We get a lot of questions about data.
Because we only use official, Florida Department of Health-published data, we are limited by the restrictions they place on the data.
They describe some of those issues here, in a document I wrote while still at DOH that has since made been revised, but still leaves much out (in addition to some information being outdated or incorrect).
I’ve added some warnings about the data to the Florida COVID Action dashboard, and am working on a more complete document in the next few weeks. I’ve emailed the DOH communications line multiple times with questions about data, and still have not received a response.
Here are two of the major warnings regarding more recent changes in data, though we’ll expand upon them in future posts.
All fields with the PUI (Person Under Investigation) prefix:
The PUI definition in DOH’s document is wrong. The state stopped updating the PUI data in late May. As described in their documentation, a Person Under Investigation (PUI) is “any person who has been or is waiting to be tested.”
By that definition, the total PUI’s should, at a minimum, equal the number of people DOH claims have been tested, and the number of “pending” tests from BPHL.
But it’s not even close. There was a period of time after I was fired where you could use the PUI data to calculate the true percent positivity, since PUI’s were individual people (cases) and total tests now included duplicate negatives.
That PUI data stopped getting updated around June 1 (not all coincidentally after I called out DOH about how the “total people tested” figure was equal to total tests, not people).
And it’s not just the total number of people tested that’s now static.
Now the all of the data with the PUI prefix is being updated with small numbers — some days equal only to the number of new cases, other days numbers we can’t figure out where they came from.
That means there’s no data available about the number of people being tested, their age, gender, race, ethnicity OR geography. Those are key elements in studying trends in cases, testing and deaths, as well as the quality of the data being reported.
Because the totals of the demographic groups (so the sum of male, female and unknown gender, or the sum of all age groups) still equals the PUI total field, we know there’s something wrong on the back-end.
DOH has told reporters multiple times since then that the data is right. If so, that means the state has tested half the amount of people they claim to have tested. But given the trajectory of testing (people and tests), it’s more likely someone messed up my script and either doesn’t know how to fix it or is deliberately hiding that data.
I believe it more likely someone messed up and hasn’t fixed it, giving the lack of experience of the people who replaced me in my role as the coder, programmer, analyst, researcher, html and java-writer, etc etc over the dashboard and public data distribution.
DOH has tried to make clear its limitations on hospitalization data for months. In the data definitions document they provide, they state:
“‘Hospitalizations’ is a count of all laboratory confirmed cases in which an inpatient hospitalization occurred at any time during illness. These people may no longer be hospitalized. This number does not represent the number of COVID-19 positive persons currently hospitalized.”DOH Data Definitions Document
Because Florida doesn’t track what’s commonly referred to as “probable” cases and deaths, we can’t really come up with a good number for hospitalization data, at least not “active” hospitalizations.
The CDC provided guidance about reporting probable cases back in April 2020. Here’s the information they provide on their website:
A probable case or death is defined as:
- A person meeting clinical criteria AND epidemiologic evidence with no confirmatory laboratory testing performed for COVID-19;
- A person meeting presumptive laboratory evidence AND either clinical criteria OR epidemiologic evidence;
- A person meeting vital records criteria with no confirmatory laboratory testing performed for COVID-19.
Here are our warnings from our dashboard:
Warning about ER, In-patient admissions and death information related to cases
There is no active count of COVID-19 positive persons admitted to a hospital ER or as an in-patient in the state of Florida.
Therefore, any statement regarding the current trend in ER and hospital admissions (and deaths-more on that below) are entirenly non-evidence based.
The Florida Department of Health (DOH) has never provided data about the number of people currently hospitalized or in the emergency room who are COVID-19 positive. They do not provide dates for when COVID-19 persons were admitted to or discharged from the ER or as in-patients to a hospital.
The only hospital surveillance data available to the public is the AHCA current bed availability data (see note about that data to the left) and the Florida Metrics data published by DOH every Sunday, which counts the total number of legal Florida residents who visited an ER in the state of Florida with symptoms matching the key terms of influenza and COVID-19 (reported by week).
Additionally, DOH does not provide the date of death for any case. The chart on the now-disabled DOH dshboard showing resident deaths by date is just that – residents only by date of death for cases where a death certificate has been received and the date of death verified by the state.
The CDC warns “delays in reporting a person’s death to the state can range from 1 week to 8 weeks or more, depending on the jurisdiction and cause of death.”
According to internal DOH records, deaths reported by DOH were reported an average 7-21 days after the COVID-19 positive person died (date from March through mid-June_, with some deaths reported as late as 44 days after death.
WARNING ABOUT HOSPITAL ICU BED AVAILABITY DATA:
AHCA provides the hospital data you see on this page.
Consequently, the number of available beds shown here likely includes beds that are currently occupied, but that the DeSantis administration has determined by an un-disclosed metric could be made available by removing the person occupying it.
We continue to monitor this situation for changes and will post any new information as it becomes available.