COVID-19: Direction of the Trend
By Jerald Hughes
& Scott Robinson
As the COVID-19 pandemic moves through its 3rd and by far the largest wave in the US, we would like to know, ‘Is it getting worse?’, ‘Has the current wave peaked?’, or ‘Are our emergency measures having any effect?’ Drawing definite conclusions on answers to these questions can only happen over an extended period of time. We can only see from a remove of two weeks or more, that a wave of infections peaked at a certain time. But short of that, we can get better indicators than just the raw data, if we look at the numbers today, compared to the same day in the previous week (see our post here of Nov 30 https://www.analogicalworld.com/post/the-daily-covid-19-picture ).
Chart 1. Daily Cases Minus Cases 7 Days Ago
Wherever this line is above 0, that day of the week is higher than it was a week ago. Whenever it is below 0, it’s lower than it was a week ago. This separation into positive and negative ranges gives us an earliest-possible indication of the trend, based on good comparisons which separate out the days of the week. What we’d prefer to see is a series of negative numbers, recording decline from the past week. As of this writing on Dec 16, we have stayed mostly in the positive-numbers zone, indicating continued spread of COVID-19.
This doesn’t completely remove anomalies. The Thanksgiving break lagged reporting of data and administration of tests, so we see some isolated numbers below the 0 line. But on Dec 1, the Tuesday after the Thanksgiving holiday, COVID-19 came roaring back, hitting higher numbers than ever. This measure also provides a clean, discrete way to measure past periods of increase, peak, and decline. The first wave of course began with all positive numbers (left side of the graph). An unbroken string of positive differences (increasing cases) continued from the very first day of recorded data, Feb 29, through April 10, at which point the curve reversed into the negative range (decreasing daily cases). Our previous blog post of Nov 15 (https://www.analogicalworld.com/post/analysis-the-dangers-of-covid-s-third-wave) identified the period of uncontrolled spread in the first wave as extending from March 15 through April 11, for a total of 28 days. These dates were obtained from the slope of the data curve.
Using the method of weekday-to-same weekday comparisons, we get very good agreement with that previous analysis, counting 27 days of uncontrolled spread.
The second wave was previously estimated to demonstrate uncontrolled spread starting on June 11, and spotted July 23 as the end of that wave, for a total of 43 days. The 7-day lag method used here observes the first shift from negative range (decreasing cases) to positive range (increasing cases) on June 10, and continuing until July 23, again showing almost perfect agreement with the slope method, for a total of 44 days of uncontrolled spread in the second wave. The previous estimation via slope of the data curve placed the start of our current wave at Oct 5. This 7-day lag method first observes a shift into the continuously positive range on Oct 3, and as of this writing we see no peak. So our 3rd wave is 74 days long so far, and still counting.
Our Nov 15 analysis failed to consider any possible effects of the Thanksgiving period. Our worst-case scenario prediction then saw the 3rd wave peaking within 77 days. In retrospect, the Thanksgiving period appears to have introduced first a delay, then an extension, of the period of uncontrolled spread. Right now it appears that our mostly likely 3rd-wave outcome is an even longer period of uncontrolled spread, but [at least], one which is more gradual than we predicted without taking Thanksgiving into account. We can directly inspect the table of data for this measure to see distinct zones of development:
For this interpretation, we are looking only at the daily numbers (WeekdayLagChange), not the 7-day averages in the rightmost column.
Nov 10 – Nov 20 (red): accelerating spread Nov 21 – Nov 25 (orange): still increasing, but not as swiftly
Nov 26 – Nov 30 (green): Thanksgiving, illusion of decline in spread
Dec 1 – Dec 9 (red): even faster uncontrolled spread
Dec 10 – Dec 15 (yellow): still increasing, but more slowly
The news from Dec 10 through Dec 15 would be good, if it led to the negative territory of the graph. However, Dec 10 -14 instead show increasing numbers. Only the very last day, Dec 15, displays a negative number. As I write, Dec 16 data is not yet in the record for covidtracking.com. Unfortunately, the data from worldometers.info, which cuts off each day at 0:00 GMT, shows another huge increase for the 16th, 7.9% over last Wednesday, just barely missing setting yet another all-time record high in daily cases.
All opinions expressed in this blog are solely those of the authors, and do not reflect those of University of Texas Rio Grande Valley, or any organizations of which either is a member.