As Coronavirus Cases Grow, New Drexel Research Cautions about Impact from Inaccuracies in COVID-19 Testing

As researchers across the globe scramble to develop and improve testing for the novel coronavirus (SARS-CoV-2) that causes COVID-19 disease, preliminary research from Drexel University suggests they have their work cut out for them.

Results from two studies, recently published online prior to acceptance in a peer-reviewed journal, suggest that lack of data on the sensitivity of current tests could significantly skew the numbers of cases that are being diagnosed.

Igor Burstyn, PhD, an associate professor, and Neal D. Goldstein, PhD, an assistant research professor, both in the Dornsife School of Public Health, are collaborating with Prof. Paul Gustafson, chair of the Department of Statistics at the University of British Columbia in Canada.

Their work stresses the urgent need to understand just how accurate the current polymerase chain reaction (PCR) diagnostic test is. And their statistical modeling approach estimates a number of cases of COVID-19 could have been misclassified among people who are tested for it in Philadelphia. The methods that they developed can be replicated to estimate the effects of inaccuracy of testing in other cities. 

Burstyn and Goldstein provided some additional insight on their work and the importance of having accurate testing for both mitigation efforts and recovery from the pandemic.

You point out the dangers of false positives and false negatives, and note that the accuracy of current PCR-based testing is largely unknown in the United States and Canada. Is that why you studied Alberta, Canada in addition to Philadelphia?

IB: We looked for best available data to answer our questions. With so many health departments scrambling to manage local outbreaks, it was in fact quite difficult to find appropriate data, and it was by mere chance that the best data we could find was from places that we have personal connections with. We are now looking to expand our work to other jurisdictions with suitable data — this would include daily counts of both positive and negative tests.

Your research shows a worst-case scenario in which there could be hundreds of undiagnosed cases in Philadelphia — even among those tested. Will we get to a point when testing is sufficient to act on with great confidence, or is this a reminder to self-isolate and call your doctor if you are experiencing COVID-19 symptoms?

IB: Diagnostic testing, even when imperfect, is essential, and was the main point of our first paper. It would be a dangerous folly to avoid testing for those who have the signs and symptoms, and are deemed eligible for such tests by the CDC. We are optimistic that the PCR-based test will improve over time and we will learn more about its accuracy, as has been the case with diagnostic tests for other diseases. Of course, calling one’s health care provider is the essential first step for anyone who suspects that they have COVID-19, and this should lead to testing as appropriate. But our fear is that if an individual receives a laboratory test suggesting that they don’t have the virus – when in fact they really do – they will receive a false sense of security and potentially infect others if they start socially engaging. This is why it is so important to understand testing accuracy.

Can you briefly explain “herd immunity”? As a vaccine is developed, and more people recover from COVID-19, how will insights into the number of people who are not susceptible to COVID-19 help to prevent future coronavirus outbreaks?

NDG: Community immunity, as it has come to be known, is the tipping point at which enough individuals in the community are immune to a disease so that those not immune will be protected. This exists already for many infectious diseases through our robust vaccination program. For COVID-19 there are a few possible scenarios. One scenario is that enough people become naturally exposed and infected due to ongoing transmission during the pandemic to “buffer” people who haven’t been infected. This is far from ideal however, because those infected may go on to develop severe disease and possibly die, which could overwhelm our health care systems. The second scenario is dependent on a vaccine being developed to provide immunity without the harmful effects from natural infection. But that is at least one year away, if even possible. In the meantime, by socially distancing ourselves and self-quarantining and isolating we in essence have an interim herd immunity.

You mention an ideal two-stage testing scenario. Please describe how that would work and whether we are likely to get to that point.

NDG: With the recent FDA-approval of the first serological test, more testing options will become available to healthcare workers. Different tests answer different questions, to reflect the patient’s and provider’s needs and concerns. For public health surveillance, both tests are useful. The serological test indicates past or present infection and can be used to estimate the number of people who have ever been infected. The diagnostic test, and the focus of our work, is useful for confirming active infection, with subsequent isolation, quarantine and contact tracing.

One possible use of a two-stage test is to first obtain a highly sensitive antibody test to rule out coronavirus (perhaps it was the common cold that made the person sick). If that is positive, then the highly specific diagnostic test rules in SARS-CoV-2 infection. This is the paradigm used for other infectious diseases, such as Hepatitis C and HIV. However, given the uncertainty in the accuracy of either test at this point we cannot be sure how much confidence in the results will be gained at present.


You note that the PCR test accuracy was less of a concern early in the pandemic. But later on — such as at present — it is more crucial than ever to know. Please explain why.

IB: If early in epidemic among four people who tested positive only two are expected to be truly infected, there seems to be little societal harm in taking greater precautions and treating all four as if they were infected. This is when the prevalence of disease in the population is very, very low. However, as the pandemic unfolded and more people became infected, we need more accurate surveillance data to fully understand the trajectory of illness, especially from a health care utilization perspective. We would not want to underestimate the need for personal protective equipment, medicine and ventilators by an order of magnitude!

What are the next steps for your research group?

NG: We recently were fortunate to receive funding from Drexel University’s Rapid Response Research & Development Fund to expand the methods we developed to work with geographic data. This means that we should be able to improve the accuracy of surveillance maps generated by health departments.

Any other important points I missed?

IB: The most important point of our work is that as long as we do not know the true accuracy of diagnostic testing, we may be under-estimating how fast the pandemic is growing and over-estimating how fast it is subsiding.

This is because we are undercounting cases.  If we continue to undercount, measures employed to slow the pandemic, like stay-at-home orders, may be lifted prematurely, causing otherwise avoidable cases and deaths. In theory, this could result in the need for another set of highly restrictive measures to stop a second wave of the pandemic.  It is better to know just how accurate the testing is instead of assuming that it is perfect. 

How is the coronavirus affecting your life? Burstyn is also collaborating with Tran Huynh, PhD, an assistant professor at the Dornsife School of Public Health, on a study of experiences of Philadelphia residents during the coronavirus pandemic. The confidential survey is available here. The results will help to better prepare and support communities through the recovery and future epidemics. Media Interested in an interview with Burstyn or Goldstein should contact Greg Richter at 215-895-2614 or gdr33@drexel.edu.

Tagged with: