A deeper dive into Ohio State’s top-rated COVID-19 testing data dashboard

In the month or so since The Ohio State University published a dashboard of aggregate COVID-19 testing data, the page has received more than 1.1 million unique views, quickly becoming one of the university’s most popular websites. Among those who have taken notice: “We Rate Covid Dashboards,” a panel of academic and medical experts rating university dashboards across the country.

Ohio State’s dashboard has received the only “A+” grade from the panel of experts, who rate dashboards on a variety of criteria that include frequency of updates, what data are reported and whether they’re easy to read. As of Oct. 1, the panel had published ratings of 211 dashboards.

“Their mission behind this whole effort is to encourage colleges and universities to be as transparent as possible with their COVID-19 strategies, including the community impact,” said Eric Mayberry, senior manager of data analytics in Ohio State’s Office of the Chief Information Officer and a co-creator of Ohio State’s dashboard.

The dashboard contains key measures for the state of Ohio – daily cases (also available by county) and hospital capacity among them – and student test result data for Ohio State presented in three formats: single day, seven-day average and cumulative, plus cumulative results of employee tests. The university also reports information about isolation and quarantine bed usage, capacity and availability, and indicators of where things stand with personal protective equipment and enhanced cleaning – key mitigation strategies for the pandemic.

At this point in the semester, the trend line of student testing data on the dashboard clearly shows Ohio State has bent its curve in the proper direction: down. Based on the latest data available today, Oct. 2, the seven-day average of positive student test results stands at 0.7% for students living on campus and 1.6% for off-campus students. The trends represent a steady decline in the seven-day average positivity rate over the month of September.

In addition, the U.S. Centers for Disease Control and Prevention recently reported that Ohio State Wexner Medical Center’s pandemic protocols – masking, hand hygiene, physical distancing, and testing of symptomatic patients and staff – were effective at protecting health care providers from contracting the virus. Fewer than 1% of Ohio State providers tested positive for SARS-CoV-2 antibodies in a multicenter study.

Understanding dashboard data

Leanne Stanley, a consulting research statistician in the Office of Institutional Research and Planning, created the dashboard with Mayberry and Henry Zheng, associate vice president for strategic analytics. They collaborated with a team of Ohio State public health, technology and analytics experts, with guidance from university leaders who strongly supported data-informed decision-making and data transparency, Zheng said.

“Many universities have public dashboards, and there are lots of reasons to do it, but probably the first and foremost is to be transparent, trying to be able to show the public where we stand in terms of the epidemic on campus and our response,” said William Miller, professor of epidemiology and senior associate dean of research in Ohio State’s College of Public Health, who helped lead development of the dashboard.

The seven-day average of positive student test results provides the most useful measurement of cases on and off campus, Miller said.

“It’s a rolling average so that today is a seven-day average for the seven days prior to that. It incorporates both the single-day information and … to give you a better indication of where we’re really at, smooths out the variability that you might have,” said Miller, also a member of the Safe Campus and Scientific Advisory Subgroup of the COVID-19 Transition Task Force.

Cumulative data – for students and university employees – is depicted in circle graphs on the dashboard. Since surveillance testing began on Aug. 14, Ohio State has conducted more than 100,000 tests of students and almost 2,400 tests of employees.

“This gives an overall sense of where things are, and the single-day result is an indicator just day by day of the number of tests that have been performed and the positivity rate for that given day,” Miller said. “And that will vary a lot over the course of a week, in part because we have many fewer tests on the weekend. And when you have fewer tests, you have much more variability, potentially, in the test results.”

Miller cautioned against making broad assumptions about any individual statistic. For example, though the daily surveillance testing is an ongoing program, there are other reasons for students to be tested: when they feel sick, if they’ve been identified through contact tracing as likely being exposed to the virus, or when a cluster of positive cases calls for targeted testing of the students living in the affected environment. As for the surveillance itself: All students at all ranks who live on campus are tested weekly, and a random sample of all undergraduate students living off campus is taken each week, which averages to about once per month for individual students.

“Mitigation and surveillance testing is trying to keep the epidemic under control and also give us a very clear look into it,” Miller said. “When we do targeted testing, your positivity rate may be higher than just when you’re testing everyone.

“So when you look from one day to the next, the thing that’s difficult to interpret a little bit is if there was a change in the testing procedures – so you can sometimes have a bump in the positivity rate because we did a whole bunch of testing related to a small cluster of cases and we found several more cases,” he said. “That’s one reason to be careful, especially when you look at the single-day rates. I think that’s really the biggest thing to be cautious of when you’re interpreting the tests of students – the reason for testing doesn’t come through in the number.”

There is a measure for the state of Ohio included on the dashboard that eventually will be added for the university, as well: the R(t), or reproduction number. Essentially, this is a gauge of whether the epidemic is expanding or contracting based on the estimate of how many other people each person who tests positive for COVID-19 infects. It takes some time to arrive at a meaningful figure because the number of positive cases over a period of time is fed into an algorithm to produce the estimate of whether the R(t) is below 1, at 1 or above 1.

“Anything above 1 is concerning,” Miller said. “If the R(t) is greater than 1, that means that the epidemic will be expanding because one person is infecting more than one other person, on average. If the R(t) is at 1, it’s basically under control, but not shrinking. And if the R(t) is less than 1, then it is going to be contracting as long as something doesn’t change to perturb it.”