It’s Too Soon to Reopen States. The Coronavirus Is Not Under Control.

https://www.nytimes.com/2020/04/22/opinion/coronavirus-states-reopen.html

Version 0 of 1.

Georgia, Tennessee and South Carolina are eager to ease social-distancing restrictions. They believe they are over the worst of their local outbreaks. Some of their assurance is because they’ve been looking at models that show that once an outbreak has peaked, it resolves about as quickly as it began.

But evidence from Western Europe is not fitting this pattern. Governors should rethink decisions to loosen restrictions anytime soon.

Models are useful for the Covid-19 pandemic because they help us plan and make decisions. They also help us to predict the future. Most of the models that have caught the country’s attention are nicely symmetric. They show cases and deaths rising rapidly, peaking and then coming down just as fast. The report from the Imperial College in Britain that seemed to make much of the country snap into focus in March outlined a variety of strategies that we could take to mitigate the spread of the infection; almost all showed a rise and fall that mirrored each other.

Models from the Institute for Health Metrics and Evaluation at the University of Washington, known as I.H.M.E., which reportedly caught the attention of the White House weeks later, showed a similar pattern. Cases and infections go up relatively smoothly, slow, and then come down smoothly.

These two models use different methodology. Imperial College runs calculations based on how many people are susceptible, exposed, infected, and then recover (S.E.I.R.). Different models can include variables — like how infectious or how deadly the virus is — to change outcomes. If, for instance, we engage in social distancing, which dramatically lowers the chance that one is exposed, a model will show that we can expect a very different course.

I.H.M.E took data on changes in numbers of infected and dying from countries impacted earlier than the United States, such as China and Italy, and used those to predict the course of the disease. It assumed that America would have a similarly shaped experience (albeit with much different numbers) and presented a range of possible outcomes.

The models were all useful in helping to plan for worst-case scenarios. They were also useful for helping guide us into decisions to flatten the curve and prevent the health care system from being overwhelmed.

Even as they are updated, though, many of the models remain symmetric. Reality isn’t.

Let’s start with new cases. According to the Johns Hopkins Coronavirus Research Center, Italy had its worst day of new cases (6,557) in mid-March. A month before, it had almost none. More than a month later, it’s still having thousands a day. Spain started seeing an increase in cases in the beginning of March and peaked about three weeks later at 9,630. About a month after that, it’s still finding around 4,000 a day. Belgium rose and flattened, not dropped. So did the Netherlands.

None of them rose, peaked and fell smoothly.

The numbers of new cases can be influenced by testing. The less you test, the less you’ll find. It doesn’t appear, though, that the more-than-expected numbers of cases are because of increased testing after the peak. And deaths, while more of a lagging outcome, show a similar pattern.

Italy’s deaths peaked at the beginning of the month, slowed a bit, but have leveled off in the last week or more. Britain peaked about a week ago and has held somewhat steady. So has France. So has Germany. Spain, while coming down, is coming down more slowly than it went up.

What’s worrisome is that none of these countries have seen a rapid decline to mirror their rapid increase. The United States, of course, really hasn’t seen any significant decline at all.

Yet many are acting as if the original models still hold. Since they are passing their peak, they believe that it’s time to start talk about loosening restrictions.

That’s a big mistake.

In order to contain a future outbreak, a city, state or country needs to be able to identify and isolate enough cases to prevent the “effective R” — the number of secondary cases per infectious case — from getting above one. This means that every infected person has to infect no more than one other. Achieving that goal would keep the numbers of cases showing up every day from increasing.

Most of the models, though, assume an effective R that remains well below one after the peak. Good social distancing policies do that, and when modeled, that’s how we see rapid declines.

“Some transmission models assume that the reduction in effective R remains fixed during a policy like social distancing and doesn’t change as long as that policy is in effect,” said Marc Lipsitch, a professor of epidemiology at Harvard. “S.E.I.R. models are approximately symmetric under static policies for that reason.”

It’s worth noting that not all S.E.I.R. models produce perfectly symmetric curves. A study by Professor Lipsitch and colleagues in Science found that intermittent social distancing policies might yield slower resolutions than upswings. Few models, if any, predicted the flattening or stalling we’re seeing in Europe, though.

It’s not clear why this is so. It could be that even though policies have remained unchanged, people have been sporadically relaxing their efforts. We’re really not sure.

What we are sure of, though, is that the rate of transmission isn’t low enough to think we’ve succeeded in getting a hold of this pandemic.

Areas that have seen their reductions in new cases stall have an effective R hovering around one. It doesn’t matter what the models predicted. Those countries — and nearly all in Western Europe qualify — are perilously close to seeing cases start to rise. There’s no reason to assume the United States will be different.

Some are nevertheless plowing full steam ahead. They’re talking about reopening society and the economy and letting people emerge from sheltering in place. It’s hard to imagine how this won’t increase the effective R a bit. That may be all it takes to see an exponential increase again.

Moreover, there’s a direct relationship between the number of cases that show up each day and the resources it will take to conduct contact tracing and isolation to prevent further spread of the virus. Locations that exit earlier will need more resources, not fewer, than those that wait until the number of new cases each day has been minimized. Almost no locations are prepared in terms of tests and public health personnel.

“We’ve engaged in policies that have slowed the number of cases,” said Ashish Jha, a professor of global health and medicine at Harvard. “We could have chosen to smash the curve, not flatten it. That would get us to a much more manageable place to reopen.”

Few seem prepared to wait long enough. Leaders are overestimating how far they’ve come, and they’re underestimating what it will take to manage Covid-19 in a more close-to-normal world. Failure to properly prepare will either result in a surge of infections and all the negative health and health care system ramifications that come with it, or a second round of sheltering in place much sooner than anyone expects. Neither is a good outcome.

“Exiting too soon is like thinking you can cut your parachute off at 2,000 feet because it slowed you down,” said Carl Bergstrom, a biology professor at the University of Washington.

It would be better to wait. The ride down from the peak appears to take much longer than the ride up. When the realities don’t fit the models, it’s time to re-evaluate the models’ usefulness.

Aaron E. Carroll (@aaronecarroll), a professor of pediatrics at Indiana University School of Medicine and the Regenstrief Institute, the author of “The Bad Food Bible: How and Why to Eat Sinfully” and host of Healthcare Triage, is a contributing opinion writer.

The Times is committed to publishing a diversity of letters to the editor. We’d like to hear what you think about this or any of our articles. Here are some tips. And here’s our email: letters@nytimes.com.

Follow The New York Times Opinion section on Facebook, Twitter (@NYTopinion) and Instagram.