Michael Nagle/Bloomberg via Getty Images
In early February, the Centers for Disease Control and Prevention was looking for ways to stop the novel coronavirus before it got out of control in the United States.
The agency selected six cities for sentinel surveillance — a sort of early warning system to detect if the coronavirus was spreading freely. The idea was to look for the virus among patients who had mild respiratory symptoms and no known travel-related risk factors — patients who weren’t being tested under CDC guidelines at the time.
To speed the project along, the CDC’s plan was for cities to piggyback on their well-established flu-tracking programs. Patient samples were already being tested for influenza strains. The agency would provide coronavirus tests for a subset of those patients to see if the coronavirus outbreak was spreading undetected.
But that quick start to the project was far from quick in most places. Fully five weeks later — contrary to statements from top CDC officials — only one of those cities had in hand any results from completed coronavirus tests, according to an investigation by NPR.
Challenges with the CDC’s coronavirus tests, struggles with logistics, clashes between federal and state officials and even hospitals’ fears of being stigmatized as a source of infection — all cost valuable time in controlling the spread of the coronavirus across the U.S., sources tell NPR.
Once the surveillance project finally got underway, its early warning came too late for some cities. Others appear to have gotten the alert just in time — or at least soon enough to significantly curb the death toll.
As they got evidence that the virus was moving far faster than they realized, health authorities in California cities were able to use the information to convince public officials to issue stay-at-home orders for most residents — earlier by a week or more than other places, at a time when unabated community spread of the virus had already begun.
“It does feel like we lost time,” says Jennifer Nuzzo, an epidemiologist at Johns Hopkins University’s Bloomberg School of Public Health. “As we learned that this disease was spreading — and spreading efficiently — between humans, we should have stood up sentinel testing in the United States.”
Virus under the radar: how the surveillance project began
On Feb. 13 of this year, Health and Human Services Secretary Alex Azar made an announcement before a Senate committee.
Five cities, he said — Chicago, Los Angeles, New York, San Francisco and Seattle — would begin looking for the coronavirus in some people who didn’t meet the CDC’s strict guidelines for getting a test. Honolulu was added to that list of cities a week later. The goal was to find out if the coronavirus was spreading undetected.
At that time in February, the CDC had been running only a few dozen tests per day at its headquarters in Atlanta. Beyond those, public health labs were not able to process the CDC’s test kits because the kits were faulty.
As a result — outside the new surveillance experiment — the only patients who could get tested had to be symptomatic for COVID-19 and either had traveled to Hubei province in China or had contact with someone confirmed to have been infected with the coronavirus. Hospitalized patients who had recently traveled to any part of China could also be tested.
J. Scott Applewhite/AP
The six cities were selected because of the high likelihood that coronavirus cases would emerge there and because, in partnership with the CDC, they already had flu-surveillance programs in place. Every flu season, the cities that have such programs test hundreds of samples from patients showing mild respiratory symptoms for various strains of influenza. They report their findings to the CDC to help track where flu is spreading and which strains are prevalent.
To adapt that system to the current pandemic, samples from patients who tested negative for flu would be eligible for a coronavirus test. Since the public health labs couldn’t run the CDC’s tests, the federal agency offered to do tests for them at its headquarters in Atlanta.
“We recognize there was a situation where we didn’t have as many tests as we wanted to do in the U.S.,” says Dr. Joseph Bresee, deputy incident manager at the CDC. “But any sentinel site, we had guaranteed testing for.”
Officials hoped to use the surveillance testing to more accurately gauge how fast the coronavirus had already spread — and to do so quickly.
“Results from this surveillance would be an early warning signal to trigger a change in our response strategy,” the CDC’s Dr. Nancy Messonnier announced to the press in a briefing on Feb. 14. “This is just the starting point, and we plan to expand to more sites in the coming weeks until we have national surveillance.”
A month later, on March 11, top health officials assured Congress that surveillance testing was underway.
“The CDC has already started that in six sentinel cities and will expand that in many more cities,” Dr. Anthony Fauci, director of the National Institute of Allergy and Infectious Diseases, told the House Oversight and Reform Committee. “We need to know how many people — to the best of our ability — are infected, as we say, ‘under the radar screen.’ “
Here’s the thing: Some of those sentinel cities were still having trouble getting their testing programs off the ground.
Chicago goes first
Chicago was the first in the CDC project to look for the coronavirus in its flu patients, beginning testing on March 2 in five hospitals, along with several other counties in Illinois.
It had an advantage. The Illinois state public health lab was the first in the U.S. to be able to process the CDC’s test kits, beginning in mid-February — somehow avoiding the faulty test kits that went to other states.
“Being able to do the testing locally definitely made it more feasible and realistic that we would be able to do sentinel surveillance,” says Dr. Jennifer Layden, chief medical officer for the Chicago Department of Public Health.
In the first week of testing, 95 emergency room patients all tested negative for the coronavirus. But by the second week, 5% of patients were found to be positive.
“We were surprised by the number of positives,” says Layden. “Identifying community transmission let us know that was occurring and changed our protocols about who we should consider testing for COVID-19.”
But in the middle of the two-week project, sampling a broader number of mild cases for surveillance became challenging, state officials say, as the chemicals used in running tests began to run low.
“We do want to expand our sentinel surveillance in order to look for the community spread,” said Dr. Ngozi Ezike, director of the Illinois Department of Public Health, in a press briefing on March 8. “And so that means you would need more reagents.”
Ultimately, the city and state waited to act on the surveillance testing findings for a full week; a statewide shelter-in-place order took effect on March 21. The city quickly became a hot spot for viral spread, seeing an early surge in coronavirus cases that is only just starting to flatten.
Success in Los Angeles after a false start
With a high volume of travelers flying in from China every day, health officials in Los Angeles suspected by late January that a wave of coronavirus cases was headed their way. The city’s first documented case showed up Jan. 26 — a traveler from Wuhan, China — and officials feared cases of infection might be falling through the cracks.
“We were interested in trying to identify people who might have been missed by the CDC testing criteria at that time,” says Dr. Prabhu Gounder, medical director of the Respiratory Disease Unit at the Los Angeles County Department of Public Health.
Los Angeles County health officials identified five hospitals for the surveillance project that served large Chinese immigrant populations in the area. After discussions, they began drafting a study protocol with one of those hospitals.
Gounder says the hospital’s medical staff was supportive, but the hospital’s board was not — remember, only one coronavirus case had been confirmed in the county at that point.
“They had a very different perspective,” Gounder says of the hospital’s board. “They were concerned that if the second case in LA County was linked to this hospital, that there would be a certain stigma that would potentially be bad for the hospital in terms of the anxiety among health care workers about exposure and also the potential for the patients to be anxious and concerned that they may be exposed to COVID by coming to this hospital.”
Los Angeles health officials declined to identify the specific hospital because they had agreed to keep discussions confidential. NPR contacted several hospitals and confirmed that discussions with the Los Angeles County Department of Public Health had taken place, but could not verify which hospital’s board had those concerns.
Gounder says efforts to find the coronavirus in the general community stalled for several weeks in the region, until another hospital called — Los Angeles County + USC Medical Center, in Los Angeles.
“This is a leading-edge indicator,” says Dr. Brad Spellberg, chief medical officer at Los Angeles County + USC Medical Center. “If you don’t look at it, you’re missing the canary in the coal mine. You’re waiting for the coal mine to collapse on you.”
Spellberg’s hospital had been preparing its own surge plan to be ready for a big, sudden influx of COVID-19 patients. But not being able to test mild cases for the illness was like “flying blind,” he says. So his hospital partnered with the county to do surveillance testing of patients with mild respiratory symptoms in the hospital and its affiliated urgent care center — though Spellberg says there was contention within his institution about testing mild cases.
“There was huge resistance,” Spellberg says. The hospital had just secured a small number of tests from a private laboratory — Quest Diagnostics — and had intended them for the sickest patients, which was in accordance with the CDC’s general guidelines at the time. “The whole system was, like, ‘You’re wasting our tests. Don’t do this.’ “
Nonetheless, testing at the county medical center’s hospital and associated clinics began March 12, and results started coming back a few days later. Over four days, the researchers learned something that alarmed them: As many as 5% of patients they tested had the coronavirus.
“It’s spreading in the community silently, and we didn’t even know about it,” says Spellberg. “These were people that had been going to work, going to social events, wandering around the community for days and days.”
“We were all a little bit surprised and alarmed,” says Gounder. “It was the first real firm data point that we had that confirmed our suspicions.”
The finding marked a key turning point in LA County’s coronavirus response. Until then, health officers had been trying to contain the virus, using contact tracing to meticulously track down every person who may have been exposed. But when a virus is spreading undetected, that sort of containment strategy alone is insufficient, epidemiologists say.
Three days after the Los Angeles surveillance study wrapped up, Los Angeles Mayor Eric Garcetti added a new strategy, ordering residents to stay at home, with certain exceptions. So far, the county’s hospitals have not been overwhelmed by COVID-19 patients.
“It is possible that the ‘Safer at Home’ order went into place just in the nick of time to stave off a New York situation,” says Spellberg.
Seattle and CDC clash over early testing
As health officials in King County, Wash., began discussions with the CDC in early February, state health officials had a suggestion: The Seattle Flu Study, a research project that was already gathering samples from patients with respiratory symptoms, could be tapped to check for the coronavirus.
“That was a surveillance system that was up and running,” says Dr. Scott Lindquist, Washington state epidemiologist for communicable diseases. “My point was, why don’t we use those samples? It was up and running — let’s do it.”
But as a research project, the Seattle Flu Study’s coronavirus test — developed by the study’s researchers — wasn’t yet authorized by the Food and Drug Administration, and the lab the study used wasn’t certified to do clinical work for patients. (The FDA didn’t authorize labs outside the CDC to handle coronavirus testing until the end of February.)
“CDC and FDA would not approve us using those samples,” says Lindquist. “I kept asking, ‘Can’t we just [get] an exception?’ “
Sensing a growing urgency, researchers conducting the flu study tested samples for the coronavirus anyway and discovered on Feb. 28 a Snohomish County teenager with a presumptive COVID-19 diagnosis who had not traveled to a country with a known coronavirus outbreak and who had had no contact with a patient who was positive for the virus.
Meanwhile, King County, Wash., health officials continued working to find other clinics to partner with for sentinel surveillance testing, as the CDC suggested — clinics that would collect samples to be sent to the CDC labs in Atlanta for processing. But the county had a number of logistical issues to sort out, such as finding the right populations to test, setting up a system to inform patients and arranging collection methods and transportation for samples to a site where they could be evaluated. That approach never came to fruition.
Eventually, researchers running the Seattle Flu Study got the necessary approvals to begin surveillance testing, so the county chose to partner with them. The surveillance study officially began on March 23.
Lindquist says sentinel testing earlier than that would have been very helpful in the weeks immediately after the state’s first confirmed case, on Jan. 21. Researchers have since found that the virus was likely spreading for weeks, undetected, during that time period.
“It could have let us know that it was here before we had the community outbreaks and transmission in the long-term-care facilities,” Lindquist says. “But we missed that period.”
Surveillance elsewhere: missteps and successes
The pattern of slow starts or delays that plagued Chicago, Los Angeles and Seattle also plagued the other three cities in the CDC’s surveillance project.
By the time New York City began its surveillance testing, it was far too late to provide an early warning. NPR was unable to get details of New York’s experience with the project, but city health officials did confirm that in March, New York collected its first samples for this project — from more than 500 patients.
On March 23, the day New York sent those first samples to the CDC for testing, the city already had more than 12,000 coronavirus cases. By the time it got preliminary test results back on March 31, showing 6.6% of patients were positive, the city’s outbreak was out of control — with more than 900 deaths, and climbing.
Hawaii got a relatively early start with surveillance, in relation to its first recorded case, and results so far suggest that this may have paid off. The state started testing for community spread with the CDC project in mid-March. That was not long after Hawaii’s first case of COVID-19 was confirmed on March 6 in a passenger returning from the Grand Princess cruise ship.
In the first week of that testing, the state found no cases of community spread (according to data ultimately released March 27), and it found one case the next week. Hawaii issued its stay-at-home order to begin March 25. Since then, Hawaii has tested more than 1,000 samples through its flu-surveillance network. As of April 17, 2.1% of those cases have come back positive for the coronavirus.
San Francisco, after having early discussions with the CDC, did not begin sentinel surveillance testing. A spokesperson for the San Francisco Department of Public Health told NPR that it was not the city’s decision, but ultimately the project “did not materialize.” The city provided no further detail.
Silicon Valley starts its own surveillance
Outside the CDC’s six original cities, some other communities were eager to establish their own surveillance testing. But they, too, ran into trouble.
For example, about 30 miles south of San Francisco, in the heart of Silicon Valley, Santa Clara County saw its first travel-related case of the coronavirus on Jan. 31.
The county wasn’t on the CDC’s original list of six test sites, because it isn’t part of the agency’s flu-surveillance network. Santa Clara County health officials had discussed starting sentinel surveillance testing on their own, but — as was a problem in Seattle and New York — the county health lab initially couldn’t run CDC test kits.
In late February, the county found its first instance of community spread.
“We needed to know pretty urgently how much [community spread] was happening,” says Dr. George Han, deputy health officer with the Santa Clara County Public Health Department.
On March 5, with help from a CDC field response team, Santa Clara County tested a sample of patients with mild symptoms at four urgent care centers and found that 11% were positive for the coronavirus. Two days after that testing wrapped up, Santa Clara and five other Bay Area counties issued shelter-at-home orders to residents.
“This data [from the sentinel testing] helped inform that decision because it was one of the few relatively unbiased sources of data that we could rely on,” Han says. “That was, I think, a key point that we needed to have in order to issue the orders.”
The action has been widely credited with, so far, preventing an overwhelming surge in COVID-19 cases in the region.
Santa Clara health officials say that if the CDC test kits had been functioning earlier in February, they would have started sentinel surveillance even sooner — and potentially have prevented even more deaths. The county recently learned that a resident died of COVID-19 on Feb. 6, earlier than the first deaths previously known and reported. That discovery suggests the virus already was circulating in Santa Clara County in late January.
What was gained and lost in a crucial few weeks
As coronavirus outbreaks have ballooned across the U.S., epidemiologists say the communities that ordered social distancing and sheltering at home earlier than others may have substantially limited the number of infections and deaths.
“A week or two can make a big difference,” says Dr. Arthur Reingold, an epidemiologist at the University of California, Berkeley’s School of Public Health. “The earlier you do it, the better.”
The CDC says the long period of time it took to begin sentinel surveillance — up to six weeks in some cities — can be attributed to the logistical challenges of dealing with a new and deadly virus.
“The rollout at these clinics — both identifying clinics that were interested and ready and willing to do this and also just getting them ready in terms of the sample shipping and collection — even though it was similar to influenza, it was different enough that it took a little more time,” says the CDC’s Bresee.
Plus, Bresee says it was important that safety protocols were in place before the projects began — health care workers testing patients for the coronavirus needed to receive and wear additional personal protective equipment, for example.
At the time, Bresee says, the CDC was also monitoring other data — such as the number of hospitalizations of people with flu-like symptoms — that suggested that cases of coronavirus infections were still limited in February.
“It’s a good question about whether this three- or four-week delay, or longer, resulted in losing visibility in what was happening,” says Bresee. “But I don’t think so.”
While it’s hard to know for sure whether earlier surveillance testing would have changed outcomes for tens of thousands of people in hospitals and millions sitting at home, leading epidemiologists say those early weeks of the outbreak were crucial.
“Three weeks is an enormous amount of time to allow cases to accumulate without knowing about it,” says Nuzzo, of the Bloomberg School of Public Health. “If we had had sentinel surveillance [activated] in a number of cities — cities where we expect to see cases first — we possibly could have caught it earlier and possibly intervened before the case numbers exploded.”
Next steps for communities and the U.S.
As many cities look forward to a time when cases of the coronavirus wane, they’re already beginning to plan new surveillance efforts to get ahead of the next wave of new infections.
“We absolutely want to restart sentinel surveillance and are preparing to do that,” says Layden, the chief medical officer of Chicago. “We want to have a good sense of the level and amount of spread that’s occurring in our community.”
Some cities are planning on setting up surveillance projects that test a sample of patients with flu-like symptoms, while others plan to use antibody tests, which check for prior infection instead of active cases, to get a better estimate of how broadly the outbreak has already spread.
The U.S. government has also identified sentinel testing in its road map for reopening the country. In this case, the project would look beyond mild cases; it would test people who seem healthy. The goal is to get a sense of how many people who don’t show symptoms are actually infected and able to spread the virus.
“We’ll be doing sentinel surveillance throughout nursing homes, throughout inner-city federal clinics, throughout Indigenous populations — to really be able to find an early alert of asymptomatic individuals in the community,” Dr. Deborah Birx, of the White House’s coronavirus task force, said at a press briefing last week, as the Trump administration laid out its plan.
Some epidemiologists say this sort of national effort could be a chance to get sentinel surveillance right — providing the critical information that health officials will need to avoid a second, bigger surge of severe illness and death.
“Going forward, we need a national testing strategy, period,” says Nuzzo. “Not just testing people for clinical purposes, but testing strategically so we can understand how much illness is in the community and how it’s spreading.”
An effort like that would require widespread availability of coronavirus tests. While the Trump administration claims that adequate testing capacity already exists, many regions of the country still report shortages of test supplies, signaling that the country isn’t yet prepared to gather the crucial surveillance data it will need.