Skip to main content

How well do workplace COVID-19 screens work? And do they discriminate?

A novel COVID-19 survey highlights strengths, drawbacks, and trade-offs of employee health screening

October 20, 2020

Author

key image
Jake MacDonald/Minneapolis Fed

Article Highlights

  • Unique survey reveals trade-offs and concerns raised by workplace screening tools
  • Screening techniques likely identify many workers as high-risk on any given day; methods vary in sensitivity
  • Demographic groups report symptoms at different rates, raising issues of workplace equity
How well do workplace COVID-19 screens work? And do they discriminate?

Lacking a fast, accurate COVID-19 medical test, many employers have implemented symptom screening protocols to safeguard employees and customers. With these screens, employees may have their temperature taken and/or answer questions about potential symptoms or high-risk behaviors prior to entering the workplace.

These screening practices raise a number of difficult issues. Do they accurately detect infection? Are some screening questions or techniques more accurate than others? Do they disproportionately flag some demographic groups, leading to workplace discrimination?

A recent working paper from the Minneapolis Fed’s Opportunity & Inclusive Growth Institute sheds light on these questions, using results from a nationally representative survey that asks about employment status, financial security, COVID-19 symptoms, health status, and personal efforts to avoid infection. The paper’s authors, economists Krista Ruffini of Georgetown University, Aaron Sojourner of the University of Minnesota, and Abigail Wozniak, director of the Institute, point to drawbacks and benefits of various forms of screening and provide guidance on factors employers should consider when designing workplace safeguards.

They emphasize several findings:

First, workplace screens “will likely identify many workers as high-risk on any given day.” Depending on the screening technique used, as many as 7 percent of workers could be flagged as potentially infected.

Workplace screens “will likely identify many workers as high-risk on any given day.” As many as 7 percent of workers could be flagged as potentially infected.

Second, the screening method matters. How a survey is designed and the number and type of symptoms that employees are asked about will affect which individuals are identified as possibly sick and prohibited from entering the workplace. This raises issues of discrimination, since different demographic groups report symptoms at different rates: Women, younger workers, and non-Hispanic Whites, for example, report multiple symptoms at higher rates than others.

Third, while indirect evidence suggests that screening doesn’t perfectly gauge actual COVID-19 status and positive screen rates are higher than actual infection rates, “these shortcomings may not disqualify workplace screens as an important public health tool,” the scholars write. Because positive screens may lead workers to engage in positive health behaviors—staying home when ill, seeking tests when symptomatic—they may lower overall disease rates. Moreover, in the absence of widespread, routine medical testing, symptom screens are likely to remain a common tool for containing the coronavirus.

How a survey is designed will affect which individuals are identified as possibly sick. This raises issues of discrimination, since different demographic groups report symptoms at different rates.

Finally, the economists discuss several points for employers to consider. As screening will likely provide high positive rates and prohibit employees from entering the workplace, it is important to design screening methods that do not inadvertently encourage misreporting—for example, by ensuring that workers won’t lose income if they screen positive. There is also a choice between screens that provide higher false negatives with fewer demographic disparities and those that yield lower false negatives but more disparities. Therefore, the harms and benefits of particular strategies may change as local caseloads change.

The COVID Impact Survey

The researchers analyzed data from the COVID Impact Survey (CIS), a nationally representative survey conducted in three waves of roughly 8,000 respondents each between April and June 2020. The survey asked individuals about fever- and COVID-19-related symptoms, exposure to COVID-19, behaviors to avoid infection, and labor market engagement. The symptom questions are similar to those used on current screening tools. About half of respondents provided their current temperature, measured by thermometer, as part of the survey.

From these results, the researchers develop seven COVID-19 screens, varying in type and number of symptoms queried. One screen is a simple thermometer check of temperatures 99 degrees or higher. Another asks about COVID-19 symptoms from a 17-item checklist. They then analyze how screens differ in detection levels, over time, and for different demographic groups.

Perhaps the strongest message from the results is that, in the economists’ words: “A substantial share of the workforce would screen positive under any of the screens.” Shares range widely, though, depending on the screen. Just 4 percent of respondents registered body temperatures of 99 degrees or higher on the day of survey response, for example, but over half said they experienced at least one COVID-19-related symptom during the previous week.

The researchers note that positive rates were fairly level over time and considerably lower when screens require two or more symptoms. They also mention that these results are likely higher than what a workplace screening procedure would find, since workers may tend to underreport symptoms if they fear being barred from jobs (and income) if they screen positive, and many of the questions in the CIS use a week-long look-back period, as opposed to 24 hours.

Are the symptom screens reliable? Do they discriminate?

The wide range of positive results across the seven screens raises the question of reliability: Do any of them provide a useful indication of actual infectiousness? After all, the researchers’ data do not include results from a medical test for COVID-19 infection. Instead, the researchers compare positivity patterns across the different screens and against other potential indicators of elevated infection risk. Although workers who screen positive under one screen are more likely to screen positive under another than those who do not, the overlap across screens is far from perfect.

No single question or screen is perfect. “Self-reported symptoms may contain information beyond a single temperature screen.” Employers should not rely exclusively on temperature-taking.

Importantly, though, no single question or screen is perfect. For example, those who screen positive on fever-symptom screens have higher rates of having at least one COVID-19 symptom than respondents with an elevated thermometer temperature reading. This, the economists write, “suggests that self-reported symptoms may contain information beyond a single temperature screen.” In other words, employers should not rely exclusively on temperature-taking when screening workers.

Examining whether workplace screens have disparate demographic impacts is “crucial,” say Ruffini, Sojourner, and Wozniak. If a particular screen flags one racial, ethnic, age, or gender group more often than another, regardless of actual infection, it may unfairly deprive those workers of their livelihoods. Yet if a screen regularly fails to catch infected workers of a given group, that group could be exposed to greater probability of infection. As context, the scholars observe that Black and Hispanic Americans have suffered higher COVID-19 case rates and higher employment losses than White Americans.

By gauging variation in screen indicators by race, ethnicity, gender, and age while controlling for marital status, income, children, and population density, the authors find that groups do indeed differ in their rates of measured and reported symptoms.

For example, women and young workers are more likely to report elevated temperature, and non-Hispanic Blacks and Hispanics are less likely than non-Hispanic Whites to report at least two fever symptoms. Women are more likely than men to report two or more fever symptoms. Young workers are more likely than older workers to report two or more fever symptoms. COVID-19-symptom screens show similar results to these fever-symptom lists.

The study reports a variety of other demographic variations and parses data from several angles. It finds, for example, that workers who report that someone in their home has been diagnosed with COVID-19 are significantly more likely to have a positive screen, but having a close friend or family member who died of COVID-19 is not predictive.

Trade-offs and dilemmas

To encourage truthful responses to workplace screens, employers may want to assure workers that they won’t lose income should they be flagged positive by a screen.

This paper is the only study to date to examine U.S. COVID-19 workplace screening techniques, and its value is enhanced through the unique information provided by the nationwide CIS.

It highlights the trade-offs and dilemmas that employers will face as they implement workplace screening. Broad questionnaires that ask about many potential symptoms will undoubtedly flag high percentages of employees as potential COVID-19 cases. Requiring more than one reported symptom for a positive result will permit more employees to enter the workplace. But demographic groups report symptoms at different rates, raising equity considerations for both employers and workers.

The economists note that to encourage truthful responses to workplace screens, employers may want to assure workers that they won’t lose income should they be flagged positive by a screen by, for example, adopting paid sick leave. And employers should be aware that screens with higher negative rates have fewer demographic disparities, and the converse. Whatever the employer’s choice, “the design of these screens, including the number and types of questions, affects how many, and which, workers screen positive.”

Douglas Clement
Managing Editor

Douglas Clement is a managing editor at the Minneapolis Fed, where he writes about research conducted by economists and other scholars associated with the Minneapolis Fed and interviews prominent economists.