Trying to estimate how much COVID is out there is tough if we assume that there is some spread above zero and also non-universal spread. We have to estimate the situation from incomplete information. We have case counts which is a function of who is getting tested with what types of tests at various points in time with various levels of background COVID disease spread. We have the number of tests and then we also have positivity rates which is merely the quotient of new infections divided by the number of tests administered in a period. We should assume that testing is non-random in most cases. We should strongly suspect that people who have already been infected are unlikely to be tested and people who are not symptomatic nor knowingly in recent contact with someone who had been infectious are less likely to be tested than people who have reason to suspect that they might be infected.
We should assume that we’re not testing enough if we are getting 80% positivity rates as there is massive community spread at that point. We are probably overtesting if we are getting 0.001% positivity rates.
I have a cheat sheet to help me think through the daily reports.
This is not perfect. But as long as positivity rate is going down in the context of flat or increasing testing, community spread is probably going down. If positivity rate is going up, a state or a region might have an upcoming challenge.
Cameron
Thanks for offering your professional insight; it’s all too easy to get washed away in the flood of nonsense churned out over social media. If you don’t mind, I’d appreciate your thoughts on this article: mercurynews.com/2020/06/07/marin-epidemiologist-larry-brilliant-virus-crisis-just-beginning/ Thanks!
Cheryl Rofer
Thanks, David. This is useful.
I’m disappointed to see New Mexico go from green to red on your map. I’ve been thinking about getting back to piano lessons, and I am trying to decide on a couple of routine medical appointments. Maybe on the appointments, better to go now than when the wave comes.
David Anderson
@Cameron: I am not qualified to assess the argument he is making. There are a couple of things (doubling times) that seem off given the data I’m seeing
Cameron
@David Anderson: Thank you!
Punchy
@David Anderson: Ah…..Dublin times. Memories….
Ohio Mom
My county appears to be on a very slow downward trend — there are a handful of sites I check daily, among them the NYT, rt.live, covidactnow.org, and my local board of health.
They probably all use the same raw data but I find the variety of presentations helpful to me in trying to get a handle on what is happening ony corner of the world.
I am assuming this downward trend is a lull before another wave. So I’m doing things like catching up with medical appointments and stocking up on nonperishables. Oh, and sewing and collecting masks.
David Anderson
@Ohio Mom: A second wave is not inevitable. If a region is aggressively testing and then aggressively contact tracing and isolating folks in the context of generalized precautions, a region can probably ride things out without catastrophic spikes
General precautions include doing as much as possible outdoors, limiting close contact, very high percentages of mask/facial covering wearing, good hygiene and avoiding crowds.
Shutdowns are to be avoided if possible. We know what tools are effective at avoiding shut-downs and we know how to use them if we allow ourselves to do so.
And then there are some states and regions the will proclaim FREEDUMB!
Wag
I think your framework is spot on. It certainly reflects what I am seeing in Colorado. Keep up the excellent work!
Rugosa
This is great, thanks. Massachusetts is starting Phase 2 of re-opening and I’m waiting to see what happens in the next couple of weeks. This gives me a useful way to make sense of the numbers as they come out.
Roger Moore
One thing I would question is the case with a rising test count but a flat positive rate. I think that is likely to be a sign of increasing infection, which is bad. It’s not a sign of rapidly rising infection, so it’s not as bad as an increasing positive rate with an increasing test rate, but it’s not good.
David Anderson
@Roger Moore: If positivity rate is constant in the face of rising testing, the assumption is either slight increase in community spread as the marginal tested individual should be less likely to be positive than the average individual, but as long as it is not spiking, not a huge concern over a few day period.
Xavier
@Cheryl Rofer: I’m not sure things are that bad. The spike recently seems to be a cluster in detention centers. Your county hasn’t had a new case in weeks.
Here in ABQ, I went grocery shopping and to a quick stop at a big box, and every single person in both places was wearing a mask.
Another Scott
I agree that your table is a useful way to think about things, but we’re still suffering from incomplete (and intentionally misleading in too many cases) information.
Other countries have shown that it is possible to reduce new infections, and even eliminate new infections (New Zealand). We should not accept that “eh, there’s nothing we can do, we can’t stay locked down forever, so we might as well open up NOW NOW NOW” attitudes that so many people seem to have. 20,000+ new confirmed cases and 1000+ new deaths a day is not success. It’s a tragedy.
:-(
In addition to rates, we need to look at numbers of hospitalizations and deaths. We should not accept that nothing more can be done.
I think Larry Brilliant’s mantra that 80% of the population wearing masks 80% of the time will knock the virus back should be tested. It seems reasonable and seems worth trying.
Also, I still think my favorite metric for “are we over the hump yet?” is Total Recovered / Active Cases at worldometers.info/coronavirus/ Is that ratio substantially over 1? No? Then we’re not over the hump yet and it’s still dangerous out there. Like any number, it can be gamed of course (look at Brazil), but it seems to tell us something about this point in time compared to successful countries.
Thanks again.
Cheers,
Scott.
Uncle Cosmo
I examined your chart in the spirit of Mr. Oberle, my HS calculus teacher (the best math instructor I ever had, out of a bunch of very good ones), whose guiding motto was Why make a simple thing complicated?
It shows that, in effect, the only factor affecting “status” is the positivity rate – if it’s flat or down the status is good, if it’s up the status is bad, irrespective of case count or test count. Is that a productive insight, or more on the order of “water is wet?”
David Anderson
@Uncle Cosmo: That is about 90% of the take-away of the chart.
Cheryl Rofer
@Xavier: On Monday, I went to the store and saw a notice on the door that because of the governor’s order, all shoppers must be masked. And they were. It’s important to have leadership for a collective-action problem. And important to the stores that they not have to try to enforce this by themselves.
New Mexico is doing better than others, but it’s too early to reduce restrictions. People have totally forgotten, or never learned, how exponentials behave. And the last lessons were only a couple of months ago. A lot of us said then that if things work out, there will be complaints that too much was done. Well, here we are. A thousand deaths a day nationally, and yeah, let’s party.
I’ll be pleased to be wrong. We’ll see.
America fails the marshmallow test – Paul Krugman
Ohio Mom
David Anderson: Ohio as you know, started out strong but now? Governor DeWine is dropping the ball on a lot of things, including mask-wearing. We are opening up things at a quick clip (amusement parks too).
I don’t know anything about our testing numbers but can’t help but feel we are tempting fate/the virus with our glee at getting out of lockdown.
If we are lucky and a second wave does not come, at least my family will be up to date on medical appointments and I’ll be ahead of the grocery shopping.
Another Scott
@David Anderson:
Relatedly, CalculatedRiskBlog.com:
Cheers,
Scott.
dmsilev
@David Anderson:
My employer is starting to phase in an on-campus reopening, and this nicely summarizes a large chunk of what we’re doing. We’ve added in requiring people to log their locations and any close contacts with other people while on campus, to facilitate contact tracing should that be needed.
(first phase will probably see about 20 percent, at most, of the normal population back on campus, and everyone else still working from home)
Dupe1970
One thing to look at is hospitalization rates and ICU usage rates especially for any respiratory ailment. Independent of testing and rates of infections versus testing we can easily look at how quickly hospitals are filling and space being used as marker for overall infection spread.
frosty
Thanks for your link yesterday: covidexitstrategy.org. My state (PA) it looks like it’s doing well except not enough testing. Maryland is a mess.
Hoodie
Becoming pretty disappointed in NC’s performance. Seems like we’re backsliding into a potential crisis and GOP legislature is trying to make it worse.
Barbara
@Cheryl Rofer: It’s the same in most parts of Virginia. Grocery stores will not let you go in without a mask. I haven’t been to other kinds of stores so far, most are not open to go into for now. One good thing about Northern Virginia is that our critical mass of government workers is predisposed to go along with most collective action measures. They give public health officials the benefit of the doubt.
bks
For the world or for the USA? Can we be in a good position while, say, Mexico is worsening?
StringOnAStick
There was a woman collecting petition signatures in front of the main local grocery on Monday, all for RW ballot measures (the ban on abortion after 22 weeks has made the fall ballot already, with an assist by extending signature collection due to Covid). She was wearing a mask, but I watched and she wasn’t cleaning pens or clipboards after each person. I think the nature of what she was collecting for and people being hesitant to handle that stuff and how close she would get to people was hurting her “sales” rate.
KsSteve
What I find particular frustrating is the lack of clarity of data we do have. Is the positivity rate “ merely the quotient of new infections divided by the number of tests administered“? I’m guessing some individuals are being tested multiple times for various reasons. Do we actually know how many people have tested positive vs how many test results are positive?
David Anderson
@KsSteve: Different states report different rates
That is actually one of the things I am working on this afternoon.