The Impact of Re-Opening Now for Tennessee

On April 27th, the state of Tennessee began re-opening. This has started small, with restaurants and retail being allowed to open for business at 50% capacity. Could a second wave of exponential growth begin again as a result? A simple model used in estimating the spread of epidemics is known as SIR, as it looks at the relative amount of Susceptible, Infected, and Removed (either recovered or dead) members of a population. Using this model, and calibrating it to the spread of COVID-19 in March, before the April lock down, gives insights into the potential increase in the spread of the disease as the state of Tennessee began re-opening on April 27th. One thing to note is that the virus continued to spread in TN during the month of April. I assume only a slight increase in the rate of transmission from the April lock down, roughly half the rate of spread in March. This results in going back on an exponential curve, albeit one that grows slowly. However, even a slow exponential curve results in significant increases in the number of active cases, although it takes a while to notice the effect. This is an optimistic forecast, as I am trying to see how good the re-opening can possibly go. See the graph below.

This starts slowly, but ramps up in June and July. As of April 30, there were slightly more than 5,000 active cases. I assume it will take a week before the additional spread is picked up in official statistics. One key date I will be looking at in May to see if this is on track is May 17, as the model projects 10,000 active cases on that day. If we do not hit that target by late in the month, that is cause for optimism going forward. I will be updating this and refining it along the way. If this forecast is close to accurate, we will be shutting down again this summer. I hope things go better than this.

For those interested in mortality, the total number of infected by mid-November is estimated to be 3.7 million additional people. At a mortality rate of 1%, that will mean 37,000 additional deaths by that point. This assumes no additional shutdowns, and that the rate of transmission stays low as the re-opening gains steam. That is likely optimistic as well.

For those interested in the details of my assumptions, read on. There are still many unknowns in modeling the spread of COVID-19, but we are beginning to get some clues. We know that the real spread of the disease is understated by the official statistics. As pointed out to me by my friend and colleague Rich Harwin, a recent study in Los Angeles County indicates that 4% of the population had antibodies for COVID-19. That is far more than the reported number of cases in that county. For this model, I assume 25% of the population is removed from the pool of people susceptible to infection due to prior exposure. This is based on work done at the University of Georgia – http://2019-coronavirus-tracker.com/nowcast.html. I think this may be a little high, but I use it to model the optimistic case.

The use of the SIR model was suggested to me by Jim Epperson, a grad school professor of mine. I assume that the time the virus stays in the body is fixed at 14 days, which is based on information I have seen and which others have used in modeling. Looking at the increase in the number of reported cases in Tennessee from March 11 to March 31 (the date of the statewide shut down), I estimated that the rate of transmission during this period is 4.9. There are estimates of the transmission rate of the disease, which is the average number of people to which an infected person transmits the disease. Theses range anywhere from 3 to 6.

I assume the slow re-opening, wearing of masks by many people, and general caution drops the rate of transmission in half, to 2.4. This is not conservative. Applying the SIR model to the last four weeks of TN data, the transmission rate appears to be a relatively high 1.75. Increasing the transmission rate to 2.4 to account for larger gatherings seems like it might be on the low end of what may actually happen. Over the last week before April 27th, the percentage increase in number of active cases was 5%. My forecast only increases this to between 6%, which then decreases over time as the number of people susceptible to infection shrinks. This is not much of an increase from 5% in late April.