Series 1: U.S. COVID-19 Data Charts

Data last updated: March 5, 2021

Overview Charts - National, Cumulative Data: Testing, Cases, Medical Treatments

***

Probably the single most interesting chart on this page is Figure 1-11. Scroll down and find Figure 1-11. It is the last chart shown in the second series of charts on this page.

Look at annotated lines “I” vs. “J”. Line I represents the trend for the number of new COVID tests being performed. The amount of testing has been decreasing since early December. Line J represents the number of new COVID cases since early December. Naturally, it is dropping as well since the U.S. is testing less. But the downward slop of Line J is much steeper than Line I. This tells you that the rate of new infections is decreasing more rapidly than the rate of new testing – which, of course, is a very good thing.

***

Now, back to the top series of charts. This first series will explore cumulative COVID-19 data at the U.S. national levels. All data comes from The COVID Tracking Project which is updated daily on their site. Within this site, the data is likely updated less often, maybe once or twice a week.

The first chart, Figure 1-1, shows four sets of data, including:

• the total number of COVID tests performed;

• the number of tests returning positive for the presence of COVID;

• the number of hospitalizations due to COVID; and

• the number of deaths attributed to the COVID-19 virus.

..

Figure 1-2 shows the same data as Figure 1-1 but uses a logarithmic scale on the Y-axis. Notice in Figure 1-1 that COVID positive cases are significantly less than the total number of tests performed, accounting for less than 10% of the total number of tests performed. Therefore 90% of the Y-scale is unavailable for the positive cases data set and the other two data sets as well, hospitalizations and deaths. Hospitalizations and COVID-related deaths are barely noticeable in Figure 1-1 due to scaling factors on the Y-axis.

When comparing data sets whose values can vary greatly, using a logarithmic scale helps in two ways:

• the detail in lower value data sets is increased; and

• if the X-axis is a time-based scale, the rates of change in the data sets can be compared by comparing the steepness of the curves whether they are increasing or decreasing. This latter point is explained in more detail in Figure 1-9 below.

..

Figure 1-3 shows more detail in medical care responses to the COVID-19 virus. In addition to the number of hospitalizations shown in Figures 1-1 and 1-2, it displays data sets regarding cases requiring intensive care units (ICUs) and respirators.

..

Figure 1-4 shows the same data as Figure 1-3 but uses a logarithmic scale for the Y-axis. Again, this enables the reader to see more detail in the number of cases requiring more extensive medical care.

Figure 1-5 shows the cumulative positive test rate for COVID-19. This has been calculated simply as:

Positive Test Rate, PTR = (no. of positive cases)/(no. of tests performed).

Overview - National, Daily Totals, COVID Testing and Cases (Figures 1-6 thru 1-11)

The next series of charts show us a more detailed daily view of the data than the first four charts do. These plot the daily number of cases using a bar graph and include a LOESS fitting curve to smooth the data between dates. A LOESS fitting curve is not unlike taking a multi-day average to smooth out the data points but is calculated differently. The graphing application used to construct these charts has a LOESS fitting function that I sometimes use.

Figure 1-6 shows the rate of COVID testing per day had been steadily increasing from March 2020 until about mid-January 2021. Around this point, new daily testing has dropped rapidly. There also appears to have been a lull in the rate of testing from mid-July to early-September before rapidly picking up again through the winter holidays.

Figure 1-7 shows the national (U.S.) number of daily new cases in COVID since March. Here you can see the distinct peaks: the early peak in April that largely impacted the Northeast; the mid-summer peak that largely affected the South; and the autumn peak which impacted everywhere but probably the upper Midwest the most. A late autumn surge in California and the Great Lake states drove the curve even higher.

Figure 1-8 simply shows the data shown in Figures 1-6 and 1-7 combined into a single chart. Note that the number of tests greatly exceeds the number of new cases. Figures 1-6, 1-7 and 1-8 all use a linear scale on the y-axis.

Figure 1-9 shows the daily positive test rate (PTR).

Figure 1-10 shows the same data in Figure 1-7 but with a logarithmic scale on the y-axis. Look at the two curves from around the start of October until about the start of December. For this period, the red curve (new COVID cases) increases at a significantly steeper rate than the black curve (new daily testing). This means that new cases were surging during this period – and not just due to increased testing. But compare the two curve from roughly the start of January. At the time of this post (February 11) the rate of new daily cases has been dropping more rapidly than the rate of new daily testing. This is shown more clearly in Series 4 chart Figure 4-5.

Figure 1-11 is an annotated chart using a subset of the data shown in Figure 1-10. It is repeated in larger form below this chart and an explanation of it follows in The Logarithmic View.

The Logarithmic View

What is the logarithmic view telling us?

I’ve included exponential fit curves on Figure 1-11 to help explain logarithmic curves might tell us. Note: The x-scale on this figure begins on June 1 as opposed to March 1 on the previous charts.

From the first of June 2020 to roughly mid July (curve sections A and B) the rate of new COVID cases (red curve) was rising quicker than the rate of new testing (black curve). We can see this by looking at the converging lines A and B. For these series of logarithmic-scale charts, when lines of two sets of data converge, it means that the rate of new cases is out pacing the rate of new testing. These two sections of curves correspond to COVID’s second wave in the U.S. (the first wave, primarily in the Northeast preceded June 1).

Curve sections C and D correspond to testing and new case rates between mid-July to the start of September. The correspond to a relatively ‘quiet’ period of this COVID pandemic, mid-to-late summer 2020. People moved outdoors as the weather warmed. Here, the rate of new cases is dropping faster compared to a less severe drop in new testing for this same period. When these two curves are diverging, new cases are dropping relative to new testing over the same period.

Curve sections E and F are roughly parallel. The rate of new COVID cases is increasing at roughly the same rate at the rate of new COVID testing when the curves are parallel.

Curve sections G and H correspond to the two month period of October-November. The rate of daily COVID testing has been steadily rising. However, the rate of new COVID cases has be increasing at a much greater rate. These period corresponds to cooler weather and a move indoors, a return to university campuses, perhaps a relaxing of precaution due to falling summer numbers, and maybe a smaller surge due to Thanksgiving holiday gatherings. Autumn news accounts of increased case loads, hospitalizations and numbers of patients in intensive care units (ICUs), especially in California, attested to this rapid increase in cases.

New sections I and J show the data since the first of December. The two new fit curves show both new daily testing and new daily cases have been trending down for much of this most recent period. These two curves are nearly parallel and suggest that the new case rate is declining slightly quicker when compared to the decline in rate of new testing for this period.

It is easier to detect comparisons of rates of change between two time-based datasets like this when using a logarithmic curve.

...

Breather: from this point forward I’m going to try to get less wordy and more to the point. I will try to introduce each chart, figure or table with the following 3-point outline: (a) what is this graphic telling us; (b) why is this important; and (c) what is the next step or view forward.

Overview - National : Medical Treatments and Deaths

Figure 1-12 shows current hospitalization rates due to COVID-19, those subset of patients hospitalized in intensive care units (ICUs), and those further subset of patients requiring respirators. Looking at all three datasets, hospital visit rates have been occurring in waves of 3-4 month spreads. The most recent wave has been the largest and appears to have peaked in early January 2021. It has been on a strong trajectory downward since. Likewise, ICU patients and patients requiring respirators to assist breathing have seen a similar wave pattern, as to be expected, and are currently declining.

Figure -13 shows the same data as Figure 1-12 except a log scale is used on the y-axis.

Figure 1-14 shows a calculated metric: hospital rates. I don’t know if this is considered a valid ‘rate’. It is calculated as follows and represents a daily snapshot:

hospital rate = (current hospitalizations/total new test results)

..

On an even more somber note, Figures 1-15 and 1-16 show a daily national (U.S.) death count and cumulative death rate, respectively. The death rate is calculated simply as:

death rate = (total deaths/total positive test cases)

Figure 1-16 shows death rates from COVID-19 have leveled off to approximately 1.8-1.9 percent. This is still high in comparison to other respiratory illnesses, but the rate has been declining since last May.

keywords:
covid, covid-19, icu, hospitalizations, trends, states, United States, USA, charts, plots, tests, positive, positivity, logarithmic,