Pennsylvania is blending results from two entirely different types of coronavirus tests, an approach that boosts state testing numbers but that experts say can paint a skewed picture of COVID-19 infection rates, cases, and testing capacity.
Diagnostic tests detect coronavirus genetic material in a symptomatic patient’s respiratory sample, confirming a current infection. Antibody tests, which use blood samples, detect an immune response weeks after an infection — even if the infection caused no symptoms.
While neither test is perfect, antibody tests (also called serology tests) have been shown to miss up to half of people who previously tested positive with a diagnostic test.
Diagnostic testing results are among the most carefully watched metrics in determining whether it’s safe to lift stay-at-home orders and reopen businesses; this data blending can make the COVID-19 rate look artificially low. But with no clear national guidance on how to collect, analyze, and report data, the U.S. Centers for Disease Control and Prevention and states have adopted varying practices.
Pennsylvania’s data practice was first reported by the Atlantic. The magazine, which runs the COVID Tracking Project, said the CDC began doing a similar conflation this week, thus “distorting several important metrics and providing the country with an inaccurate picture of the state of the pandemic.”
A CDC spokesperson emailed that the agency plans to stop blending the data “in the coming weeks," but did not explain why it began doing so.
Pennsylvania Department of Health spokesperson Nathan Wardle confirmed that the state is combining results from tests that diagnose current coronavirus infections with antibody tests that look for immune response to a past infection.
Earlier this week, the department was faulted for releasing nursing home data that minimized the devastating toll COVID-19 has taken on staff and elderly residents.
The U.S. Food and Drug Administration says antibody tests may not be used alone to diagnosis a patient. It is hoped — but not yet proven — that the tests can be used for disease surveillance, and to identify people who can safely go back to work because they have some protection from reinfection.
Depending how results from antibody and diagnostic tests are combined, it presents a picture that is, at best, hard to interpret, and at worst, misleading.
For example, antibody tests tend to have a lower positive rate than diagnostic tests, because antibody tests are used on the general population, and only an estimated 3% to 5% in the United States have been exposed to the virus. So combining results from the two types of tests can drive down the overall positive rate.
That’s exactly what has happened over the last three days, since the CDC began adding antibody test results to diagnostic test data, according to the Atlantic. On Monday, a page on the agency’s website reported that 10.2 million diagnostic tests had been conducted nationwide since the pandemic began, with 15% of them — or about 1.5 million — coming back positive. On Thursday, the testing total had jumped to 12.9 million, 1.7 million of them positive, which pushed down the overall positive rate to 13%.
By combining the data from different tests, the CDC has made them both “uninterpretable,” Ashish Jha, director of Harvard University’s Global Health Institute, told the Atlantic.
Pennsylvania’s blending of the data could confound the interpretation in several ways.
In mid-April, when the state Health Department began receiving antibody test results from commercial labs and other sources, the department broadened its definition of a case, according to Wardle. Positive diagnostic tests were combined with positive antibody tests of people who also had experienced COVID-19 symptoms or high-risk exposure.
If antibody testing ramps up, that methodology could inflate the case count. So far, it hasn’t. Of the state’s total cases, Wardle said, less than 1% are based on antibody tests.
But Wardle said the Atlantic article is correct that Pennsylvania is also adding negative antibody test numbers to its negative diagnostic test numbers. That could make Pennsylvania appear to be doing more diagnostic testing than it actually is. And using that artificially large number of negatives in calculations could drive down the positive rate.
On Thursday, the state website listed 65,392 total cases, 304,514 negatives, and 4,869 deaths.
The practice of conflating different testing data is not consistent across the state. Philadelphia, which has by far the biggest COVID-19 outbreak in the state, is not reporting antibody test data to the state Health Department, nor is it including it in data on the city’s website, department spokesperson James Garrow said Thursday.
Pennsylvania Health Secretary Rachel Levine said at a news conference Thursday that if someone has a negative antibody test, “then they wouldn’t be included at all” in tallies on the state website.
Gov. Tom Wolf said Thursday that he wasn’t aware that the state mixes antibody and diagnostic test result data, saying, “Maybe there’s something that I don’t know.”
Levine also said the governor’s color-coded reopening formula is based on confirmed cases, not probable cases based on antibody testing.
“What’s important to note is we only use the count of confirmed cases when we’re looking at any metrics for counties going from red to yellow or yellow to green or any other transition,” she said. "And we’ll make sure it’s all very clear on our website.”