I’ve been watching the data daily as cases begin to surge in many parts of the US. I have indicated that I suspect that the case volumes this winter could be 2-3 times as high as the surge last winter. I detected a new signal today in the data that suggests that my assumptions are correct.
I have been using the first derivative (red line) using the prior three weeks of data to monitor acceleration and deceleration of cases since very early in the pandemic. As I’ve studied it, I’ve also noticed that this derivative slope can forecast reasonably well about 10 days ahead, where you can see it extending further to the right than the epidemic curve below it in blue.
Visually, what concerned me was that the tail seems to be climbing much more steeply than in the past. I am disregarding the very middle of the graph since all of the up and down noise is mostly due to lack of reporting of data during the holiday weekends.
When zooming in, it is even easier to see just how sharp the climb is with that line.
That led me to look at the second derivative to see if that rate of change was really increasing faster than in the past using the past week of values from the first derivative.
In the graph below, the red line is the same as above (the first derivative), but on a slightly vertically compressed scale. The dotted black line is the second derivative, which can be thought of as the rate of change of the rate of change.
Again, ignoring the holiday noise in the middle, the outlook for the second derivative shows it exceeding any prior time in the past sometime over the Thanksgiving holiday weekend, as indicated by the horizontal yellow line.
I’m expecting case counts to get into the upper 200,000s sometime over that weekend. Unfortunately, we won’t even know that until the middle of the following week for a number of reasons: people are less likely to get tested while traveling, testing sites are less likely to be open, testing capacity will be stretched after the weekend, more at home tests are being used that don’t get reported, and reporting will lag because of the weekend.
I will throw out one last appeal to reconsider travel, especially on any kind of mass transit, over the holiday. Avoid contact with anyone outside of your immediate household. For many people that don’t heed this, it will be the last Thanksgiving they share with some of their loved ones.
I will update the two primary graphs shown daily until sometime after Thanksgiving, but leave the ones above in place as a point of comparison.
11/21 Note that both the first and second derivatives are turning downward. Don’t be deceived by that. It’s an artifact from cases not being reported over the weekend in many jurisdictions. It will look that way most Sundays and Mondays. That will likely also be the case Thu-Mon of Thanksgiving weekend.
The second derivative is much more sensitive to these changes by design. It’s only using the past week of data, while the first derivative is using three weeks. That was done to look for longer trends in the first derivative while the second one helps identify real changes that are occurring aside from the expected reporting variances.
11/23 As expected, the brief downward trend over the weekend for both derivatives has ended. Tomorrow should provide the best data set to get the clearest picture for another week, but even this is hampered by unreported at home tests and the 1-2 day turnaround for the two major send out labs in the US.
11/24 The virus is proving that it does what it’s going to do and with some variation. There are a lot of variables that eventually become a part of case counts so changes in trajectory are expected.
We will have a good data set tomorrow, but after that we are going to be in somewhat of a data blackout similar the Apollo missions going behind the moon. We have a guess as to where it’s going, but not enough data coming in to know with much certainty. That will be a problem until the middle of next week.
11/26 As expected, the holiday weekend is making it appear that the situation is improving. Other evidence paints a different picture, such as hospitals cancelling elective surgeries, shortages of pharmaceutical and medical supplies, and some hospitals implementing crisis standards of care.
11/30 A little data from the holidays is starting to trickle in. It won’t be readily visible on the graph for a few more days because of how they are designed. The first derivative is smoothed using the data for the past seven days, so it’s still under the influence of the holidays. Currently, the slope that is projected for 10 days out as a point estimate is 582. That coming change is more apparent as the 2nd derivative is approaching zero. I also still suspect that these numbers are underreported due to more home testing as people had been preparing for the holiday weekend.
12/2 Now that we are getting closer to having caught up on the data lost over the holiday, the country is back on track with where I had seen it going. It was bad enough with delta. We are superimposing another pandemic ON TOP of delta with the arrival of omicron. I’ve indicated the data gap in the slop line with a double-ended arrow. Realistically, that entire last part should be slid up so there was no downward dip. The second derivative suggests that the country is heading into the second fastest rate of rate increase we have seen so far.
The nation is speeding toward a cliff without brakes.
12/3 This looks bleak for Christmas.