98-death “surge” on July 11 is actually spread over 54 days

BY LEN CABRERA of The Alachua (FL) Chronicle.

"On the July 11 update, the Florida Department of Health Dashboard shows 98 deaths more than the July 10 update (and an additional 10,486 cases). That number of deaths may sound shocking to people who haven’t been paying attention to the data reporting; they could be downright confusing to anyone looking at the bar chart showing deaths on the Dashboard, which shows less than 10 deaths on July 11. (The graph from the Dashboard is shown below.)

Below the chart is this fine print: “Death data often has significant delays in reporting, so data within the past two weeks will be updated frequently.”

The FDOH line data file that feeds the Dashboard shows that some of those 98 deaths date all the way back to May 17. The file has one record (row) for each COVID-19 case in the state (total of 254,511 rows on July 11). For each case, the data denotes whether the person was hospitalized or died, but not the date of either. By comparing the file to the July 10 file, we can see which COVID cases the 98 deaths came from. The graph below shows the deaths by date of the cases from the July 10 file in blue, with the changes in the July 11 file in red.

This is a quick graph based only on sum of deaths by case date. Sadly, it’s not possible to easily match up the records between the July 10 and 11 files because the “ObjectId” column, which assigns a unique number to each record, changes from day to day. Because of medical privacy laws, FDOH cannot have personally-identifiable information in the publicly-available data file.

Of those 98 deaths, 12 of them are associated with COVID-19 cases from over 4 weeks ago (as far back as May 17). Another 11 were 3 weeks ago, 21 were 2 weeks ago, and 54 were within the last 2 weeks.

Unfortunately, the FDOH data does not include the date of death to filter situations like that May 17 case that was reported as a death on July 10. It could be that the person died weeks ago, but the information finally made it into the system on July 10. It could be that the person did in fact die on July 10, 54 days after a positive COVID-19 test. There’s no way to tell from the publicly-available data. But the state’s graph does not show these 98 deaths on July 10, so we know they occurred on earlier dates; we just don’t know exactly when."