Tennessee Surpassed Best-Case Scenario Under Vanderbilt’s COVID-19 Modeling

May 14, 2020

Public health nurses and Tennessee National Guard medics take swabs from drivers in McMinn County.
Credit TN Dept. of Health via Twitter

Pandemic modeling from Vanderbilt University finds Tennessee was able to drive down the COVID-19 transmission rate well ahead of the initial projections made on April 10. As a result, the number of people simultaneously hospitalized has plateaued below 300 statewide.

A month ago, even if the state made “continued progress” to slow the spread of the virus, concurrent hospitalizations would have at least hit 1,200 by mid-May. But Vanderbilt modelers say the state started slowing the spread of COVID-19 a month earlier than expected, meaning that on average not every person with a positive case was getting at least one other person sick.

“This effectively reduced the growth of COVID-19, as it meant that cases (and hospitalizations) were no longer growing exponentially,” health policy professor John Graves writes in the updated modeling report released Thursday morning.

Vanderbilt isn’t making many new predictions, though Graves says the modeling has been revised. He says it’s difficult to make predictions as the state has expanded testing so rapidly to people who are showing no symptoms.

Hospitalizations have still climbed, especially in Nashville and Memphis, which are home to the state’s largest hospitals. In the Nashville area, hospitalizations have nearly doubled since mid-April, but still remain below 150 at one time. Graves, who directs the Center for Health Economic Modeling at Vanderbilt, says hospital capacity is not stressed.

While backing off bold predictions, Graves says it will be another week or more before the state can begin to see the affects of easing restrictions on daily life.

“We believe it is too early to assess the impact of businesses reopening across the state or of more Tennesseans resuming activities outside their homes,” Graves writes. “Therefore, data presented here should be considered a new ‘baseline’ for monitoring changes moving forward.”

‘The Point Of Modeling Is Not To Be A Fortune Teller’

In an interview with WPLN News, Graves addressed questions about how his model has performed since its initial release. He says he’s “heartened that we weren’t optimistic enough” and that he’s proud of “how quickly Tennesseans rose to the occasion.”

“But that’s not to say that we’re going to stay at the situation we are right now,” he says. “It’s totally unknown.”

The point of modeling is not to be a “fortune teller,” Graves says. It’s to help public officials understand various scenarios, ranging from the best to the worst case.

“If you want to know the future, you’re better off going to a fortune teller, not using a pandemic model,” Graves says. “So our best-case scenario happened a month earlier than we anticipated, which we are ecstatic about.”

Graves says modeling the pandemic in real-time has been a bit like trying to put a fitted sheet onto a 100-corner bed. As soon as one data point becomes clear, another becomes more fuzzy — like the introduction of targeted mass testing. The events revealed huge numbers of asymptomatic cases in meatpacking plants and prisons but relatively few hospitalizations.

The revelations forced Graves to change estimates built into his model, such as the number of cases presumed to be going undetected and the overall hospitalization rate.

“Doing that in real time has been in challenge, especially when you’re dealing with exponential growth as you do with these novel infectious diseases,” he says.

He says the state is now “walking a tightrope” because the transmission rate is just below 1.0, where not every positive case results in another positive case. Moving just a few points higher results in some of the original runaway scenarios.