This CDC Chart Sucks

There is bad data visualization and then there is the “COVID-19 cases in the United States by date of illness onset” chart shared by the CDC. It is truly remarkable in how badly it was put together and how badly it is maintained. This is the chart.

I have seen a lot of people ignore that caveat in the upper right to make the claim that covid cases are falling dramatically. They are not.

I have also seen a lot of people who read that caveat, saw the note that illnesses before 3/16 may not have been reported, and assumed every number to the left of that date is stable. It is not.

In today’s update, the CDC updated dates as far back as March 10 (6 days outside that window) BY MORE THAN 100 CASES. That revised March 10 up by 16%. Not a minor adjustment. What’s more, March 10 is the only date in the chart — outside their disclaimed window — that shows a decline.

On March 23rd, when I began tracking this chart, March 10 was already outside the stated disclaimer of unreported cases. In the last four days, March 10 has been adjusted upward a remarkable 253% and that date is unlikely to show a decline at all when the data is updated again on Monday.

Looking beyond the identified incomplete data in the 3/16 to 3/26 window, and the wildly fluctuating numbers going back a week or more before that, the case counts today were revised up as far back as January 30.

The CDC needs to seriously improve, get rid of, or dramatically increase the disclaimers on this chart. This is incredibly misleading data being put out in the name of public information. It is unintentionally misleading, but it is misleading nonetheless.