20 9:04 AM
horizontal rule
14

Extroverts fucking.


Posted by: heebie-geebie | Link to this comment | 11-30-20 9:09 AM
horizontal rule
15

Don't twist my words.


Posted by: ogged | Link to this comment | 11-30-20 9:11 AM
horizontal rule
16

You're just mad that you're missing out.


Posted by: heebie-geebie | Link to this comment | 11-30-20 9:15 AM
horizontal rule
17

It makes me grumpy that my state's daily dashboard always looks like things are trending downwards in the past 3-4 days, but only because of data that hasn't arrived yet. It's so consistently wrong that I think it shouldn't be included in the visualizations, or highlighted as being incomplete data, or something.


Posted by: Nathan Williams | Link to this comment | 11-30-20 9:31 AM
horizontal rule
18

It sounds like something that would be easy to model.


Posted by: Moby Hick | Link to this comment | 11-30-20 9:38 AM
horizontal rule
19

This might be about 80% of what you're after:

https://mobile.twitter.com/AdamJKucharski/status/1332307259184074752


Posted by: Alex | Link to this comment | 11-30-20 9:58 AM
horizontal rule
20

Is there a ceiling to testing capacity? A bit like the geiger counters at Chernobyl only giving readings up to 4 (or whatever).


Posted by: Charlie W | Link to this comment | 11-30-20 10:52 AM
horizontal rule
21

On the other hand, UK deaths have been more-or-less constant while cases have fallen significantly.

I have been tracking this fairly obsessively and the metric I am using is the average week-on-week change (to get past problems about lower reporting levels on the weekend). This dropped below 1 for cases on 16 November, and it appears to have dropped below 1 for deaths on 25 November. This implies that deaths are tracking cases at a distance of about 9 days - this is also what we saw in the spring. Case and death data from 27 November onwards is too recent to be stable - the figures will go up as more cases sampled on that day are processed and reported.

Cases have now been dropping at a weekly rate of around 0.72 since 18 November. The figure for 26 November was 0.73 of the figure for 19 November; I anticipate this relationship will hold until at least 11 December.


Posted by: ajay | Link to this comment | 11-30-20 11:19 AM
horizontal rule
22

I don't see how the Heebie Model is different from herd immunity. In particular, how it the HM would produce a series of spikes and retreats, particularly on the nearly globally coordinated scale we've seen (well, I just follow what happens on Kevin Drum's [and I second comment 6] tallies of deaths, so North-America-north-of-Mexico-And-Selected-European-Countries scale, but the wave we're in the middle of is pretty darn broad). It seems to me that a bunch of small fires that fizzled out would produce a gradual slowing of transmission, viewed at large scales, rather than waves. And apparently there is no new more virulent or transmissable strain out there. So I have settled on behavioral changes as the best (though still unsatisfying) explanation.


Posted by: | Link to this comment | 11-30-20 12:04 PM
horizontal rule
23

dammit, me.


Posted by: chill | Link to this comment | 11-30-20 12:04 PM
horizontal rule
24

21: I've been using DIVOC-91's weekly moving average (on a per-capita basis for comparison), which appears to have slightly different numbers than you; this is my saved link for US/UK comparisons. For deaths I'm not seeing a weekly drop until the 29th's data, which as you say is a little suspect. The official government website also is saying there's an increase in deaths in the last 7 days. I'm curious why our numbers are different, but either way it's close to 1 and falling.


Posted by: dalriata | Link to this comment | 11-30-20 1:05 PM
horizontal rule
25

My state still seems to be doing a great job testing. I had a shallow nasal swab 11/23 at a pharmacy (waited in the drive-through line for six hours, results came back negative on 11/28) and a saliva test on 11/24 at a municipal building (waited ten minutes in the drive-through line, results came back negative on 11/27). Neither charged me anything or asked for insurance, though you can bet I'll continue doing the municipal testing from now on. When it's not the week of Thanksgiving, results have taken 24-48 hours.


Posted by: | Link to this comment | 11-30-20 4:35 PM
horizontal rule
26

One alternate hypothesis is that there was a big wave as a result of the election, and election-related social gatherings, which started petering out around Thanksgiving.


Posted by: Spike | Link to this comment | 11-30-20 6:36 PM
horizontal rule
27

The election and Halloween.


Posted by: Moby Hick | Link to this comment | 11-30-20 7:38 PM
horizontal rule
28

Also, schools being open.


Posted by: Spike | Link to this comment | 11-30-20 7:50 PM
horizontal rule
29

I've read that isn't usually a problem, at least for little kids.


Posted by: Moby Hick | Link to this comment | 11-30-20 8:28 PM
horizontal rule
30

If herd immunity occurs at 70% then I would expect that the curve of the total number of cases would look something like: .7(1-(exp(-t/T)). Where T is somehow related to the rate at which the disease spreads.


Posted by: jeepers | Link to this comment | 11-30-20 11:19 PM
horizontal rule
31

I should have included some assumptions in my previous comment. I think you would see something like the behavior I described if there was no active intervention and there was enough mixing of the population to avoid creating insulated clusters of people. In the real world it's not a nice smooth curve obviously. But I think that as the population approaches herd immunity the curve would start to asymptotically approach 70%.


Posted by: jeepers | Link to this comment | 11-30-20 11:27 PM
horizontal rule
32

I've read that isn't usually a problem, at least for little kids.

We just had three cases in a local elementary school. I suspect the "kids don't pass it around" line is in large part wishful thinking on the part of people who are going nuts having their kids at home all day.

For example, there was recently an article in the Washington Post lauding Rhode Island for keeping its schools open, in contrast to schools in the DC area where most things are remote. What it didn't mention was that Rhode Island's current outbreak appears to be substantially worse than DC's.


Posted by: Spike | Link to this comment | 12- 1-20 10:00 AM
horizontal rule
33

I think if the bars and restaurants were shut for long enough to get the cases down in the community, it should be saf-ish with testing/screening.


Posted by: Moby Hick | Link to this comment | 12- 1-20 10:51 AM
horizontal rule
34

Yeah, I think so too. But what happens in practice is the bars and restaurants don't shut down and the schools get opened anyway. If there could be a long-term commitment to keep the bars closed (and the bartenders paid) it could be made to work, but I don't think we live in a country where that is going to happen.


Posted by: Spike | Link to this comment | 12- 1-20 12:21 PM
horizontal rule
35

Things work better on islands like New Zealand and Rhode Island.


Posted by: Moby Hick | Link to this comment | 12- 1-20 1:01 PM
horizontal rule
36

Boris Johnson's long-repeated joke about the mayor being the hero of Jaws, for rejecting "silly regulations on health and safety", where he declares it was good for the majority even if "some small children were eaten by a shark", ghoulish from the start, is especially terrifying looking back now.


Posted by: Minivet | Link to this comment | 12- 1-20 1:58 PM
horizontal rule
37

I think the evidence that elementary school aged kids are less likely to get and spread COVID is backed up pretty well. Doesn't mean they don't spread it at all, and doesn't seem to apply to older kids (maybe including older elementary school).

The speculation that this is because they've had the mumps vaccine more recently is interesting.


Posted by: Unfoggetarian: “Pause endlessly, then go in.” (9) | Link to this comment | 12- 1-20 2:01 PM
horizontal rule
38

The mayor in Jaws was reëlected. Trump wasn't.


Posted by: Moby Hick | Link to this comment | 12- 1-20 2:06 PM
horizontal rule
39

38: More Proof that this Election was FRAUD!


Posted by: Opinionated Trump | Link to this comment | 12- 1-20 2:16 PM
horizontal rule
40

The speculation that this is because they've had the mumps vaccine more recently is interesting.

I speculate that its because kids are short and therefore aerosols and droplets from them aren't as likely to travel very far.

Evidence from India suggests that children can be spreaders because they are more difficult to recognize as carrying the disease.

Worst of all, the proportion of children who are infected but don't show any symptoms, he says, is higher than adults. So they may not even be recognized as potential carriers.
"[Many] kids are silent spreaders in the sense that they don't manifest the disease with symptoms," Laxminarayan says. "They happen to get infected as much as anyone else, and then they happen to spread it to other people."

Of course, that's India but I think the broader point is that the degree to which kids are spreaders is highly dependent on context. Can America sustain the kind of context where groupings of children do not cause broader viral spread through family networks? I'm not particularly confident that it really can.


Posted by: Spike | Link to this comment | 12- 1-20 3:05 PM
horizontal rule
41

I speculate that its because kids are short and therefore aerosols and droplets from them aren't as likely to travel very far.

This seems hard to credit - surely the vector that would lead to the most community transmission would be kid-to-kid outside the home, followed by kid-to-adult inside.


Posted by: Minivet | Link to this comment | 12- 1-20 3:59 PM
horizontal rule
42

That's not going to happen if the nanny was already infected over the summer because she's also a waitress on weekends.


Posted by: Moby Hick | Link to this comment | 12- 1-20 4:02 PM
horizontal rule
43

"kids don't pass it around" is vague. The rule of thumb people seem to be going by is 15 minutes of close contact indoors between adults results in a high chance of transmission if one of them has it. OK, so maybe similar contact with a kid would result in only a medium chance of transmission, or even a low chance. But who spends 15 minutes at a time with kids? If you're around kids at all, you're spending 7 hours with a dozen of them at a time (a dozen, hah, that would be amazingly small class size) in rooms with bad ventilation. And if you're lucky, half of them cover their mouths half the time they cough. Kids might not pass it around during 15-minute interactions but schools could still be petri dishes.


Posted by: Cyrus | Link to this comment | 12- 1-20 5:58 PM
horizontal rule
44

But who spends 15 minutes at a time with kids?

Most of them get really repetitive after like five minutes. But there's pretty good evidence that the spread at schools, at least for the younger kids that you would expect to be licking all the door knobs, is pretty low.


Posted by: Moby Hick | Link to this comment | 12- 1-20 6:04 PM
horizontal rule
45

Whatever the pause in US cases and deaths was due to, it's over now. That it was both cases and deaths synchronously points to it being due to mismeasurement, not an actual decrease.


Posted by: dalriata | Link to this comment | 12- 4-20 3:57 AM
horizontal rule
46

I'm a bit shocked to learn that Gisgard d'Estaing was alive long enough to die from covid.


Posted by: Moby Hick | Link to this comment | 12- 4-20 6:31 AM
horizontal rule