The Washington PostDemocracy Dies in Darkness

We stopped tracking coronavirus cases at the University of Florida. Here’s why.

Our covid ‘dashboard’ had reported more cases than any other university in the country. But the data was increasingly unreliable.

Students wear masks on the University of Florida campus in Gainesville on Sept. 1. (Phelan M. Ebenhack/For The Washington Post)
Placeholder while article actions load

For nearly two years, I oversaw the coronavirus “dashboard” at the University of Florida. On that site, we posted the number of tests performed at the university each day, the percentage that were positive and the total number of cases. We also relayed how many students and faculty members were in isolation or quarantine. The dashboard was a tool that people on our campus referred to, and that the national media monitored (along with similar dashboards at hundreds of other schools) as they tracked the coronavirus situation at colleges and universities. In a typical week, some 3,500 people consulted the site, and there were more than 240,000 page views over the course of the pandemic.

Our dashboard attracted an unusual amount of attention because we identified more coronavirus cases than any other higher-education institution in the country — some 14,500 from May 2020, when the university began steps to return to in-person instruction, through 2021.

On Dec. 31, we stopped updating the dashboard because I concluded that the numbers we were posting were no longer useful. This generated complaints, including accusations that we were covering up cases at a time when the situation was worsening by some measures. But so many students and staff were forgoing the school’s official testing site, and other sites that reported to us, in favor of at-home tests — and not reporting the results to us — that we lost confidence that our totals bore any relation to reality. We also knew that many students were experiencing mild covid-19 symptoms but attributing them to allergies or a cold, and so not reporting them. Still others didn’t test because they didn’t want to be barred from dining rooms and classrooms — a trend that increased over the course of the pandemic.

Higher-education reporters took note of our move — as did our community, sometimes vociferously. Among those who complained to me, some argued that transparency is an intrinsically important goal; some also said the dashboard helped them to make decisions about their own behavior (signaling when it was safe to go maskless in public indoor spaces, for example). But transparency is an illusion when the data is bad; likewise, you can’t make good decisions by looking at incomplete or misleading numbers. We at the University of Florida concluded that the pandemic had entered a new chapter, and testing and reporting strategies had to change, too.

Nationally, we are moving from the epidemic phase to the endemic phase, in which vaccinated people are less likely to get infected and far less likely to become seriously ill. Case numbers, even if they could be accurately measured, are far less important than such things as hospitalization numbers paired with vaccination status. As of Jan. 1, we ceded all authority to the Alachua County Department of Health to collect and report coronavirus-related information, ending the categorization of data at the university level.

This is a dangerous time in the pandemic for people like me. Don’t forget us.

From the beginning, the dashboard was integrally connected to the campus’s virus-fighting strategy — known as UF Health Screen, Test & Protect, or STP. (I directed that program from its start in May 2020 through the end of 2021.) We created it to support the return-to-campus effort, and we were affiliated with the county health department. All test results for students, faculty and staff were reported to STP, regardless of where the tests were done or which laboratory did them. We conducted more than 500,000 tests on campus, and private labs also reported to us the test results of university community members.

For the first year of the massive effort, we required testing for select groups that were at higher risk: basically, anybody attending face-to-face classes, living in dorms (or fraternity or sorority houses) or training in a medical field. We also did extensive contact tracing. The data was comprehensive, and, at a time when little was known about what the epidemic trends would be in a variety of university settings, it allowed us to develop knowledge of what worked in our setting.

For instance, in the fall of 2020, we had a steep rise in cases roughly five to seven days after students resumed classes. From the data we could see that most of the transmission was occurring, unsurprisingly, in residence halls. So we moved testing into the dorms; infected individuals were either placed in special isolation housing, asked to quarantine or given the option to leave campus. (Exposed people typically quarantined in place.) The spike lasted about four days, but the isolation plan led to a rapid drop in cases.

The data collected for the dashboard also let us see the effect of things beyond our control. There was another spike on campus in October 2020 that coincided with the opening of the bars frequented by students. More recently, the large (but still smaller overall) rise in cases during the delta variant wave was mitigated by the aggressive vaccine campaign that preceded it.

Among universities, as I mentioned, we led the country in the total number of coronavirus cases — an unenviable category in which to be No. 1. Many commentators inferred that this must have reflected what they saw as a generally disastrous approach to the pandemic in Florida in general. In reality, I would argue, the high numbers stemmed from our aggressive testing program (as many as 3,000 tests a day), which gave us better information than some of our peer institutions had.

I disagree with many of my state’s coronavirus policies: Beginning in the summer of 2021, the university was forbidden to require masks, to mandate testing or to ask students about vaccination status. But the university itself has had decent policies to fight the virus (less so, admittedly, after those state rules went into effect): We tested and followed the numbers; we gave three N95 masks to every member of the faculty and staff (and kept them on offer afterward); we conducted (as mentioned) aggressive contact tracing, isolation and quarantine. We required masks for as long as we were allowed — and voluntary use remains high. We vaccinated high-risk faculty and staff as soon as vaccines were available and vaccinated 19,000 students in the first week students were eligible.

The argument over closing schools is still stuck in 2020

As of this writing, we are unaware of any deaths that occurred among our employees and students from a work or classroom exposure. (Overall student enrollment at the university is about 53,000.) To my knowledge there were two deaths of staff members that were due to exposures off campus.

We’d already begun to transition the services my office provided — testing, vaccination and so on — back to the county and student services in November because of decreasing cases and increasing endemicity. But we retained the ability to rapidly scale testing back up. And when the omicron variant hit, we scaled up big time: We went from about 1,000 tests per week in early December to 1,500 daily now. Of those, roughly 20 to 25 percent are positive. Life on campus, however, is surprisingly normal. Students who test positive are isolated or leave campus, but they can continue their coursework online. A few public events have been canceled or delayed, but many continue.

How would numbers on a dashboard change behavior these days? Everyone should wear a proper mask when indoors, get vaccinated and boosted, and stay home if they are sick. You don’t need to track cases day by day to grasp these fundamentals (assuming you could track them accurately).

We absolutely need more data in the fight against the coronavirus. We need to know who is being hospitalized for covid as opposed to with it. We will need to know how future variants differ in their ability to cause severe disease. What we don’t need is a catchall of whatever information happens to be available — including wildly imprecise case counts that include people with symptoms resembling a cold. That’s why we decided to stop updating the dashboard at the University of Florida.