COVID continues to proceed at a crawl. Nationally, modeled active cases are up about 29,000. This could be the bump I was expecting from extensive university testing, but it’s occurring a couple of weeks later than I anticipated. You all probably saw the news reports of a surge in new cases over the past 2 days, but this is quite misleading. Over this time period, Texas increased their total case count by 27,673 cases. However, 21,967 of these are old cases newly reported. To Texas’s credit, they did not report these as new cases, however, Johns Hopkins and The COVID Tracking Project did. Of the 27,673 cases, Texas reported 1,742 on Monday as new, and 3,964 today as new. Johns Hopkins seems to make no attempt to correct this sort of data, but The COVID Tracking Project does a pretty good job of it (but it takes some time). I think they’ll revise the Texas numbers sometime over the next week or so, but in the meantime I’ve corrected for this in my numbers.
IHME has revised their model again on September 18. They are now projecting that daily deaths peak on December 21, 2020 at 3,620 per day. However, they also model “universal mask usage”, and under that model the April 16, 2020 daily peak holds. I have no idea how their model is built, but they have been overestimating daily deaths for the past two months. In reality, we’ve observed a slow, but steady decline in daily deaths for 7 weeks now.
Testing continues at a high rate, and we set a new record 3 days ago of just a hair under 1,000,000 tests. In total, the U.S. has recorded over 93 million tests. The New York Times ran an article at the end of August where they suggest that the tests are too sensitive. In case you didn’t see it, here it is:
Here is the relevant quote: In three sets of testing data that include cycle thresholds, compiled by officials in Massachusetts, New York and Nevada, up to 90 percent of people testing positive carried barely any virus, a review by The Times found.
I expect we’ll see more about this as time progresses.
Feel free to send me your questions about my assumptions, methodology, or modeling in general.
- Modeled known active cases in U.S. 279,198
- Likely date of active case peak (Chalke modeling): July 23
- Likely date of peak deaths (IHME): December 21 (last revision on September 18)
- Total Test Results reported today: 719,578 (high)
- Total Pending tests reported today: 7,999 (very low)
- National reported case Growth Rate today: 0.51% (very low)
Shane Chalke Interviews
Groom Ventures has agreed to host a website that will archive my daily reports, and supplement with other commentary. John Groom worked at one of my companies back in the day, and is an excellent writer. The website is: www.howmuchrisk.com For those of you that post my daily report on Facebook, let me suggest you link to this site, as the direct Facebook posts don’t seem to copy the graphs.
Here is the national picture. We’ve added about 29,000 cases in the past 9 days. I suspect this is the widespread asymptomatic testing in university settings, but haven’t analyzed it yet. One thing that puzzles me lately is the phenomenon where geographical areas will reach a low level of COVID prevalence, and, rather than continuing the decline, it remains stationary. Will this ultimately be like the common cold (another Coronavirus), where a certain percentage of the population always has it?
Look at the daily new cases. We’re looking at a slow but steady decline since mid-July.
The daily death count is exhibiting a very slow decline for the better part of two months now. Here is the picture:
The decline is slow, and not particularly obvious on the above graph, but if we isolate the last 7 weeks, and show the regression line, it’s easier to see.
On to the states.
Here is Arizona. They reported a large increase in cases on Sep 17 and 18. This was a result of a change in reporting protocol, where they are now adding the results of antigen testing. As always with a change in protocol, this bump will flatten back down in a week or so.
SC has been relatively flat for nearly 3 weeks. Note that South Carolina double counts cases, as they treat each positive test as a new case (per The COVID Tracking Project).
Here is Florida – down dramatically from the top, but little progress in the past 2 weeks.
California has had a great recovery, but now flat for the past week. College testing? I think so, but don’t know for sure. As always, I need to report that California is one of the states that counts tests rather than people.
Georgia is also flat for a week now. Georgia hit a high of 0.25% of the population, so I’d be surprised to see any real increases here. GA is now down 55% from the peak. Note here again, the case numbers are exaggerated. Georgia counts each positive test as a case (according to The COVID Tracking Project).
Texas had a significant increase a week ago, but is already starting to decline. As I discussed above, Texas reported 21,967 historical cases over the past 2 days. I removed them for purposes of my analysis, but they will eventually show up in the historical data, increasing modeled active cases at some point in the past. Note that Texas also reports positive tests as cases, so is doing some level of double counting (per The COVID Tracking Project).
NC and VA tend to track each other. No real progress in the past week in either state, but I predict NC will fall further soon, based on day to day new case reports.
Here is the daily death report for NC, flat for 2 months.
And here is the daily new case count for NC. A noticeable drop from the peak in mid-July, and then 5 or 6 data points above the trend line in late August, then declining again. Based on this, we should see daily deaths in NC decline soon.
The COVID Tracking Project did a major revision of Washington data in late August, which created the sharp decline you see at the end of the month. It’s popped back up again, based on yet another change in the reporting protocol. These data aberrations have settled now, and we’re back seeing a decline.
Here are NY and NJ – I think the interesting observation here is that even though both states peaked very early and recovered rapidly, they don’t seem to make much progress once cases get to a low level. It seems asymptotic. It makes me wonder if the course of this disease is to simmer indefinitely at low levels.
Here is Massachusetts. On September 2, Massachusetts changed their case definition to the more restrictive August 6 definition released by the CTSE. As a result, they removed a whopping 7,000 cases in a single day. Since my model only showed about 2,500 cases at that time, this dislocation sent my model into negative territory. I zeroed it out, but this shows the problems with trying to model from continually shifting data definitions. This highlights the dire need for national standards on data collection and reporting. Where is the CDC? In any event, there is not much COVID in Massachusetts.
…And here is Michigan. Michigan peaked at a very low 0.089% of the population, so may continue to grow if my theory is correct. Nonetheless, MI has been relatively flat for 2 months now.
PA has been pretty flat for over 2 months. I believe PA has more growth to come, as the peak was very low (0.091% of the population).
And finally, here is Colorado. Modeled active cases have nearly doubled since my last report. I have no idea why, but have built up a healthy mistrust of Colorado data based on their many revisions and changes in protocol.
So that’s it for today. I’ll report again next week.
The numbers are still very small as a percentage of the population. Unless you’re in a high density area, your chances of contracting COVID are very small. However, even though the probability is very small, you still don’t want to get it. Everyone please continue to be as cautious as you feel necessary.
–Shane Chalke, FSA