Recent Florida Death Data Reporting is Painting a Confusing Picture

Carter Mecher MD -- Medical Advisor
August 14, 2021

What some might not realize is that it appears that FL changed its reporting of deaths this week. The result is that it gives the appearance that FL deaths are declining.

In early June, FL transitioned from reporting daily data on COVID to weekly reports.  At the time COVID activity was relatively low in FL and this transition was not unreasonable. But things have changed since early June.  Today, FL is the epicenter for the delta wave in the United States.  All eyes are on FL.  There is a great deal we can learn from the experience in FL as other states prepare for delta.  The weekly and multi-day reporting of cases and deaths has led to confusion and misinterpretation of the data FL is reporting.  FL is providing daily data on numbers of patients hospitalized and in the ICU.  That data has proven invaluable for tracking the delta outbreak in FL and assessing its severity.  It is time for FL to return to daily reporting for cases and deaths.  The public can only see the weekly reports from FL.  We can indirectly observe daily numbers of cases, deaths, hospitalized and ICU patients through graphical and tabular displays on websites unrelated to the FL DOH. As an example, compare the very different graphical displays of trends in daily deaths from just two commonly used sites: (1). WorldoMeters; and (2). Newsnodes. Until this can be resolved, the best data available for assessing the severity of the delta wave in FL remains the hospitalization and ICU data.

Some Additional Background:

I used the Wayback Machine to  access cached web pages from the state of Florida Department of Health website (weekly reports from May 28 through the present).  I took screenshots of the summary data at the beginning of each report as well as the death curves at the end of each report and pasted them into the attached images for reference (data source). Florida reports cumulative deaths over the course of the entire pandemic through the end of each week and also provides a weekly number of deaths that is much smaller than the weekly difference in the number of cumulative deaths.  I assume that the weekly death number Florida is reporting is the number of deaths with a date of death within the past week to explain this discrepancy. Below is a table that includes the weekly data from the Florida reports that were cached (missing weeks covered by Jun 18-Jul 1 because cached web pages weren't accessible). From Jul. 8 to the present, the number of cumulative deaths increased by 2,603, while the sum of the weekly deaths reported in the weekly Florida reports totaled 706 -- nearly a fourfold difference. To be fair, Florida back corrects the weekly data in its future reports.  The prior weeks’ numbers are continually adjusted upwards in subsequent reports.

None of this was that important because websites that were tracking and reporting on daily FL deaths (e.g., WorldOMeter, BNO, etc.), were reporting daily deaths reported by date (and not by date of death)--at least up until a week ago.

It is interesting to go back to Twitter feeds and monitor the tweets over the past couple of weeks. One source that has been actively reporting data has been BNO. BNO had been tweeting the daily FL numbers (cases, deaths, etc.) each day at about noon. That is typically when the Florida data was made available. The last tweet from BNO reporting daily FL death data was on Aug 6. FL doesn't report death data over the weekend. On Aug 6, BNO also tweeted about the discrepancy between the number of weekly deaths being reported by FL and the calculated increase in cumulative deaths over the week.

There was a tweet on Aug 7 noting the number of cases in FL was the highest ever reported.  On Monday there were a number of tweets and other media reports re the issues between FL DOH and CDC and the misinterpretation of the data FL was reporting.

Starting Monday, Aug 9, FL reporting changed. Instead of reporting data by noon, FL daily updates were being reported in the late evening.  That routine is very different from the usual practice throughout the pandemic. I pulled the daily data being reported on the WorldOMeter website as an example (with a graph of deaths). It appears as if FL is seeing a dramatic decline in deaths over the past week. This is unexpected given the continued rise in the #hospitalized and the #ICU.  I don't recall an example where cases are still rising and have not yet peaked, the number of hospitalized patients and ICU patients is still rising and has not yet peaked, but deaths were already falling dramatically a week earlier.

The number of hospitalized patients or ICU patients is a number that is current for that date (a simple counting of COVID census in the hospital and ICU).  There is no need to back correct daily numbers--those numbers do not change except to correct transcription errors.  Cases can be reported by date of symptom onset, date reported, date test positive, etc.  Deaths can be reported by date of death or date reported.  It looks like FL (at the peak of a serious outbreak) has switched from reporting deaths by date reported to deaths by date of death.  That creates the appearance of sudden decline in deaths.  That is very misleading. This is especially concerning given that many are closely monitoring FL (given that FL is at the leading edge of the delta wave in the US) to better understand how this outbreak might behave in other parts of the country.  GA, for example, provides a toggle in its graphical display to view deaths by date of report or deaths by date of death. Either curve is fine.  What you can't do is present curves where a portion of the curve is daily number of deaths by date reported and then suddenly change at given point in time (in the same graph) to number of deaths by date of death.  That is in effect what appears to be happening. Because FL is reporting weekly reports, it is easy to miss the constant upward adjustment in daily deaths (due to the lags of reporting deaths). Doing this near the peak of an outbreak (when deaths are rapidly accelerating) has a more pronounced effect.

For comparison, here are the daily numbers of deaths reported yesterday for the top dozen states.  And here are the states ranked by the #Hospitalized per million.  

To better illustrate what is happening,  I pulled the daily data on numbers of daily deaths being tracked by a website monitoring FL (WorldOMeter is just one example). I pulled the daily deaths being reported and graphically displayed by WorldoMeter.  I also then pulled the weekly cumulative death totals being reported on WorldOMeter to match the same dates when FL is reporting its weekly data.  I then calculated the net increase in cumulative deaths on WorldOMeter and compared that to the sum of the daily numbers it is reporting.  The numbers match pretty well.  Recognize that FL is only reporting deaths of FL residents in its weekly reports whereas WorldOmeter is reporting all deaths (the cumulative numbers are always a little larger than the FL numbers).  The net weekly cumulative deaths match between WorldOMeter and the FL DOH until the last week (also see graph). This is the reason for the misleading decline in the deaths. FL is not creating the daily death curves (other websites are doing that based upon the numbers FL is reporting each day).

Join our Team

We are actively hiring and would love to connect with talented individuals that want to be on the forefront of change. Contact Us.