How one university built a COVID-19 screening system

“Failure is not an option.” This was the stark, five-word directive that landed in Paul J. Hergenrother’s cell phone in early May.

It came from Martin D. Burke, Hergenrother’s colleague in the Chemistry Department at the University of Illinois at Urbana-Champaign (UIUC). Burke was leading a program to establish a COVID-19 test-and-trace system that would enable the campus to reopen during the summer, and Hergenrother had just been put in charge of developing the test. Both scientists were fully aware that without a suitable diagnostic assay, the whole venture would fail. “I knew that already, but it was . . . reinforced,” Hergenrother says, laughing.

Over the months that followed, researchers and staff worked at breakneck pace to set up Shield, a screening program that aims to identify people infected with SARS-CoV-2, even when they don’t have symptoms, and stop them from infecting others.

Today, anyone with a university ID card who comes onto campus must provide a saliva sample for analysis at a dedicated UIUC lab. People get their results within a few hours, and those who test positive are immediately told to isolate. Anyone who wants to access a building on campus must show a recent negative result on a cell phone app. This homegrown system—built with a mixture of scientific ingenuity, community effort, and millions of dollars—processes about 10,000 tests every day, with undergraduates tested two or three times per week. During late August, Shield performed about 2% of all COVID-19 tests in the entire US.

Universities around the world have adopted a remarkably diverse range of coronavirus strategies since the academic year began in the fall, and Shield is one of the largest mandatory COVID-19 screening efforts at any university. In contrast, some universities have opened their campuses without having any routine diagnostic screening in place at all, and many others have simply shifted their fall semesters online, effectively keeping their campuses in lockdown.

It is too early to tell whether UIUC made the right decision to create Shield and fully reopen its campus. But as a growing chorus of researchers and politicians advocate for frequent coronavirus screening in workplaces, in travel hubs, and even nationwide, UIUC’s experience holds important lessons about the strengths and limitations of that strategy.

It highlights the trade-offs between the speed and accuracy of different diagnostic tests and the vital role of robotics and automation. It demonstrates the inexorable math that even a cheap test, repeated often enough, can rack up an eye-popping bill. And it shows that despite the regular hoopla around exciting new diagnostic technologies, the assay itself is just a small part of a successful screening system. Public hygiene measures, sampling logistics, user compliance, IT systems, and well-trained contact tracers all must work in perfect harmony, at high speed, to identify infections before they turn into outbreaks. “There are multiple pieces to this whole puzzle in which fast and accurate testing is just one piece,” says Awais Vaid, lead epidemiologist and deputy administrator at the Champaign-Urbana Public Health District and part of the Shield team.

Invisible enemy

With the pandemic gaining pace around the world, UIUC suspended face-to-face classes on March 16 and asked students living in residence halls to pack their bags. When university leaders started discussing how the campus could safely reopen in the fall, it soon became clear that the characteristics of the virus would make that a huge challenge.

Simply testing and isolating symptomatic individuals on UIUC’s campus would clearly not be enough to break chains of transmission because asymptomatic or presymptomatic people can spread the virus through aerosols released during normal speech, for example. So in late April, UIUC researchers began discussing whether it would be possible to set up from scratch a system for screening asymptomatic people.

In a standard COVID-19 test, a health-care worker—clad in appropriate protective equipment—inserts a swab deep into a person’s nose. The worker puts this nasopharyngeal (NP) swab into a vial containing viral transport medium, a solution that washes any RNA present from the swab and preserves the specimen during its trip to a lab. Technicians then use various reagents to extract and isolate RNA from the sample, making it ready for reverse transcriptase quantitative polymerase chain reaction (RT-qPCR), which allows them to detect the RNA from SARS-CoV-2.

The Shield team decided that approach wouldn’t work for routine screening. Swabs, transport medium, and vital reagents were all in short supply in the spring, and the university would need an army of trained health-care workers to collect the samples. Students and staff also might be less willing to submit to the uncomfortable swabs several times per week.

A better approach for their needs emerged from a study by researchers at the Yale School of Public Health that found that saliva samples were actually better than NP swabs for diagnosing COVID-19 (N. Engl. J. Med. 2020, DOI: 10.1056/NEJMc2016359). If students could provide samples by drooling into a sterile tube, it would speed up the collection process, simplify the logistics, and potentially avoid some of the supply bottlenecks.

Raw saliva cannot be plugged directly into a PCR machine because it is stuffed with enzymes that disrupt the test. Hergenrother and his colleagues realized that they needed a simple way to deactivate those enzymes and to liberate RNA from within the viral capsid without relying on scarce and costly chemicals. In short, they needed a new test protocol.

He assembled a team of four researchers, with diverse scientific backgrounds, and outlined their challenge in a Zoom call on May 3. For the system to be fully operational in time for undergraduates’ return to campus in August, the testing program had to begin in early July. “That meant we needed to have the test developed within 6 weeks,” he says. “We were all under tremendous pressure.”

‘Order whatever you need’

One of those researchers was Diana Rose E. Ranoa, who joined Hergenrother’s lab as a postdoctoral fellow in November 2019. With a background in immunology and molecular biology, she’s usually responsible for testing the anticancer properties of compounds developed in the lab. But on May 4, she and her colleagues started setting up a research lab to develop the new saliva test. “The university pretty much told us, ‘Order whatever you need,’ ” she says.

So they ordered three Thermo Fisher Scientific QuantStudio 7 RT-qPCR machines and borrowed three simpler QuantStudio 3s from other research groups while they waited for the new equipment to arrive. They also ordered inactivated SARS-CoV-2 virus from BEI Resources, a repository that provides samples and reagents for studying infectious diseases, so that they could spike their own saliva samples with the virus to test different spit processing methods.

After trying out a range of conditions, the researchers soon found a workable protocol that met Hergenrother’s challenge. A subject drools saliva into a 50 mL tube and caps it; lab workers then heat the sample at 95 °C for 30 min to deactivate the virus. After that the sample can be safely uncapped to add a simple buffer solution and a detergent, which makes the saliva less viscous and disperses the RNA in solution, making it ready for RT-qPCR (bioRxiv 2020, DOI: 10.1101/2020.06.18.159434).

For the PCR assay itself, the team used a commercial Thermo Fisher TaqPath kit containing primers and fluorescent probes that can detect three genes from SARS-CoV-2. Checking for three genes at once improves the reliability of the test. The Shield team had also worked with the kit manufacturer to ensure there would be enough supplies once UIUC scaled up to the full screening program. “By the end of May, we knew that the technique worked,” Ranoa says.

They also managed to find a location on campus where they could run thousands of tests every day: UIUC’s veterinary diagnostics laboratory. It was already set up for COVID-19 work—it had tested samples from the Bronx Zoo and found that one of its tigers was COVID-19 positive. That made it relatively straightforward to certify the lab under the Clinical Laboratory Improvement Amendments (CLIA) statute, a legal prerequisite for any lab offering diagnostic testing for humans in the US. The Shield team moved into the lab on May 30. “There was nothing in there; it was just a space,” Ranoa says. “So we started shopping for equipment.”

Their spree included two more PCR machines; a trio of Beckman Coulter Biomek i5 automated workstations, which transfer samples between sample tubes, 96-well plates, and 384-well plates; and half a dozen 89 L circulating water baths, big enough to heat the flood of saliva samples that would soon pour into the lab. While they waited for that cornucopia to arrive, it was time to tackle the next problem: how to get the samples from students’ mouths into the freshly minted lab.

To manage the logistics of sample collection, the Shield team turned to Laura Wilhelm-Barr. As senior director of special events at UIUC, she normally helps coordinate campus-wide events, such as commencement, that might see 20,000 people on campus. “We had a lot of experience in crowd management,” she says.

In early June, she produced a map of possible sample collection sites around the 26 km2 campus, their locations coinciding with convenient parking spots and the most common routes taken by undergraduates. Open-sided tents started going up at those sites a couple of weeks later, until there were 40 sample collection lines housed in 17 tents around campus.

The procedure the team developed starts when students arrive at a tent. They swipe their university ID cards to get a bar code sticker for their sample. Then they drool in the tube and leave it in a rack for collection. Every hour, a car picks up the next load of saliva and delivers it to the lab. “There’s this huge chain that has to keep flowing, or things get bogged down,” Wilhelm-Barr says.

Some students might regard the regular pit stop at the sampling tent as an imposition. But Moca Yok, a third-year student in community health at UIUC, says that it has not been inconvenient. “They have many testing sites here on campus. It’s all reasonably close to where people live,” she says.

Access app

By July 6, the system was ready to go, and UIUC started testing faculty members, running about 700 tests per day. “But testing on its own doesn’t do anything per se,” Hergenrother says. “You also need to remove positive cases from the population. If you don’t do that part, then you’re just keeping a scoreboard.”

Any CLIA lab that performs a COVID-19 test has to report the result to the state’s disease surveillance system, which then filters it down to the relevant county based on the person’s address. In Champaign County, where UIUC is based, public health staff have 24–48 h to put that case into isolation and track down anyone else who recently spent more than 15 min in continuous close contact with that person, so that they too can quarantine.

UIUC needed to create a parallel system that would notify people of negative test results, which could be used like a passport to give students access to buildings on campus. Fortunately, the university already had a head start on that.

William C. Sullivan is a professor of landscape architecture at UIUC, but in 2018 he pivoted to lead a new initiative called Smart, Healthy Communities. It developed an open-source operating system called Rokwire to support the rapid development of apps that enable healthy lifestyles—by suggesting walking or cycling routes to appointments, for example.

His team used Rokwire to develop a free app called Safer Illinois, which gives users their COVID-19 test results and prompts them when they need to take another test. Users can also opt in to receive exposure notifications, a system that uses Bluetooth to warn if they have been in close contact with someone who subsequently tests positive.

“Part of the reason it’s successful is that it limits access to places that people want to go,” Sullivan says. Students need a negative test result to access university facilities, which drives them to the testing sites and improves compliance with the system, he says.

Back to school

UIUC started testing undergraduates on Aug. 15, and the diagnostic lab went into overdrive. Roughly 20 technicians work in 8 h shifts to keep the lab running around the clock, and on Aug. 24 they carried out 17,656 tests, their highest daily total to date. “I’ve seen all different types of saliva,” Ranoa says, laughing, “all different colors and consistencies.”

Many of the lab’s testing data are aggregated on a public dashboard, which displays the number of daily tests, the number of positive results, and the case positivity rate. On Aug. 31 there was a huge and unexpected spike: 230 new cases in a single day. The combination of frequent testing and contact tracing allowed the Shield team to find where clusters of positive cases were breaking out—largely in dormitories, sororities, and fraternities. Many of these cases were traced back to a handful of parties on campus that became superspreader events.

Screening can never guarantee low case numbers if people are unwilling to comply with public hygiene rules. So on Sept. 2, UIUC ordered a partial 2-week lockdown for students and also suspended a handful of students for breaking COVID-19 rules. That has helped improve compliance with the system, Vaid says: “Anecdotally, I heard that students were most scared of having to go back to their parents’ basement and spend the next 4 months there” if they were suspended for breaking the rules, he says.

The university also started testing some undergraduates three times per week and reduced testing for staff. “Our data shows that faculty, staff, and grad students are following social distancing guidelines very strictly, so we seldom see positive cases in that population,” Ranoa says. Targeting the tests at undergraduates helps ensure the team doesn’t bust their lab’s capacity, she adds, ensuring that results are now returned in 5–12 h.

The Shield team also realized that the flow of test results through the public health system introduced a time lag of up to a day before people got their results. Even the fastest test is pointless if people cannot act on it. So UIUC created a crack team of contact tracers, dubbed Shield Team 30, who contact positive cases within 30 min of the result—by text, phone call, or even in person. For a while, Shield also set up “high priority” racks at sample collection stations so that if people think they have been in contact with symptomatic individuals, they can expedite a test.

Broadly, these measures have worked. The weekly case positivity rate declined and stood at 0.47% on Oct. 25. Rates below 5% are one indication the epidemic is controlled in a community, according to the World Health Organization. By that time, Shield had carried out over 600,000 tests and identified almost 3,000 positive cases. So far, none of these people have been hospitalized with COVID-19, and none have died from the disease.

Each individual test requires just $10 of reagents. But add in other running costs, including consumables and staff, and Hergenrother estimates that the whole Shield program burns through about $3 million–$5 million every month. That’s on top of the $6 million–$7 million that UIUC estimates it took to establish the system, including the CLIA lab.

“Even though the test is cheaper than any comparable PCR test, it is still expensive,” Vaid says. “To set up a system like this, you need a complex laboratory, staffing, and collection sites everywhere, and it all has to be done within a few hours to be effective. It takes a lot of logistical support and a lot of financial support as well.”

Jump in the pool

UIUC isn’t the only institution that decided to set up a testing system rather than switch to online-only instruction. An analysis by researchers at Cornell University concluded that running the fall semester entirely online could actually result in more infections and hospitalizations among students and staff than holding classes in person. Most students planned to return to the area anyway, so requiring them to undergo frequent testing gave the university a better chance of controlling infections. Cornell is now running a screening system that expects students who live on campus or in the Ithaca, New York, area to provide a nasal swab twice per week for testing.

Other universities have turned to established names for their testing needs. In March, Broad Institute of MIT and Harvard turned part of its CLIA-certified lab into a COVID-19 testing facility to support local hospitals and clinics. It now tests about 70,000 nasal swabs per day using RT-qPCR, with results usually provided within 24 h, to support screening programs at more than 100 public and private colleges and universities, along with nursing homes and homeless shelters.

In collaboration with Color Genomics, a health technology company in Burlingame, California, the Broad lab also provides COVID-19 screening services for more than 80 companies across the US. “It gives employees a tremendous sense of safety knowing that everybody coming in has tested negative twice that week,” says Sekar Kathiresan, chief executive of Verve Therapeutics, a biotech in Cambridge, Massachusetts, that uses Color’s service.

Meanwhile, a New York biotech firm called Mirimus has developed a weekly saliva-testing protocol called SalivaClear, used by schools, universities, and even a movie studio. Samples are delivered to Mirimus’s labs, where automated robotic systems combine groups of 24 samples into a pool. This pool undergoes RT-qPCR—if it returns a positive result, each sample is then tested individually to identify the infected person. These individual tests rely on a protocol developed by researchers at the Yale School of Public Health, that is similar to the UIUC test (medRxiv 2020, DOI: 10.1101/2020.08.03.20167791).

The pooling strategy should save money and boost Mirimus’s testing capacity to about 50,000 samples per day by December, the company’s CEO and cofounder, Prem Premsrirut, says. “Individual testing eats up so much resources and reagents,” she says. “There’s just no need to be testing everyone on an individual basis.”

Another way to cut costs and avoid supply bottlenecks involves using microfluidics, which can test for SARS-CoV-2 using mere microliters of reagents. On Oct. 8, biotech company Fluidigm announced that it was launching a saliva-based testing program for US colleges and universities that would cost as little as $5 per test. The program would provide surveillance testing to keep track of COVID-19 rates on campuses, not diagnostic results for individuals. Its Advanta Dx SARS-CoV-2 RT-PCR assay relies on a single-use 192-well microfluidic cartridge, which plugs into the company’s Biomark PCR machines. “Our approach shows that PCR is not intrinsically expensive,” Fluidigm’s CEO, Chris Linthwaite, says. Each Biomark machine can run about 6,000 tests per day, and by the end of the year, Fluidigm should be able to produce about 2 million of these cartridges per month.

Plug and play

UIUC is now rolling out its screening system more widely, branded as Shield Illinois. Seven more labs are being prepared across the state, each able to process 10,000 tests per day for various organizations at a price of about $20–$30 per test. A spinout company called Shield T3 is readying five mobile labs in trucks that can drive to a site and carry out 10,000 tests per day.

Could such screening efforts be scaled up even further? The UK government, for example, announced in September that it would invest £500 million ($657 million) to support trials of a nationwide screening system, dubbed Operation Moonshot, which would test the entire population of the UK on a weekly basis.

Some researchers believe that national screening programs would be ruinously expensive and potentially divert funds from more effective strategies, such as education about public hygiene measures. “I don’t know that routine asymptomatic testing is going to help in a cost-effective way,” says Matthew A. Pettengill, a clinical microbiologist at Thomas Jefferson University Hospital.

Others argue that cheaper point-of-care diagnostics, which don’t need to be run in a lab, would enable this kind of population screening. For example, Abbott’s BinaxNOW detects SARS-CoV-2 antigens using a lateral flow assay, like a home pregnancy test. Each test costs $5 and takes 15 min to show whether viral proteins are present in a nasal swab.

Yet antigen tests like this are inherently less sensitive than RT-qPCR because they do not amplify viral material. In some cases, that can mean infectious people are incorrectly given the all clear—as seen at the White House, where visitors were being screened with BinaxNOW right before the outbreak that infected President Donald J. Trump.

Antigen tests are also more likely to produce false positive results than RT-qPCR. If they were used to test millions of people every day, it could cause many thousands of people, and their contacts, to isolate needlessly.

“What we really need are point-of-care tests for saliva that are just as sensitive and accurate as PCR and that can be done in 15 minutes,” Hergenrother says. So far, no saliva assay can combine the accuracy of PCR with the convenience and speed of an antigen test, he says. But as new diagnostic technologies become available, Hergenrother says, it should be straightforward to incorporate them into the UIUC screening system. “The logistics are a big part of it, so having that up and running means we can plug other diagnostics in as they emerge,” he says. “If our own test isn’t obsolete in about 6 months, I’m going to be disappointed.”

Mark Peplow is a freelance science writer based in the UK.