Back in April I never thought I’d be writing about COVID in November, but here we are. Nationally, new cases are setting records, but many of the wave 3 “hot spot” states are leveling or in actual decline. There is a lot happening with the numbers right now, so I’m going to report a bit more often, probably twice a week over the next few months. I think it’s important to analyze the high watermarks in the wave 3 states as they peak and decline, and it won’t be long before we start to see the effect of vaccinations on the numbers. I’m also going to write later this week about South Dakota (spoiler – it’s in decline), and an update on Sweden (also in decline).
Let’s start today’s discussion with a look at peak active cases by the state as a percentage of the population. I have long written about my theory that COVID peaks at a low % of the population (I’ve said between ¼ and ½ of a percent) and then declines. This has been borne out by observing wave 1 in April and wave 2 in July. For just a bit of background, here is the introduction from my report on August 6, 2020, a bit over 3 months ago.
I’ve talked a lot over the past weeks about how COVID known active cases tend to peak in the range of 0.25% to 0.50% of the population in a given area, then decline. For the longest time, it has puzzled me. However, I’ve been seeing more and more theories about full or partial immunity in a large swath of the population. Although I know little about medicine, this is quite congruent with the data, so I tend to believe it has some validity. Here is an article published today on CNN by Dr. Sanjay Gupta, postulating that as much as 50% of the population carries some level of built-in resistance to COVID: https://www.cnn.com/2020/08/02/health/gupta-coronavirus-t-cell-cross-reactivity-immunity-wellness/index.html
This may be the missing piece of the puzzle. I’m seeing a ceiling in known active cases of, let’s say 1/3 of one percent. Now bear with me for a little arithmetic — When a location reaches this threshold, roughly 5 times the active case count has confirmed positive, so that means something over 1.5% of the population has been confirmed to have had COVID. However, we know that the actual prevalence of the disease is a large multiple of known cases. There is quite a range here based on the various antibody studies, but the average consensus seems to be about 10x. Assuming that, when a locality reaches the ceiling, perhaps 15% of the population has had COVID in total. If we add this 15% to 50% of the population that has some level of innate immunity, we have 65%, which is right in the range of herd immunity.
Let’s see how it looks today. Here are the 18 states I track with their peak known active cases as a percent of the population:
The horizontal line is the 0.50% of the population mark, where I’ve hypothesized that the COVID population ceiling exists. However, we have several states above this level. Of these, the first four on the left (WI, CO, IL, IN), have recently slowed, leveled, and 3 of them have begun their decline. Of the remaining states, most have had their high watermark in the past, but may still be growing or not (see details below). The only states that set new high water marks yesterday are OH, PA, TX, MI, VA, CA, and NJ. Of these, MI and NJ are now slowing.
So we’re still seeing known active cases turn around at a fairly low level of the population, although a bit higher than I first theorized. I believe this is because we are now “discovering” a higher percentage of actual COVID cases than we were in wave 1 and wave 2. You might remember that reported cases represent a fraction of the total prevalence of COVID. Random testing studies from April to August point to a prevalence to a reported ratio in the 10 – 20 range. I used two of these studies back in late April to make my initial calculations of mortality rates, which, by the way, still seem to be valid. (See my mortality rate discussion at the bottom of this report from April 28th). The problem for those of us mathematical modelers is that we haven’t seen any robust domestic random test studies since August. Many experts are still using the factor of 10 as rule of thumb for prevalence to reported case ratio, but could this still be reasonable? Back during wave 2 in July, we had a reasonable idea that a factor of 10 was in the ballpark. That means that my hypothesized ceiling of known active cases at 0.50% would translate to about 5.0% of “true” active cases. However, we now have a cumulative total of about 12 million reported cases. Are we to believe that this implies that over 120,000,000 Americans have or have had COVID? That would be 1 out of every 3 people. I don’t think so. I have to believe that the prevalence to reported ratio is lower now, due to ubiquitous testing. Take a look at the daily reported test numbers.
Back in July, we were reporting daily test results in the 750,000 range. However, yesterday we set a new record of just under 2 million reported tests. I do believe that any given new COVID case is more likely to be “known” than it was in July. If this is true, then we’d expect to see higher percentages of the population at peak than we saw in wave 2. In any event, the states on the left side of my high watermark graph are all turning around, so I think the general behavior still holds.
My first daily death projection
I’ve often shown you this calculation, which is daily deaths in the U.S. per 1,000 modeled active cases.
Here we see continued declines in this metric for the past 9 weeks. From this we can conclude that the disease is far less deadly than it was in wave 1 back in April. But can we conclude that wave 3 is less deadly than wave 2? Not necessarily. We know that, with some lag time, cases predict hospitalizations, and hospitalizations predict deaths. Although it has not been true for wave 1, during wave 2 we learned that deaths lag cases by about 3 weeks. Now look what happens when we calculate daily deaths per 1,000 modeled active cases, but with a 21-day lag:
Now that’s remarkable. The metric is very stable since wave 2, averaging about 2.75 deaths per day per 1,000 modeled known active cases. As of yesterday, I model 1,167,945 known active cases. From that, I project about 3,200 cases per day 21 days from now. Interestingly, the latest IHME model update is projecting 2,200 daily deaths by December 12, about 2/3 of my projection. I hope to be wrong.
Well, I think that’s a lot for today. I’ll report all the various stats in my next report in a few days, but for now, I’ll just show the progress in the 4 wave 3 hotspot states in some state of improvement.
It was just recently that Wisconsin was the most troublesome are of the country. After rapid growth, the state began leveling a bit over a week ago.
Here the slowdown is only a few days old, but Colorado is now at 0.70% of the population, so I’m inclined to believe the slowing will continue.
Illinois has been in decline for 8 days now, and at 0.69% of the population, it makes sense.
Indiana began slowing 8 days ago and saw it’s first actual decline yesterday. Indiana is at about 0.66% of the population.
Well, that’s all for today. In my next report, I’ll go through the complete slate of stats, as well as a report on Sweden.
Shane Chalke, FSA