Many countries that controlled their COVID-19 cases in the spring are now seeing rises in infections, raising the prospect that they’ll face a second wave of cases, as many epidemiological models had predicted. But in the United States, the number of cases has never dropped to low levels. Instead, it varied between high levels of infection and very high peaks in cases. Why is everything so different in the states?
While there are plenty of possible reasons, a series of new studies essentially blame all the obvious ones: the United States ended social distancing rules too soon, never built up sufficient testing and contact-tracing capabilities, and hasn’t adopted habits like mask use that might help substitute for its failures elsewhere. The fact that some of these studies used very different methods to arrive at similar conclusions suggests that those conclusions are likely to hold up as more studies come in.
One of the studies, performed by a US-South African team, looked at the relaxation of social distancing rules in the US. Its authors created a list of restrictions for each state and the District of Columbia and tracked the number of COVID-19 deaths in each state for eight weeks prior to the rules being terminated. The number of deaths was used as a proxy for the total number of cases, as the erratic availability of tests made the true infection rate difficult to determine.
Most states started relaxing these rules in late April. But, as the authors note, they did so without the capability of controlling infections through other means. “Relaxation of such measures is intended to be accompanied by appropriate behavioral practices (e.g., mask-wearing and physical distancing) and control measures (e.g., contact tracing and increased availability of testing), so that epidemic control can be maintained,” they wrote. Given that testing capacity was limited and flouting of behavioral practices were widespread, that simply wasn’t possible.
So, the authors gathered data on COVID-19 deaths from states after they’d lifted restrictions and compared the two trajectories. Linear regression models were used to take the number of COVID-19 deaths and estimate the likely reproductive number for the virus in each state and DC.
Of the 51 examples, 44 had seen the virus’ reproduction slow while social limits were in place. As a whole, the authors estimate that the US saw the virus’ reproductive number drop by an average of 0.004 per day during this period. While not dramatic, this meant that 46 had a reproductive number of less than one by the time they started relaxing their social distancing rules—a situation that would ultimately mean the end of the pandemic.
Unfortunately, that decrease ended with the relaxation of the rules. After they were gone, the estimated reproductive number went from declining by 0.004 per day to rising by 0.013. Only eight states and DC were able to keep the reproductive number under 1.0 after the rules were relaxed, meaning the pandemic was back on the path to growth.
There are obviously a lot of state-by-state differences in the restrictions put in place and how high the infections were when those restrictions were first put in place. So it’s no surprise that, when the researchers break out each of the states, there’s no simple pattern to either the “before” or “after” of the restrictions. But both the overall results and national average clearly suggest that the pandemic-focused restrictions were ended too soon.
And not enough
And if that weren’t enough, an epidemiology modeling paper that’s focused on a somewhat different question reaches the same conclusion. The work, done by a group of researchers at Texas A&M, is actually focused on what we’d need to control the pandemic without returning to heavy restrictions on social interactions. But, in the process of in the process of finding out what we’d need to control the pandemic, the A&M team figured out what those restrictions might be accomplishing right now.
For the work, the researchers built a standard epidemiological model and used mobility data from companies like Google and Open Table to adjust its properties for both periods of social restrictions and after reopening. They also added in data on state-level cases and deaths and then validated the model using historic data.
When they actually analyzed their model, it more or less reproduced the results above. For all but five states, the effective reproduction value of the virus was less than one early in the pandemic, “mainly achieved during the state shelter-in-place.” Once those restrictions were lifted, the model showed that infections started to increase, and by mid-July, 42 states and DC were likely to have viral reproduction rates that would enable the pandemic to expand.
By the last date used in their analysis—July 22—the chance to control the pandemic was pretty much over. Only three states, all in the Northeast, would be able to without adding back additional social restrictions. Absolutely none would be able to do so if they relaxed any existing limits. Even if states were to double existing testing and contrast tracing, only eight could manage to bring the viral reproduction number down to the point where the pandemic could be brought under control. Another 30 would need to do that and increase social restrictions. The rest would need to go back into a severe lockdown.
“We showed that, in most states, control strategies implemented during their shelter-in-place period were sufficient to contain the outbreak,” the authors conclude. “However, for the majority of states, our modeling suggests that reopening has proceeded too rapidly and/or without adequate testing and contact tracing to prevent a resurgence of the epidemic.”
Wear a mask already
The authors acknowledge that their model has a notable weakness: it assumes that personal protective measures such as face-mask use and physical distancing are adopted roughly in proportion to the number of people obeying the state-mandated social restrictions. That’s not an unreasonable assumption, but it prevents the model from being able to analyze the effect of these personal measures separately from official policies on limiting social contacts.
That brings us to a draft paper that’s not yet been through peer review but addresses the issue directly using data from Ontario, Canada. Its authors compared the infection rates in 34 different public health districts within Ontario before and after the adoption of mask-wearing mandates. Like the A&M group, the authors use Google mobility data to control for the frequency of personal interactions. Overall, they estimate that the use of masks probably dropped the infection rate in Ontario by somewhere between 20-40 percent.
The thing is, none of this should be at all surprising. From the start, public health officials said that the social restrictions were needed to control the infection rate so that testing and contact tracing could be effective at keeping the pandemic in check. Data from the pandemic has only served to indicate that this initial advice was exactly right. The United States’ response, however, has been to lift the restrictions before the infection rate was controlled and to limit testing sufficiently to make contact tracing nearly impossible. As an added bonus, the country has made some of the possible alternative ways of limiting the pandemic, such as the use of protective masks, a political issue.
So, while the papers give us some indication of what will be needed to keep the United States from seeing the pandemic continue to spread out of control, they also serve to highlight how we’ve done pretty much everything wrong.