Medical Examiner

Georgia’s Reopening Depended on Missing Data

There’s a lag between when COVID-19 is found and when we know about it.

A group of people sit and have a picnic in the park
Historic Fourth Ward Park in Atlanta on Sunday. Kevin C. Cox/Getty Images

Georgia Gov. Brian Kemp ended mandatory shelter-at-home orders on April 30. His reasoning for doing so was based in part on the apparent decline in COVID-19 cases reported by the Georgia Department of Public Health. But was Kemp right? Was the number of COVID-19 cases in Georgia really in a decline in the days leading up to the April 30 reopening? The data Kemp consulted to make his decision show an apparent decline. But a later, more accurate picture of the situation in the state reveals something worrisome: The number of cases in Georgia was not actually declining when Kemp lifted his order. Georgia was, and still is, in a “plateau”-like trajectory, with cases fluctuating day to day, with no strong evidence of a systematic decline.

To understand this gap, we must go back to April 29. On that date, case data for Georgia appears to show a peak on April 14, followed by a rapid decrease in new cases to nearly zero on April 29 (see top panel of the infographic below). However, the period between April 15 and April 29 should be treated more cautiously. We know that there are lags between infection, confirmation, and reporting. With the benefit of hindsight, we can look at the data available in this same period—14 days later on May 13. As is apparent in the bottom panel of the infographic on the blue line, the case report evidence tells a very different story. New cases between April 15 and April 29 fluctuated, but the overall trend suggests COVID-19 was not declining. Instead, the number of newly recorded cases has been about the same since early April.

One chart of Georgia COVID-19 cases shows numbers declining. The other shows that numbers are not declining.
Stephen Beckett and Joshua Weitz

There are several reasons the epidemic may appear to be declining, even when cases are steady. First, there are delays between testing and reporting due to the time it takes to transport samples and backlogs in analysis, and then the time it takes to record and report those tests to the state. Second, confirmed COVID-19 cases are usually assigned to the date of symptom onset or the date the test was performed, not the date when the results come back, which could be days to several weeks later.* This means that numbers will constantly be retroactively adjusted. And while this may be acceptable practice in the context of intervention efforts, if we don’t acknowledge and account for the delay, it will provide a faulty barometer to those who want to use recent case reports as a real-time measure of the epidemic trajectory.

One of the criteria in the “Opening Up America Again” guidelines issued by the White House stipulates that cases should be on a downward trajectory over a 14-day period. But this 14-day period is precisely when case numbers are underreported. The infographic shows that while new cases appeared to be falling between April 15 and April 29 when viewed on April 29, in hindsight, new reported cases were at similar levels in late April as in early April. Notably, the highest recorded number of new cases up to that point was recorded on April 27, days before reopening. It was not clear at the time it was announced that Georgia met the conditions for reopening. Today we know those conditions were not met.

Given the characteristics of the COVID-19 pandemic, it seems clear that reopening criteria must consider the impacts of lagged data rather than treating the most current (and underreported) data as the basis for confirming a decline in cases. One way to deal with this could be to try to compare gaps in reported cases vs. final totals after reporting windows have passed. Such information could guide hindsight casting and provide improved estimates of real-time cases during reporting windows.

Moving forward, Georgia, and every other state, must prepare long-term strategies to minimize and contain the spread of COVID-19 while fostering safer economic reengagement. The vast majority of the population remains at risk of infection. Mask-wearing, strategic social distancing, approaches to test/trace/isolate, expanding testing to inform strategies, sustained public awareness campaigns—all of these are approaches we need to take now.

If decision-makers in the state of Georgia want to use evidence-based criteria to continue to reopen, they must not conflate reporting lags with actual declines to suit preferred narratives. Instead, they must deal with the evidence as it is, and use evidence-based criteria to take the necessary steps to continue to drive down cases.

New Zealand, a country of nearly 5 million, has had multiple days of zero cases and has begun to safely reopen many sectors of its economy. Georgia, a state of over 10.5 million, has now had more than a month where cases have held steady, often with 600 to 700 reported cases a day. Furthermore, these case reports are almost certainly a significant underestimate of the true prevalence. Let’s not be satisfied with a plateau. Aiming for a real zero is what matters.

Correction, May 18, 2020: This piece originally failed to note that tests are most frequently dated to when symptoms started. Only if that date is unknown are they dated to when the test was taken.