Is there less testing? Is the positivity rate declining too?

Maybe people are changing their own behavior in response to the earlier rising cases without the need for mask orders.

I thought they closed bars in many places. They did here. I think bars were a huge part of the problem.

At a glance, the decline in new cases seems to match the decline in testing so it may not be receding as much as it not being detected.

https://covidtracking.com/data/state/florida

1: Florida's positivity rate is limping downwards. (Their testing has also decreased.)

Is that enough to say that the epidemic hasn't actually waned there?

4 If the positivity rate is high that could well be the case.

3: Maybe that's it. I think that made a huge difference here, and it does appear that bars are closed in Florida.

Bars, plus a decline in testing: at what point do case numbers stabilize?

Hopefully we're not about to reopen any major institutions on a massive scale over the next few weeks.

When fifty to sixty percent of Americans have been exposed.

We're doing the societal equivalent of when Sideshow Bob keeps stepping on rakes.

Heh.

Deaths haven't really come down (yet).

I'm a little unclear on what "Deaths by Case Date" means. Isn't that necessarily going to show a graph with falling numbers, given the delay between (what I assume to be) the case date and the death? I mean, those numbers seem like they only become meaningful a month or more into the past.

But truly: are you guys saying the appearance that cases have peaked is an illusion? Or are we saying that closing bars and sporadic mask orders is enough to keep daily numbers from increasing?

I'm not sure. Utah's test rate has plummeted but the percentage positive rate is holding steady. I've heard that some people are refusing to get tested, but that's at the level of anecdote. The criteria to get tested tightened up again compared to earlier this summer, so that's probably contributing, too. Testing centers are shutting down early due to the insane heat, too.

I'm also wondering if there are just more mild nearly asymptomatic cases, and no flu-like symptoms to rule out so people aren't bothering to go on.

And I'm talking about inferring true infection rates. As in, the number in reality that we don't have access to, which we guess at with case numbers and positivity rates.

If Florida's epidemic hasn't actually waned, then they're juking the stats pretty widely. State stats (floridahealthcovid19.gov) have positivity rates coming down from a peak c. 6 weeks ago, but also fewer ER visits that include reports of cough, fever, and shortness of breath, and less flu-like and COVID-like illness. I also don't really know why. A bad equilibrium in most of the state combined with pockets of near-herd-immunity in Miami?

Our positive rate continues to go down as test volume increases (currently at 3.2%). I don't believe there is a sate wide mask order in Maryland, but every business I've seen is requiring masks to come inside.

I agree that (re)closing the bars was probably a big factor.

I really miss bars. I mean, I was at an outdoor one on Tuesday, but it wasn't the same.

California is finally seeing some real drop-off after our second peak. The Central Valley is still far and away the highest affected region with 50 cases/day/100,000 (also highest deaths), most of the metropolitan regions fall between 17 and 23, counties outside major metro areas are 11, and San Diego is 10, surprisingly.

But we're only a little down from peak. The first peak was 25.1 statewide on July 24, which troughed to 16 on Aug. 5, second peak was 24.4 last Friday Aug. 14, most recent day I have is 20.9 for Aug. 17. So I agree, we have no idea what's driving the changes and there could for all we know be a similar third wave. Positivity rate is around 6%.

The California numbers are complicated by the malfunctions in the state registry. For a while (it's not clear exactly what period -- end of July to early August maybe?) the registry significantly undercounted cases. This dip was followed by a steep bump as the state attempted to catch up. I think there were about 300,000 cases total involved in the undercount-and-catch-up. So, to some extent the pattern described in 19.2 is due to the registry problem.

Closing bars, together with people in hotspots dining out less as a result of COVID being in the news more. Also we may be past family reunion season, which drove a lot of outbreaks.

Insofar as I saw anyone indoors, which wasn't often, masking in Palm Beach County was universal when I visited. I'm sure there will be a school bump, though.

I don't know, but I suspect part of it is that transmission within a household is a big driver, so the increasing caseload was one sick person infecting their household of (let's say) 4. The folks who are going to get it early (essential workers, doctors, etc) are going to pass it to their household, then the chain mostly is broken. After that, you'll see something more like regular circulation around a population with a roughly fixed rate that depends on the local regulations.

I hear anecdotally that leisure travel has been up the past couple of weeks aa folks try to get a summer vacation in before school starts. I'm worried about Thanksgiving - how much guilt is a family going to pour on about the annual gathering where 20-30 people travel, sit indoors, eating and drinking?

We need to invent a glory hole for eating.

I mentioned a few days ago that our school district is trying to acquire an on-site covid testing system. They posted the slide show, so now I have an exact quote:

"The test is one of the three EUA approved rapid tests. The manufacturer states 96% sensitivity and 100% specificity; which means it will generate a true positive 96% of the time and a true negative 100% of the time."

There is no way that exists. What I'm hoping is that the local point person just doesn't understand sensitivity and specificity (since they also don't understand semicolons).

What does "96% sensitivity and 100% specificity" actually mean?

I think that is what it means. It's plausibly not true, but I think that's what the words mean.

Could it be something like:

"Provided the virus is being shed at the point of swabbing, and gets on the cotton ball, then we have 96% sensitivity and 100% specificity (but that is a huge caveat that leads to false negatives)"?

A true positive 96% of the time the individual tested is infected. A true negative 100% of the time they aren't.

If 27 is what they mean, they are cheating.

But then such a test doesn't exist, right? And it's surely not a 15 minute test that is affordable by a school district!

26 is correct. A specificity of 100% does seem unlikely.

My personal theory is that when there's an outbreak, the number of cases rises for about a month, then decreases. That's true of most of the nations on Kevin Drum's charts, with the U.S. being unique in having a second one-month increase. Also true of most of the individual state charts at the NYTimes link, some of which have two humps of about a month each. Prevention programs can stop an outbreak from beginning, but it will stop eventually regardless. Maybe for reasons suggested by ydnew. Maybe because people start taking more precautions after the first few weeks of things getting progressively worse, even if not required by law. Maybe some virus-level mechhanism.

The death chart hasn't declined yet, but it is no longer rising, and probably will start declining in a week or so. Interestingly, although the second hump has many more cases per day, it also has many fewer deaths. Probably because nursing homes have mostly got their acts together by now, and somewhat because medical treatments have improved.

Bold prediction: There will be more outbreaks, some visible as humps on a chart, but with progressively fewer deaths in the future, even without a vaccine. Most colleges and schools that are opening will stay open.

Even with lowering death rates, we look likely to hit a quarter million deaths at minimum before 2021.

I don't know how long the nursing home improvement will last. Much of it comes from not allowing visitors. I can't be the only one realizing how much more quickly deterioration is happening in the absence of visits, even with calls/video.

25: Sounds like the Quidel antigen test, whose manufacturer claims 96.8% sensitivity and 100% specificity. The Quidel test is the one that gave Mike DeWine a false positive, so I would be wary.

I mean, for screening, pretty good is pretty good, so even if the claims are overstated it's okay. But they've got to be overstated.

I still think he just lied. Calling in sick when you don't want to do something you can't really say "no" to is a thing as old as society.

The California numbers are complicated by the malfunctions in the state registry. For a while (it's not clear exactly what period -- end of July to early August maybe?) the registry significantly undercounted cases. This dip was followed by a steep bump as the state attempted to catch up. I think there were about 300,000 cases total involved in the undercount-and-catch-up. So, to some extent the pattern described in 19.2 is due to the registry problem.

I download my data anew from data.ca.gov each time I run it, so it's possible it's corrected for that if they gave the backlog the right dates.

Late July to early August is the period of the trough - so if it's not corrected, maybe it was actually steadier than we thought, and the second peak was less of one? Interesting.

Huh. It turns out maybe college football really is a bad idea. Who knew?

36 is where I landed, too. Quick, repeatable, inaccurate tests are pretty good for keeping an eye on outbreaks. But the lying weirds me out.

Standards for that are really low right now.

My personal theory is that when there's an outbreak, the number of cases rises for about a month, then decreases. That's true of most of the nations on Kevin Drum's charts, with the U.S. being unique in having a second one-month increase. Also true of most of the individual state charts at the NYTimes link, some of which have two humps of about a month each. Prevention programs can stop an outbreak from beginning, but it will stop eventually regardless. Maybe for reasons suggested by ydnew. Maybe because people start taking more precautions after the first few weeks of things getting progressively worse, even if not required by law. Maybe some virus-level mechhanism.

That's possible. The pessimistic take is that because we have done much less than other countries to prevent needless contagion, especially among low-income people who are forced to work and nursing facility residents, and we keep pushing the boundaries on how quickly we can reopen, the virus will keep slowly burning through the population at a lowered but still bad rate, with periodic outbreaks.

My further evidence for fearing this is that looking by these eight regions, there have been high peaks (25+ cases/day/100,000) in only four of them so far: Mid-Atlantic, New England, Southeast, and Southwest. Far West is borderline, peaking at about 23. That leaves Rockies, Great Lakes, and Plains states as potential sites of future outbreaks.

35: False positive is a hell of a lot better than false negative in this context though. If worst case is 1/25 of pos results have to unnecessarily quarantine until a more accurate PCR result clears them (a week tops? Less than a day if you're governor?) I'd take that all day over a similar percentage of infected folks unknowingly circulating. Now 0% false neg may be too good to be true but there's a lot of microbio involved I know nothing about. As LB says, even mediocre rapid tests could revolutionize daily life. I might be ok going to a movie if everyone was tested at the door and any infected were ushered out before the trailers ended. Same thing for schools, workplaces, etc.

I'm not smart enough to understand how the link in 42 supports your point, but I take it anyway.

44: I was looking for better regional divisions of the US than the typical "Northeast/Midwest/South/West", and ran across those divisions used by the Bureau of Economic Analysis, which I horked. No relevant data is at the link other than those state groupings.

Nate Silver notes that part of what we're seeing are different peaks and drop-offs within different parts of a state.