Skip site navigation

Diagnostic testing is critical to managing the pandemic, especially since some people who carry the virus display no symptoms and a vaccine remains, at minimum, months away. But to date the US has struggled to conduct enough testing, thanks to an inadequate supply of test materials as well as confusion over who should be tested. An additional diagnostic, the antigen test, emerged late this summer as a vital tool for monitoring the spread of COVID-19. But government health departments are not clearly or consistently reporting how these tests are being used. While antigen tests hold considerable promise for disease surveillance, their true impact won’t be known until we know the extent to which these tests are being used.

First, some background. The gold standard for identifying a case of COVID-19 has been a test that uses polymerase chain reaction (PCR) technology. PCR tests detect viral genetic material (RNA), and they rely on an “amplification” process in which RNA is copied multiple times to enhance the signal. PCR tests are hailed for their sensitivity even to small amounts of virus, but their performance comes at a price, both literal and figurative. They cost $50 to $100 to perform, and results often take a few days. For testing to be an effective tool to help slow the spread of the disease, results should take no longer than 48 hours. Any longer than that, and windows for proper isolation and contact tracing are lost.

Governments are now seeking to close the gap with newly available antigen tests, which promise faster results at a lower cost. Antigen tests have been used in the past to quickly diagnose respiratory viruses such as influenza A and B. They identify the presence of viral proteins, termed antigens, found on the surface of the virus. A positive antigen test reveals that a person is likely currently infected with the virus. 

Antigen tests confer advantages over PCR tests. First, they cost as little as $5 per test, making it economical to administer the test frequently. The US Food & Drug Administration (FDA) has approved six antigen tests under its Emergency Usage Authorization provision. Second, all of these tests can be performed outside of a lab, and some, like one from Abbott, do not even require specialized equipment: Healthcare workers simply take a sample from a patient using a nose swab and deposit it on a testing card along with a reagent. Results are available within 15 minutes.

The cost and flexibility of these tests align with the screening-based strategy from the US Centers for Disease Control and Prevention to test asymptomatic people who haven’t been exposed to SARS-CoV-2. The CDC recommends serial antigen testing in closed congregate settings, such as long-term care facilities; this strategy is crucial to support the reopening of schools and workplaces. 

The Trump Administration is sending 100 million antigen tests to states and territories and 50 million to communities at a heightened risk of viral spread, such as long-term care facilities and schools that serve minority populations. It hasn’t all been smooth sailing: When the Department of Health and Human Services sent tests made by Quidel and BD to nursing homes across the country this summer, demand far exceeded supply, and nursing home administrators were confused by the federal government’s guidance for using the tests. And earlier this month, Nevada’s public health department told nursing homes to stop using antigen tests because of a high rate of false positives.

Antigen tests have their benefits, but their drawbacks should not be ignored. Studies have shown that viral antigens can be detected only 5 to 7 days after the onset of symptoms, but the duration of infectiousness may be longer—up to 10 days after symptoms begin. Even under such ideal conditions, antigen test sensitivity, when benchmarked against standard PCR, ranged from only 84 percent to 97.6 percent. As a result, the CDC recommends treating antigen test results as presumptive evidence and suggests follow-up PCR testing, especially if a person has a high chance of being infected—meaning they either have symptoms or they’ve had contact with a known case. 

A positive antigen test can be used to identify a probable case of COVID-19, but not a confirmed case, according to the most recent guidance provided by the Council of State and Territorial Epidemiologists, which represents epidemiologists across the United States and provides technical support on epidemiological issues to agencies like the CDC. Other epidemiologists have suggested that antigen tests should primarily be used for symptomatic testing, and that the use of such tests for asymptomatic screening may be dangerous due to the possible errors that could occur.

Currently, about 1 million tests are reported in the US each day. But, critically, many state and US territories’ public health departments aren’t making clear which types of tests these are, both in terms of which tests are administered and which ones come back positive.

This matters because the different tests deliver different results, and these results, in the form of test positivity metrics, are being used to make critical decisions, like when to reopen schools and communities. PCR tests are used to diagnose people who already have a high likelihood of being infected with SARS-CoV-2 (such as presenting to a hospital with symptoms or having contact with known cases). Antigen tests are deployed to diagnose cases where there is no suspicion of infection (for example, screening in workplaces or random surveillance of the population). As a result, PCR test positivity is typically higher than antigen test positivity, as we would expect a higher share of the tests used to diagnose likely cases to return positive results. When test positivity figures mix two test types in their numerators and denominators, they muddle the picture of testing capacity they are meant to present. Lumping the two test types into one positivity calculation makes it difficult to assess how a state is faring on either front.

An analysis by the COVID Tracking Project finds that states are inconsistently reporting antigen testing data—if they are reporting them at all. What’s more, the numbers that are reported are suspiciously low, given what we know about the prevalence of antigen testing.

The COVID Tracking Project is a data transparency project, rather than a public health organization, so our recommendations focus on the most useful way to report COVID-19 data to the public. Earlier this year, we recommended that states report antibody test totals and results separately from PCR tests to maximize the usefulness and clarity of the data. We recommend the same for antigen tests: States and territories should report antigen test totals and results separately from both PCR and antibody tests. Doing so will benefit all researchers, media organizations, and members of the public who use the data in their work and decision-making, and it will be especially useful to the residents of each state and territory, who need to understand the full realities of local outbreaks and testing strategies.

How states are reporting antigen tests

Right now, only 12 of the 56 states and territories we track are reporting separate antigen test totals, while nine states are combining this data with information for other types of tests. Nineteen states are providing PCR test results only in their viral test reporting, and 16 states have unclear policies about how they handle antigen testing. What’s more, The COVID Tracking Project has yet to see a breakdown of antigen tests at the federal level. So, without better data from states, a significant portion of the US testing landscape will remain invisible.

Based on the COVID Tracking Project’s most recent analysis of publicly available documentation, the 12 states and territories reporting separate antigen counts are Arkansas, Guam, Iowa, Kentucky, Massachusetts, Minnesota, Missouri, Mississippi, North Carolina, South Carolina, Texas, and Utah. Our analysis focused on total test counts, rather than the numbers of COVID-19 cases identified through antigen testing. Meanwhile, at least nine states are known to combine antigen and PCR total test counts in one figure: Alaska, Alabama, Arizona, Florida, Illinois, Indiana, Kansas, Michigan, and Oklahoma. 

Our complete list of which states are reporting what information is available in our detailed spreadsheet. However, we would like to point out a couple of notable examples and trends which demonstrate the state of this reporting.

Missouri exemplifies the antigen test reporting approach that we hope to see from all states. The state foregrounds its count of total PCR tests on the front page of its dashboard, which is useful for researchers looking to determine representative test positivity rates. Meanwhile, Missouri also reports clear, separate values of antigen and serology tests in units of both specimens and people on a page dedicated to testing.

Other states are not so clear in their reporting. Utah, for example, foregrounds a total test number that combines PCR and antigen tests at the top of its dashboard. While separate counts of PCR and antigen tests are available in a time series chart, the figures change daily and are not easily accessible to casual dashboard viewers. The COVID Tracking Project has been unable to verify our total antigen testing figures for Utah, which are calculated by summing the daily values reported by the state.

We have also found that several states, including some of the 12 that report separate antigen test counts, are reporting state-calculated test positivity rates that combine PCR and antigen tests. This type of calculation presents a misleading picture of testing capacity. Finally, Virginia is a unique case in misleading test figures: While this state separates out a PCR test count, it lumps antigen and antibody tests—an even more disparate combination than antigen and PCR.

Among the states that report antigen test totals, most fail to release time series of these tests. Time series are important for understanding how test positivity and antigen test use have changed over time. While the COVID Tracking Project provides antigen time series for all nine states in our API drawn from our daily data collection, time series reported directly by states tend to be more accurate. Only Massachusetts, Texas, and Utah publish antigen test history going further back than September. 

The COVID Tracking Project is not the first to collect information on how states are addressing antigen tests; KHN recently painted a similar picture of states’ policies. The publication found that while 40 states require reporting of all antigen test results to state health departments, 18 states were lumping this figure into the total test counts. (Some of the states that KHN identified as lumping antigen and PCR tests were categorized as “unclear” in our annotations because their official documentation didn’t specify what types of tests were included.) This further indicates a lack of transparency in how states are communicating their testing performance to the public, echoing past alterations with antibody tests

One more twist: separately reported antigen test numbers appear to be undercounts

Even for states that are reporting clear, separate counts of antigen tests, we suspect that these values may significantly understate the true impact of these tests. As antigen tests may be administered on-site in hospitals, nursing homes, schools, and other facilities, they are not built into a state public health department’s reporting system in the same way that PCR tests now are. 

The body that advises epidemiologists, CSTE, recommends that a positive antigen test should only be considered as “presumptive laboratory evidence” that may be used to identify a probable case of COVID-19. We suspect that some states may be following this directive and are logging antigen tests as probable cases, but we do not have clear evidence as to how this practice may be impacting overall test numbers. Probable cases also often reflect counts of positive antibody tests, individuals with COVID-19 symptoms, epidemiologic linkage with a confirmed case, and death certificates. (For more details on probable cases, look out for a forthcoming COVID Tracking Project analysis.) As of the new directive, the probable-case designation based on antibody tests is no longer appropriate. 

If the sites that are doing antigen testing en masse—such as long term care facilities—are reporting their testing counts, these counts are not being made public. The federal COVID-19 nursing home data reported by the US Centers for Medicare and Medicaid Services includes 29 fields on testing, including whether the facility has access to testing and whether residents in the facility were tested since its last weekly report. But none of these fields report actual testing counts. These facilities have received thousands of antigen tests since July, yet we have very limited data from federal or state sources on how many of them have been used.

Counts of antigen tests in Texas, as reported by the Texas Department of State Health Services, are difficult to trust. Consider: As of October 5, the state had reported 12,700 positive results out of 137,000 antigen tests. But 15 Texas counties (out of a total of 254) reported a total of 28,300 probable COVID-19 cases—the majority of which were diagnosed via antigen testing, according to the Houston Chronicle—during that same time period. This disparity calls the accuracy of the state’s antigen test data into question.

Other measurements also suggest dramatic underreporting. Data shared with Carnegie Mellon University by test maker Quidel revealed that between May 26 and October 9, 2020, more than 3 million of the company’s antigen tests were used in the United States. During that same period, US states reported less than half a million antigen tests in total. In Texas alone, Quidel reported 932,000 of its tests had been used, but the state reported only 143,000 antigen tests during that same period. 

Given that Quidel’s antigen test is one of six in use, the true number of antigen tests performed in the United States between late May and the end of September was likely much, much higher, meaning that only a small fraction are being reported by states. 

American manufacturers are on track to make millions of antigen tests per week by the time winter hits. Facilities ranging from doctor’s offices to large corporations are on track to use them. Some universities and sports leagues are using these tests already. Public health departments will need to make significant changes to their data collection and display in order to accurately report the results of these tests—and in order for the public to best understand the spread of COVID-19. 


Many people at The COVID Tracking Project contributed to the research and data-compilation efforts that made this story possible. We would like to thank Kara Schechtman, Matt Dempsey, Elizabeth Eads, Daniel Lin, Jennifer Clyde, Jesse Anderson, Nadia Zonis, Erika Thomson, and other contributors for their tireless work on this and many other data quality efforts.


Betsy Ladyzhets is a Research Editor at Stacker and works on data quality and the COVID Racial Data Tracker at the COVID Tracking Project.


Quang P. Nguyen is a PhD candidate in the Department of Epidemiology at Dartmouth College.


More “Testing Data” posts

How Probable Cases Changed Through the COVID-19 Pandemic

When analyzing COVID-19 data, confirmed case counts are obvious to study. But don’t overlook probable cases—and the varying, evolving ways that states have defined them.

20,000 Hours of Data Entry: Why We Didn’t Automate Our Data Collection

Looking back on a year of collecting COVID-19 data, here’s a summary of the tools we automated to make our data entry smoother and why we ultimately relied on manual data collection.

By Jonathan GilmourMay 28, 2021

A Wrap-Up: The Five Major Metrics of COVID-19 Data

As The COVID Tracking Project comes to a close, here’s a summary of how states reported data on the five major COVID-19 metrics we tracked—tests, cases, deaths, hospitalizations, and recoveries—and how reporting complexities shaped the data.