Slow relaxation of COVID-19 rules helps push recurrence back

Countries like the United States have never really gotten the pandemic under control, while others, like Brazil, haven’t even slowed the pace of infections. But elsewhere, many countries that took dramatic action to limit the spread of COVID-19 have seen the rate of infections plunge, leaving them with the issue of how to successfully emerge from the restrictions they put in place.

One of those countries is Spain, where infections have dropped from a peak of over 9,000 a day in late March to only about 300 a day at present. Two researchers based in Barcelona (Leonardo López and Xavier Rodó) decided to look at different ways Spain could have exited its restrictions while protecting future public health. After building their model, they set it loose on other countries, including the US. Their work lets us test what might happen if the immunity developed in those infected fades over time or the public’s fear of the virus subsides after it’s under control.

Lockdown vs. open up

How did Spain get cases in check? In part, by following the advice of public health experts. On March 29, it placed everyone in non-essential jobs on a strict lockdown. Two weeks later, with cases dropping, restrictions started being eased. But the easing was done cautiously, with a variety of restrictions being kept in place as cases continued to drop. López and Rodó were interested in looking at how this reopening could be handled in a way that most effectively limits future returns of widespread infections.

To do so, they built an epidemiological model that was able to incorporate a variety of distinct factors. The foundation of their work was a standard susceptible–exposed–infectious–recovered model, in which the population is moved among the above groups based on the spread of the virus. That spread is based on what we know of things like the virus’ infectivity, how long people can spread it before and after developing symptoms, and so on.

López and Rodó added some critical features to this model. To begin with, it’s able to handle the fact that some aspects of the virus’s behavior aren’t precisely defined—our estimates of how long people are infective before symptoms appear are a bit rough, to give one example. So the authors added a probabilistic element, where the model could be run multiple times with uncertain values varied around their most likely values. This helps show how robust any result is by highlighting whether it’s sensitive to small changes in things we’re uncertain about.

To address the issue at hand, the researchers included an isolated population, which could not become infected. They acknowledge that this is obviously unrealistic, since some of the people in isolation will undoubtedly become infected. But they argue that it provides a low estimate while avoiding potentially alarmist results.

Finally, they include the ability to have some features of the pandemic and public health response vary over time. These include factors like a fading of the immunity that’s been seen in other coronaviruses. In addition, they allow the degree of compliance with social-distancing rules to vary over time, as fears during periods of high infection will probably boost compliance initially but fade over time. These factors influence existing pieces of the epidemiological model. For example, the fading of immunity would move people out of the “recovered” pool and back into the “susceptible” group, while fading of social-distancing compliance could be simulated by increasing the rate of new infections.

Finally, and critically for the research question here, the process of ending lockdown could be modeled directly as a time-varying process of moving more people out of the isolated population and into the susceptible one. This let them explore different rates of ending the lockdown, from an immediate, complete elimination of the isolated population to different fractions being removed over time.

Spain and beyond

The researchers started by initializing their model with the actual data from Spain, from the start of the pandemic up to its peak in that country. People were released from the confined group until the pre-pandemic working population of Spain was back in the susceptible group. The researchers modeled various dynamics for this “deconfinement,” ranging from letting everyone out immediately to gradually allowing increasing numbers of people back into the working population. They also varied the length of the lockdown period.

The results are about what you’d expect. The longer the lockdown lasted, the longer it took for a second peak of infections to occur. And a gradual release of the isolated population tended to delay the appearance of the new peak even further. (Spain’s actual policy, which was also simulated, leaves it vulnerable to a a new peak in July, according to this model.) In the case of a 90-day lockdown and a gradual rate of release from isolation, the second wave was delayed to beyond the time scale of their study, which ended in early 2021.

In all cases, however, the number of deaths remained low, as one of the most vulnerable populations—the elderly—was kept in isolation since it wasn’t part of Spain’s pre-pandemic work force. The impact on deaths, however, was still reduced by shorter lockdowns and in scenarios where the release from isolation was immediate, rather than gradual.

The researchers then turned their attention to other countries. Indonesia was chosen because it doesn’t have pronounced seasons, while Argentina was picked because it is about to enter winter. Despite those differences, however, the model suggests a similar pattern of viral spread, with a second wave coming around January: sooner for shorter lockdowns, later for longer ones. For countries like Japan and New Zealand, however, the aggressive approach to handling the pandemic has left them in such good shape that there was no scenario in which a second wave occurred before 2021. If they did a longer lockdown and a more gradual reopening, the second wave might not arrive until nearly a year from now.

The big exception to all of this is the United States. In the case of shorter lockdowns—which is what we’ve essentially done in many states—the current rise grows to about 10 million infections before building to an enormous peak in late summer, with over 50 million infections. Longer lockdowns, at best, delay that to the winter and shave the peak by about 10 million. In short, even the best case of a hundred-day lockdown only stops the current peak and delays a resurgence so that it occurs around the same time as those in the worst cases for other countries.

Variations

López and Rodó also looked at some variations on the other interventions their model explored. They tried changing the number of people released from confinement during their gradual reopening. While they reasoned that they could put much of a country back to work by having two times the people outside confinement as there were inside, they found little difference between the results from that and the results when 10 times as many people were unconfined as remained isolated. So as long as the release is gradual and the highest-risk individuals are kept confined, it’s possible to handle reopening safely.

This contrasts sharply with trying to alter the rate at which people are allowed out of the lockdown. Boosting the rate by a factor of 10 always produced a substantially higher second wave, regardless of how long the lockdown was for. The wave also arrived more quickly.

The researchers also modeled risk awareness by altering the rate at which the susceptible population took precautions that limited the spread of the virus, like maintaining social distancing and using face masks. They found that high levels of public awareness, when coupled with lockdowns and a gradual reopening process, was capable of blocking a second wave of pandemic entirely. The complete absence of these sorts of behaviors, in contrast, accelerated second waves so that they occurred earlier in 2020.

Next up, López and Rodó looked at the role immunity would play by allowing people from the recovered population to move back into the susceptible pool, testing rates that assumed everything from a short immune period of four months to one lasting a year. Not surprisingly, having longer periods of immunity could double the amount of time between subsequent waves of infection and reduce the size of the peaks in each subsequent wave.

In summary…

Most of the findings here are consistent with work from other researchers. We’re already fairly sure about issues like whether lockdowns are effective and how social distancing and face mask use limit the rate of the virus’s spread. But López and Rodó put some quantitative weight behind the idea of gradual reopenings, which are being tried in a number of US states. “Gradual deconfinement strategies always result in a lower number of infections and deaths,” they conclude, “when compared to the sudden release of moderate to large portions of the population.”

The other thing that is clear is that we’re going to be engaged in an extended struggle with our own indifference and frustrations. “The use of social distancing, face masks, gloves, and other individual protection measures has a massive impact in reducing the current peak of active cases,” the authors write, “but diminished awareness over time of the threats of the pandemic may result in a new larger second epidemic wave.”

The other thing that’s clear from these models is that ending lockdowns while case numbers are still high is dangerous and can extend the ongoing peak. In this case, the US seems determined to demonstrate that the predictions of the model are borne out by real-world data.

Nature Human Behavior, 2020. DOI: 10.1038/s41562-020-0908-8 (About DOIs).

https://arstechnica.com/?p=1686967