This week marks a major improvement in the total tests data available for the state of Florida. For the first time, Florida is publishing a full time series for total tests in test encounters. This addition to Florida’s public COVID-19 data allows us to make changes to the state’s data in our original API field for total tests data, totalTestResults, to better reflect the reality of test utilization in Florida as we head toward winter.
The COVID Tracking Project’s
totalTestResults field for Florida will now reflect test encounters, or the number of individuals tested per day, instead of the number of unique individuals ever tested in the state. This change will increase Florida’s cumulative figures reported in the totalTestResults field by 3.7 million tests. These numbers are distributed over the state’s entire time series, so changes on any one day are smaller; for example, the change increased the daily rise in totalTestResults on October 26 by 24,575 tests. The unique people tested metric that we previously reported in totalTestResults was also reported in totalTestsPeopleViral, and will continue to be reported in this field and on our website for as long as we have access to this data.
Why change total test units?
We’ve written extensively about the shift in our preferred total test metrics, but in a nutshell, we’re making these changes to better reflect the realities of US testing strategies and utilization, which have changed since the pandemic’s early months. Repeat (or “serial”) testing is now a major component of many states’ testing strategies, and we have prioritized total test units that capture these tests. Since late summer, we have therefore prioritized total testing metrics in the following order: test encounters (unique people tested per day), specimens (all samples tested), and unique people tested. A full list of states whose totalTestResults we’ve switched and their units is available on our total tests page.
What this means for Florida’s data
The big increase in Florida’s cumulative testing numbers in totalTestResults you’ll see today is driven by repeat testing: Those 3.7 million additional tests represent individuals who received tests on multiple different days, but weren’t picked up in Florida’s totalTestResults before this switch, since this field was previously filled from a unique people metric. The new data we’re using in our totalTestResults field, made available on Florida’s open data hub on Monday, picks up those repeat tests by counting the number of people tested per day (excluding those who test positive repeatedly).
Though Florida’s totalTestResults will now reflect test encounters, you can still get the data for all the cumulative testing totals Florida provides from our other API fields. We will continue to capture the number of unique people tested per lab in the totalTestsViral field of the API. (As of today, Florida’s reporting of total unique people tested has changed: We now capture them from a query to Florida’s COVID-19 Cases ArcGIS layer by summing the T_NegRes and C_AllResTypes fields.)
Finally, we’ve made one more change to our case data for Florida to ensure that cases are aligned as closely as possible with the total tests data the state provides. The COVID Tracking Project has chosen not to calculate test positivity, but we know that many users of our data calculate test positivity using the data we compile from states and territories. To avoid introducing additional noise into these calculations, we are making a change in our case data policy for Florida.
Previously, we summed Florida’s resident and non-resident cases and tests in our case and total test figures. The new figures for test encounters, however, only include residents. To align our case and test reporting, we are now capturing and reporting all cases and tests for Florida residents only. This change is in line with CDC recommendations for reporting on nationally notifiable diseases. Because Florida separates out cases identified by antigen and PCR testing, this affects all three of our cases fields: probableCases, confirmedCases, and positive. Our total cases number reported in Florida will decrease today by 10,024 (1.28 percent of total cases reported) with the removal of non-resident cases. We will also be removing 22,828 non-residents from the unique people tested metric in Florida, which includes both PCR and antigen tests, to reflect this new policy.
A reporting note for close readers of Florida data
Florida has previously provided numbers in its state reports that captured serial testing, but we’ve been unable to use them for several reasons: Test encounters were reported only as percentages, which prevented us from using exact numbers, and they were reported only as a daily number rather than as a cumulative one. Another chart on the state report, “Number and percent of Florida residents with positive test results,” provided raw numbers but only for each day rather than cumulatively—preventing us from aggregating an up-to-date cumulative number, since daily data is retroactively altered as a part of normal data cleaning for epidemiological datasets.
Finally, Florida also reports a testing number at the bottom of its “testing by lab chart,” which we store in our totalTestsViral field, but we have elected not to switch to this metric because it doesn’t align with any of our existing total tests data definitions. According to responses we’ve received from Florida’s state health department, the number at the bottom of this chart represents the number of unique individuals tested per lab—so if an individual was tested at two different labs, they’d be counted twice in the figure, but if an individual is tested at the same lab they’re counted only once. Though the number of unique people tested per lab reflects repeat testing to a greater degree than unique people, it is not comparable to total test metrics reported by other jurisdictions. For totalTestResults, we only use testing figures with standardized definitions, since it is the field we sum to assemble our picture of national testing.
Many states have moved toward greater transparency about their test data reporting methods, and we’re making changes to better represent what they publish. We’re also introducing a “new” way some states are counting tests—one we think all states and territories would be wise to embrace.