Yeshiel Xinhua / eyevine / Redux

Updated at 3:10 p.m. ET on April 19, 2020.

Even as Donald Trump has delineated his plan to relax social distancing, the United States remains very much in the dark about who has the coronavirus and who does not. We have a shortage of COVID-19 tests, and we simultaneously have the highest number of confirmed cases in the world. Consequently, not every American who wants a test can get one. Not every health-care worker can get one. Not even every patient entering a hospital can get one. Because of the shortages, we are rationing tests, and medical facilities and public-health officials are prioritizing the sickest patients for them.

If the goal is to restart the American economy, the United States isn’t performing anywhere near enough tests. Worse still, we are testing the wrong people. To safely reopen closed businesses and revive American social life, we need to perform many more tests—and focus them on the people most likely to spread COVID-19, not sick patients.

COVID-19 testing has been an unmitigated failure in this country. This month, according to the COVID Tracking Project, a data initiative launched by The Atlantic in March, the number of tests performed in the United States has plateaued at about 130,000 to 160,000 a day. Rather than growing rapidly—as all experts think is absolutely necessary—the daily number of tests administered in some jurisdictions has even decreased. In New York, for instance, 10,241 tests were performed on April 6, but supply limits forced a huge drop a few days later to 25 total tests. Quest Diagnostics, one of the two biggest firms that run tests, just furloughed 9 percent of its workforce. In addition, Deborah Birx, the White House coronavirus-response coordinator, said during a briefing last week that, of 1 million test kits distributed for use in Abbott Laboratories' high-throughput testing machines, only 88,000 had been used; news reports suggest that shortages of supplies and personnel were to blame.* Testing bottlenecks such as these are major obstacles to getting Americans out of their homes and back on the job.

How many tests do we need in order to safely relax social-distancing measures, reopen nonessential businesses and schools, and allow large gatherings? According to the Morgan Stanley analyst Matthew Harrison and the Harvard professor Ashish Jha, we should be conducting a minimum of 500,000 tests a day. One of the authors of this article, Paul Romer, has called for the capacity to run 20 million to 30 million tests a day. Even this has been criticized as insufficient for the task of identifying enough of the asymptomatic spreaders to keep the pandemic in check.

Current guidelines from the Centers for Disease Control and Prevention give priority first to hospitalized patients and symptomatic health-care workers, then to high-risk patients, specifically those over 65 and those suffering from other serious health conditions, with COVID-19 symptoms. Under this system, asymptomatic individuals are not tested, even if they had contact with people who tested positive.

This is an enormous mistake. If we want to control the spread of COVID-19, the United States must adopt a new testing policy that prioritizes people who, although asymptomatic, may have the virus and infect many others.

We should target four groups. First, all health-care workers and other first responders who directly interact with many people. Second, workers who maintain our supply chains and crucial infrastructure, including grocery-store workers, police officers, public-transit workers, and sanitation personnel. The next group would be potential “super-spreaders”—asymptomatic individuals who could come into contact with many people. This third group would include people in large families and those who must interact with many vulnerable people, such as employees of long-term-care facilities. The fourth group would include all those who are planning to return to the workplace. These are precisely the individuals without symptoms whom the CDC recommends against testing.

Not testing suspected COVID-19 patients will not harm those patients. Because we do not have any treatment targeted for the new coronavirus, confirming an infection generally does not change the way a patient’s symptoms are treated. Patients suspected of having COVID-19 should be presumed to be infected and receive care accordingly. Symptomatic patients should be tested only in the rare case where a positive test would meaningfully change what type of care is delivered.

To shift the focus of testing away from the sickest patients and toward the people most likely to spread the coronavirus, we will have to conduct millions of tests a day. Millions of health-care workers in the United States are in positions that may expose them to infection: physicians, nurses, respiratory therapists, midwives, pharmacists, phlebotomists, hospital cleaners, and others. By one estimate, 3 million people work in grocery stores. To screen everyone in these two groups once a week will require about 1 million tests a day. We currently lack the infrastructure for that. And that is before we add the approximately 800,000 police officers, 290,000 bus drivers, and 60,000 sanitation workers—and patients without any symptoms in the health-care system. We will need millions more kits to test asymptomatic potential super-spreaders and people planning to return to work. Taking a sample from seemingly well people just once isn’t enough; effective surveillance over time requires repeated testing.

How can we close this gap between our needs and current capacity? We need a national strategy over the next 10 weeks, one that draws on the many strengths of our research system. It should leverage the thousands of research laboratories at U.S. universities, medical schools, and health-care systems that have the capacity to perform polymerase-chain-reaction tests for COVID-19—the only type of test that catches infections in the crucial early days. We also need to encourage rapid adoption of the saliva test that now has an emergency approval from the FDA and expedite the approval of tests that require fewer reagents and staff.

Another promising pathway is to pool many tests and run them together. If a pooled sample tests negative, everyone in the pool is negative. If it is positive, the members of the pool can be tested individually. A more sophisticated version of this approach uses genetic “bar codes” that make it possible to trace back which of the many samples in a pool was the one that had RNA from the virus, without any retesting.

How can we get this testing capacity up and running? One idea is for Congress to award in the next stimulus bill, say, $150 million in unrestricted research funds to the first five universities that can process 10 million tests in a week or less. That will unleash a huge amount of latent testing capacity. Another catalyst could be to subsidize businesses that agree to test all their employees as they return to work. This testing should not be in competition with tests for health-care workers and the other groups. To encourage early and widespread adoption, the amount of money a state gets per test decreases with time from the announcement.

A plan for using these approaches to scale up testing will only be a first step toward sustainably containing the coronavirus. When someone tests positive, officials should identify close contacts, find them, and test them. To do the tracing, we may need to hire 100,000 to 200,000 additional public-health workers. Those workers would need to have technology that makes it easy for them to speak with those contacts and direct them to a test site.

This type of voluntary contact tracing is labor-intensive and requires some training, but it does not require highly specialized skills. Technology can speed it up without risking a permanent erosion of privacy or the further intrusion of for-profit firms into our personal lives. These health workers must offer two vital assurances to the American public: Data will never be commercialized, and all data will be destroyed within 45 days, when they no longer serve a public-health purpose.

If we adopt and follow a coherent plan, we can have a testing regime that keeps us safe without compromising our freedoms. But what Americans must not do is assume that the current rate of testing is sufficient to bring our society back to life. Until the United States can successfully perform a lot more tests—and starts testing the right people—the public will have little reason to believe that the virus is in check.

*This article previously misstated which of Abbott Laboratories' COVID-19 tests were going mostly unused. It also omitted the original source of the data about Abbott's unused tests: The information was reported by Deborah Birx.

We want to hear what you think about this article. Submit a letter to the editor or write to letters@theatlantic.com.