ACA.lafcorona197.072420

National Guard Specialist Mary Felton waits for residents to complete the free Covid-19 test at Cajun Field on Thursday, July 23, 2020 in Lafayette, La.. The testing will run 8:00am to 4:00pm through August 2.

Among the slew of data points that health authorities across the U.S. release daily, one is becoming increasingly fashionable: the “positivity rate,” commonly understood as the percentage of new tests that are positive.

It is a multifunctional stat that proves case increases reflect spread of the disease and are not simply the result of more testing. It is also seen as a concrete benchmark for safe reopening. If the positivity rate in a community remains below a certain threshold, it is seen as evidence that testing capacity is adequate and the virus has been contained.

Yet for all its persuasive power and convenience, calculating the positivity rate is a bit more difficult than it seems at first blush. And depending on which method is used, the results can vary quite dramatically.

The difference in how positivity is calculated is “very technical, but very important” as the rate becomes a focal point in public policy, said Cyrus Shahpar, who led the Centers for Disease Control and Prevention's response to global public health emergencies in 2016 and 2017.

Louisiana Department of Health officials, following the CDC, calculate positivity as the percentage of all tests on a given day that are positive — including those for people testing positive a second time. Much of the public, meanwhile, thinks of positivity as new cases — excluding retests — as a percentage of the new test results.

The latter method tends to result in a lower rate, as it removes a swath of positive results, and it’s that method that tends to get the most attention. To cite one closely watched example, the Johns Hopkins University daily coronavirus testing tracker calculates a rolling positivity rate for each state by reporting new cases as percentages of new tests, based on the raw daily totals that states report.

The differences in ways of calculating positivity rate is important because the results are significantly different, and the state has identified a specific positivity goal — below 10% — as one threshold for safe reopening. Asked Tuesday about reopening bars, for example, Gov. John Bel Edwards replied that “now is certainly not the time to entertain that,” because case counts are still high with “positivity that exceeds 10% throughout the state.”

“You’re starting to see it nationally, decisions being made on positivity, especially around schools,” said Shahpar, who now works with Resolve to Save Lives, a team of global health experts that has evaluated how states report coronavirus data. “The ideal case would be we are all doing it the same way, and I don’t know that we are right now.”

State officials say that strictly looking at all tests, including retests, is the more accurate methodology, because it ensures consistency of variables when calculating percentages.

If someone tests positive once and then again three weeks later, the second test result is not included in the daily case count, which is the numerator in the common way of calculating positivity. But the second positive tests are still included in the overall test volume, which is the denominator, resulting in a deflated percentage.

“It gets really complicated to try to change these data points around. That’s why we thought, rather than trying to make up a metric and be accused of playing with the data, we are going to go with strict epi methods and CDC,” said Alex Billioux, the assistant state health secretary.

State health officials also prefer their method because it looks at tests by the day the tests were taken, not by the day results were reported to the state. The daily case and test numbers get reported to the public as they come in from various labs that provide results at different speeds.

It usually takes a few days for health department staff to sort these results by test date, meaning the state’s positivity rate always refers to past dates. Daily tracking organizations prioritize the most recent available data to help the public understand trends in real time. The best way to do that is to calculate cases as a percentage of tests, said Jeff Asher, a data analyst who writes columns for The Times-Picayune | The Advocate explaining coronavirus trends.

“We do it that way because that’s the only data available,” Asher said. “If the state wants to provide us with additional data to do a more detailed, accurate assessment, I’d be happy to. In the absence of that data, we have to make do with what we have.”

The difference in calculation methods would not matter much if they yielded similar percentages, but they don’t. The number of positive tests exceeded the number of cases by more than 40% as of July 15, according to the most recent data available on Wednesday. That translates to a similarly large gap in the positivity rate, depending on how it is calculated.

On July 21, for instance, Edwards said the state’s positivity rate for the preceding week was 15.5%. That would have surprised anyone following Johns Hopkins’ posts, or those of most other daily trackers that look at newly reported cases as a percentage of new tests. By their estimation, the state’s positivity rate that week was 10.3% — within reach of the state’s goal.

Edwards himself has confused matters by using both methodologies in different situations. On July 8, when the governor referred to a daily positivity rate derived from the more intuitive calculation that tends to be available in real time. The state had recorded 1,891 cases that day, he said, along with 18,139 new tests, meaning the “positivity on those tests is just slightly over 10%.”

Edwards did not clarify that he was using a different methodology on July 21, when he cited the 15.5% rate for the preceding week. His spokeswoman, Christina Stephens, said Tuesday that Edwards follows the method used by the health department and CDC when making public policy.

“That said, it is natural to want to do the ‘back of the envelope’ math by dividing cases by total tests,” Stephens said in an email.

Billioux said he understands the desire to calculate positivity by new cases, because new caseloads are reported daily and people are hungry for the most up-to-date information.

“I am not going to say it’s wrong,” Billioux said. “I’m just saying from an epi standpoint, it’s not the way we would measure it, because it ignores the fact there are positive tests being done and are reported to us.”

The difference in how positivity rates are calculated is not merely about timeliness. There is also some disagreement about what’s most useful from a public health perspective.

While government epidemiologists say that strictly looking at total test volumes is the only way to come up with an accurate positivity rate, other experts such as Shahpar say this doesn’t adequately capture how the virus is spreading.

“Ideally we want to know about people, not the test itself,” Shahpar said. “If it’s five people getting tested 10 times a week, it’s more important to know there are five people than there were 50 tests done,” Shahpar said.

At the same time, the state and CDC method provides an important measurement because people testing positive a second time are, in fact, still positive, and thus capable of spreading the virus.

“Presenting the profile of what proportion of people got tested are still shedding virus is a relevant, valuable public health metric,” said Susan Hassig, an epidemiologist at Tulane University.

Whichever variables are used, Hassig said health authorities should be clearer with the public about what they are calculating. Epidemiologists are not typically accustomed to the limelight, Hassig said, and the field is struggling to figure out which information is best to showcase and how it should be presented.

“It’s really hard to judge from the geek, nerdy epidemiology side, just how detailed information should be that you are communicating with the public,” Hassig said. “There is a great deal of concern about people zoning out.”

Jeff Adelson contributed to this report.


Email Ben Myers at bmyers@theadvocate.com. Follow Ben Myers on Twitter, @blevimyers.