Thread by @zorinaq

Every day Florida reports new COVID-19 deaths, but some of these deaths may have occurred days/weeks earlier as it takes time to process and record them.

The data at fdoh.maps.arcgis.com/home/item.html… shows deaths by exact calendar day.

I charted both datasets.

1/N

1st observation: it takes on average about 4 days for a death to be reported. We can see this lag more clearly in periods of increase.

2/N

2nd observation: in periods of increase, the true number of deaths is always higher than the number of reported deaths, because of the lag.

3/N

These artifacts exist in the COVID-19 deaths data from all states. Very few report "deaths by date". A few that do:

Florida: fdoh.maps.arcgis.com/home/item.html…

Indiana: coronavirus.in.gov/2393.htm

Mississippi: msdh.ms.gov/msdhsite/_stat…

4/N

Keep in mind that because of the time needed to assemble the "deaths by day" dataset, data in the last week or so is very incomplete.

Here is a chart I have assembled to monitor how deaths by day counters evolve over time:

5/N

In one extreme example, one death, that occurred on 2020-06-08, took 36 days to be recorded: notice how the line for 2020-06-08 jumped from 24 to 25 in the rightmost part of the chart.

6/N

This chart also shows that the death counters of the last two days, 7/16 and 7/15, have increased faster & more sharply than any other day.

So even though the data is incomplete, it's a strong indicator that 7/15 and 7/16 are likely to establish record high number of deaths

7/N

I finally got more data to complete the chart:

https://twitter.com/zorinaq/status/1284789258868256770

8/N