Mortality, Hospitalization and Survival rates adjusted for reporting lag, July 24, 2020

Data and charts by both the Wolf and the Badger

As cases continue to increase in Florida (12,444 new cases for today), a lot of discussion lately has focused on the mortality rate (and to a lesser extent, the hospitalization and survival rates).

We consider a 30-day lag in our hospitalization data shown below, which is based on the Florida Department of Health’s case line data.

If we look at the “survival” rate (cases where no report of an emergency room/department, in-patient hospitalization, or death is recorded) using a 45-day lag, we get a clearer picture about the rates of “case severity” across age groups.

We used the most severe healthcare/death outcome to assign cases below, with a hierarchy of ER/ED visit/in-patient hospitalization/death.

So if you die, you would only be represented in the “died” section. If you went to the ER and were also admitted to the hospital, you would be put in the in-patient hospitalized category, as that is generally considered “more serious” than an ER/ED visit.

If you were never admitted to the ER or as an in-patient to a hospital and you did not die, you would be placed in the “survived with no hospitalization on record” category.

The chart below is interactive, and you can remove age groups from the display by clicking the group on the chart legend. To add it back, just click it again. To reset the chart at any time, double click anywhere on the plot.

We’re also working on creating a new experience for users to improve optimization, add new data resources, and make the interface a more natural reading environment. Here are some of the sub-dashboards that will be included, which are now public and you are free to use.

Please note: these are designed for desktop experience and are not mobile compatible yet.

Florida Case Overview:

Health and Hospitals Dashboard:

Cumulative cases by city, by week (updated every Sunday):

Long term care/assisted living facility cases and deaths:

Missing Deaths in Leon County

Article by the Wolf

In Florida Department of Law Enforcement’s records, FDLE counts 36 total COVID-19 related deaths in Leon County .

That’s considerably more than the 20 COVID-19 deaths DOH currently reported.

Of course, FDLE receives ME reports and responds to emergency calls and assigns those to the county of death. With an intensely rural surrounding area, many of Leon county’s deaths aren’t residents of Leon county, meaning they’d be assigned to the county where they live in DOH records, according to DOH.

The 16 additional deaths that FDLE reports can mostly be attributed to DOH assigning the cases to the victim’s county of residence.

However, there are quite a few in the FDLE report from July 22 that don’t appear in the county of residence for the victim, which is also listed in the FDLE data alongisde the place of death. Two deaths include out-of-state residents from Georgia who died in Tallahassee hospitals, and another two do not have a county listed in the FDLE report, but their demographic profiles do not fit anyone who died from surrounding counties at that time.

Missing deaths by county, date of death, and age, race and gender.

Leon County, 4/25/20, 62 year old black female.
Liberty County, 6/1/20, 63 year old black male.
Liberty County, 7/12/20, 60 year old white female.
Gadsden County, 7/11/20, 86 year old black female.
Unknown County, 5/21/20, 80 year old black male
Unknown County, 5/27/20, 87 year old black female

Here’s an example of how the FDLE and DOH data don’t line up through last week:

CountyFDLE DeathsDOH Deaths
Indian River1828
Palm Beach479724
St. Johns817
St. Lucie5088
Santa Rosa418

Part of our tracking project here at Florida COVID Action is to chase down those disparate figures and see which deaths were simply moved to another county, or never included in DOH data at all (and even removed from previous DOH data in some cases)

You can support us in our ongoing efforts as we build out our team and our software systems.

We’ve also been working on building an entirely new experience for the public, which will help solve our issues with load time on our current dashboard system. Here’s a preview of what’s to come:

Data Apparition at Liberty Prison from June 2020.

Article by the Wolf

Acronyms used in this article:

DOH = Florida Department of Health
DOC = Florida Department of Corrections
ME = Medical Examiner’s office

Liberty CI = Liberty Correctional Institution

Data about cases and deaths in prisons frustrate our team to no end.

To start, DOH’s data for correctional cases and deaths almost never match what the DOC is reporting.

For example, DOC shows 34 “COVID-19 Related Inmate Deaths,” while DOH shows 29. There are 115 total deaths across DOC’s facilities listed as “natural” or “pending” since March 1. All COVID-19 inmate deaths verified through our project are listed as “natural” in DOC records.

But at least one is missing from the list of DOC deaths, or the list of DOH deaths.

An inmate at Liberty CI — John Doe for now — was transferred to Tallahassee Memorial Healthcare where he died on June 1. He was a 53-year-old black male, according to FDLE and ME records.

In FDLE records, his death is attributed to Leon county where he died, since the ME’s report was issued from Leon County. Typically DOH assigns deaths based on where the person lived, not where he or she died, and that’s true of correctional facilities in most cases, as well, we’ve learned.

However, in DOH’s case line data, no 53-year-old black male has died in either of those counties, and no one is reported to have died in Liberty on June 1, according to DOH’s deaths-by-date data.

Even though there was an article about a death at Liberty CI later in June (a man that does not fit the description of the inmate referenced above), as of this writing, DOC shows no deaths at Liberty CI since February.

From the FDLE report:

“The decedent …[] …was an inmate at the Liberty Correctional Institute. There had been numerous cases of COVID-19 positive inmates at the institute. He had been tested ~2 weeks ago and was negative at that time. For the last 3 days he had been having fevers, dry cough, shortness of breath, headaches, body aches, palpitations, and some nausea. He was taken to the Calhoun Liberty Emergency Room where a chest x-ray revealed bilateral lobar pneumonias. He was transported to Tallahassee Memorial Healthcare where a rapid PCR test for COVID-19 was positive. His medical condition continued to deteriorate and he was eventually pronounced.”

His date of death is listed as June 1, but the report of his death was not received by FDLE until June 8.

So… where’d he go? Why isn’t his death included in DOC’s records or DOH’s records? How many other “natural” and “pending” deaths in Florida’s prisons aren’t being reported by DOC, DOH or both? And how many people like our John Doe died and were never even recorded at all?

Perhaps if DOH could start responding to our multiple emails, this data apparition could be resolved.

You can download a redacted version of the FDLE data report for July 22, 2020 here. Please note, however, that not all deaths reported by FDLE/ME’s offices are reported in DOH records, and vice versa.

API feed changes to key data on our site

Update by the Wolf

We’ve moved our key API data feeds to a new account to help us with data automation and system optimization.

The updates are reflected in our data library, as well.

If you’d like to join our data users listserv to receive key updates about changes to our data schema, you can subscribe to our blog or email us at:

Our new data hub, where all data on our dashboard can be downloaded, is housed here.

Our updated data feed links are below.

Stand-alone Florida COVID Action APIs:

LIVE! COVID-19 Hospitalization Data from AHCA

Article by the Wolf

You asked — we delivered.

The Florida Agency for Healthcare Administration (AHCA – a separate entity from DOH) quietly launched its new actively-hospitalized COVID-19 cases/hospital capacity data two days ago.

Naturally, this new data was published through a Tableau view-table, and not an easy-to-ready-and-add-to-other-public-platforms format.

So we wrote a script (a few scripts, actually) to pull said data from the website and to deliver to you both on our main dashboard (under “Health and Hospitals” and through a public REST service!

Of course this means we had to move some stuff around on the dashboard (cases by day moved to “Testing” tab), and we’re keeping the live-count of available hospital beds from the old AHCA API on the home page for now. We’re working on getting automation set for the new data, but we won’t be running updates every 15 minutes like the new AHCA data, or even every 30 minutes like the old AHCA data.

Of course, if AHCA finds it in their hearts to produce their OWN public REST service, like the one they helped the Department of Emergency Management and ESRI build back in April, we won’t have to purchase the additional server time or worry about the changing web token.

You can click the blue tweet button below to ask AHCA to generate the API so the public can have access to the live counts through multiple platforms.

Hey, @AHCA_FL! Please make your #COVID19 #hospital #data available through a public API! Pretty please, #AHCA_FL with cherries on top!!! @GeoRebekah @HealthyFla #AHCA #Florida

For now, we’re working to clean up the process, finish our bridge script, and work out an update schedule.

Rest link:

Item page:


On the fatality rate of COVID-19

Article by the Badger

The fatality rate is the number of cases that result in death divided by the total number of cases (for each gender-age group) excluding cases within the past 45 days (or at the time of writing this, May 25). The basic idea is to only include cases that are not time-censored; that is, they have had enough time to develop until recovery or death.

Despite the repeated assertions by certain politicians that Florida has the best data (still waiting on hospitalizations, Ron), it is actually lacking in many areas that hinder a more complete understanding of the situation. For example, case data is overwritten each day with no record of the days between when Died field goes from ‘No’ to ‘Yes’ (time until death). We cannot even be entirely sure that cases are the same from day to day as there is no unique id and the count of cases by case date changes each day (always lower for some reason).

The only way to estimate time from case until reported death is to look at the change in the number of deaths from one day’s case file to the next. Unless you have been saving each day’s file (we have, as have many others), it would be impossible to know how many new deaths were reported and get an estimate for when a certain proportion of deaths have come in. This is where the 45 day lag comes in.

Now that I’ve wasted all your time, I’ll get to the point. We just look at the percentage of deaths (by case date) for each consecutive case file. Using case dates that are at least 60 days old, we looked at how many of the total/almost total deaths that have been reported by day 45, which is about 96% of them.