WLRN depends on donors to remain South Florida’s leading nonprofit, most trusted source of news and information. Support our mission by giving monthly as a sustaining member of Friends of WLRN or make a one-time donation of your choice. Thank you. Click here to give.
One of the most important things for the public during the COVID-19 public health emergency is access to reliable information from the state government.
Data can help the public weigh its need to respond to what is going on during a pandemic. When mapped out it can clearly show where cases are popping up, while demographic information can signal which populations might be most at risk from the spread of the novel coronavirus.
But what if the government is sharing incomplete, flawed, or otherwise questionable data?
A concerned citizen contacted WLRN this week to share information they have been gathering that raises potential red flags about data the Florida Department of Health and other statewide agencies have been sharing about the crisis. The person wanted to remain anonymous because they are not authorized to talk about the subject, but they did send us links, screenshots and pointed us to spreadsheets that, taken together, raise questions about the state data.
Through the tip and further reporting, we have identified a series of instances where information about reported COVID-19 cases altogether disappeared, was removed and then re-added, or might have been altered in some kind of fundamental way, to the point where it is unclear what’s actually being shown to the public.
“The state data has not been transparent, at the very least. You see some cases that are there in the dataset one day and gone the next day, with no explanation for why that’s happening,” said the tipster.
The data itself has been released in a dizzying variety of formats. At one point it was simple graphic tables, then Excel spreadsheets, then PDF documents. The columns of what information is being shown is inconsistent, making it exceedingly difficult to track particular cases, and track nuggets of information that are released through the different formats.
For instance, Thursday’s data contained a column for people who have died. Friday’s version puts that in a separate table entirely. The case numbers for both days don’t match up with one another. One day the case numbers are sorted in chronological order, the next by county.
“What that does is just create another layer of complication when you’re trying to track this on a statewide, or even a local basis,” said the caller.
COVID numbers behaving strangely
In the early days of the pandemic in Florida, when the numbers were somewhat low, each daily announcement from the Florida Department of Health included a table that listed the location and travel information for each case, along with demographic information about who had contracted COVID-19.
On March 12, the Florida Department of Health listed several new cases on its daily press release. Included in it were two cases in Sarasota County -- a 50-year-old man and a 70-year-old man; both were out of state residents and it was “unclear” whether they were travel related.
A 24-year-old woman was added in Alachua County. She had traveled to Europe, according to the release. Local media in Alachua County reported conflicting accounts of the person's sex.
The following day, on March 13, the 24-year-old woman from Alachua County no longer appears on the table. But suddenly a 20-year-old male in Alachua County appears, without explanation.
The day after that, on March 14, the 50 and 70 year-old men from Sarasota County are no longer on the chart. Yet a 50-year-old female in Sarasota County suddenly appears. In Alachua County, the 20-year-old male who was just added the previous day is no longer there. Instead, a 24-year-old male from Alachua County is listed.
To recap: The original Alachua County case — announced on March 13 — was a 24-year-old female. Then this person disappeared from the chart, followed by a 20-year-old male who disappeared from the data after one day. Then, a 24-year-old male appears, with no reference to whatever happened to the 24-year-old female who was reportedly the first case in the county.
This past Monday, on March 16, the state migrated from these tables to a statewide map, showing the case load by county. By inspecting the data on the back end of the website, it was possible to access the spreadsheet the map was using to populate the map.
A spreadsheet downloaded that day showed three Broward County cases that were on the table the previous day were now missing: a 41-year-old male whose case was unclear if he traveled; a 49-year-old female whose case was travel-related; and a 48-year-old female whose case was travel-related.
In the most recently released data, information matching the 48 or 49-year-olds in Broward County is still nowhere to be found; A 49-year-old female from Broward who did not travel is listed, as well as a 48-year-old male — not a female.
It is unclear if these are the same people with updated data points, or if these are entirely different cases.
On Monday, the data showed a 22-year-old female who lived in Broward County who had recently traveled to Ireland. By Wednesday, that data was not reflected in data powering the statewide mapping tool. And finally on Friday, her information appeared once again.
The 50 and 70-year-old males from Sarasota County who disappeared on March 14 have now reappeared in the data. The 20-year-old from Alachua County who was listed on March 13 has disappeared from the dataset altogether.
If all of this sounds incredibly confusing, that’s because it is.
These are just a small number of inconsistencies that we found analyzing day-to-day data released by the state. There could well be more, but analyzing the data is complicated by the constantly changing formats through which the data is released.
'Consistency is key'
“Consistency is key when you’re reporting this kind of public health data,” said Nora Friedman, a senior data analyst at the Chicago Center for HIV Elimination, under the University of Chicago hospital system. “For new things that you’re tracking it’s normal to kind of track and release data on the fly, but a few weeks into this, it surprises me that there’s not one template they’re going by.”
Friedman said she would guess that the problems with the data can be chalked up to confusion rather than deliberate misdirection. The state might actually need some help analyzing and reporting the data to the public, she suggested.
"We're just trying to evolve and get the information out as quickly as possible, and that means we're fine- tuning the way that we present the data to the public," a representative for the Joint Information Center on COVID-19 for the State of Florida told WLRN.
The representative said that the inconsistencies can be chalked up to the fact that the daily reports say that the data is "provisional and subject to change." Notably, the statewide map that visualizes the data present in the daily reports does not contain that provision.
"When we get a reported case, there is an epidemiological investigation that takes place. We try our best to figure out where the person might have contracted it, and move down the list contacting everyone we think that person might have been in contact with since contracting it," said the representative. "Sometimes that information can lead to changes after the fact of what you see when you're looking at the data from one day to the next."
For example, sometimes a person tests positive in one county, and later authorities learn that the person is a resident of another county. That could initiate a change in where that particular case is listed, leading to be removed from one county and perhaps added to another county, without any public announcement about why that has occurred.
Friedman and the person who initially contacted WLRN said that this leaves open questions about data transparency that the public and local government agencies are relying on.
“The one thing I’ll say is that it would be good to have some kind of corrections table for when someone is removed, to say ‘X person was removed for X reason,’” said Friedman. “That would help the public have more faith in the data, and that what they’re looking at is the most accurate and up-to-date information available.”