Updated With New Editorial Note: August 19, 2020
You’d think that something like the percent of COVID-19 tests that are positive would be simple and straightforward. Turns out it’s not.
With the metric informing really important decisions like in-person school instruction and driving decisions like when bars and nightclubs can open- it’s super important to have a metric that’s as accurate as possible. We’re not there yet, but it’s fixable.
Why is the Percent Positive Number Important?
The number of new cases diagnosed per day often gets the headlines, but the percent of tests that are positive is a better indicator of community spread and especially the adequacy of testing.
It’s important to be able to test widely including people who are contacts of cases but only mildly ill or even asymptomatic so they can isolate if they are positive. The percent positive number gives a decent picture of whether testing is adequate to achieve that goal. As a rule of thumb, a 5% positive rate or lower indicates decent testing capacity and moderate to low community spread.
How is it Calculated?
This seems like a straightforward question, but there are at least 3 different ways people are calculating the percent positive rate in Arizona. Understanding how the percent positive is being calculated is super important now that the metric is being used to make decisions about how and when businesses like bars, nightclubs, gyms and waterparks as well as schools can and should open.
Here are 3 different ways one can calculate a percent positive metric:
Method 1: Straight Up Division
This approach takes the number of positive tests that are reported each day to the state health department divided by the total number of reported positive and negative tests. The vast majority are positive PCR tests but a handful are IgM positive serology tests with diagnostic support.
This is the percent positive number that you are probably most familiar with because it’s used by the COVID Tracking Project, ASU Biodesign Institute and Johns Hopkins University.
It’s also the number that you’d get from the ADHS website if you divide the daily number of reported PCR positives by the total PCR tests. It’s also the fastest way to check the percent positive because it can be calculated immediately every day.
However, there are some people that take issue with this method because, according to the ADHS Director, some labs are reporting the positive tests but not the negatives (despite the fact that Executive Order 13 requires all labs to report both positives and negatives). If that’s the case, then this method will give an artificially high percent positive estimate because some of the denominator is missing (the negatives).
This method of calculating the percent positive doesn’t reflect when the person was tested. Because of the historically long turnaround times, the results that come in each day can be as much as 2-weeks old.
The COVID Tracking Project, ASU Biodesign Institute and Johns Hopkins University all display the daily percent positive using this method.
Method 2: Backfill the Epi Curve
Another way to calculate the percent positive is much slower but more accurate. Using this approach, you can find out what the percent positivity was 2 weeks ago, but it doesn’t help you know what the percent positivity is real-time.
Because of the slow turn-around times and reporting delays the percent positive using this approach can only be calculated after the fact… even 2 weeks after the fact. With this approach you only know what the percent positive was last week or even the week before.
Under this approach, test results that come in each day are backfilled to the date in which the specimen was collected. So, a test that comes back from the lab today, but was sent to the lab 10 days ago, won’t be counted in today’s percent positive number. Rather, it would be backfilled to count as a positive on the date the sample was collected.
This approach gives you a more refined look at the percent positivity rate, but it can’t be calculated for several days after the fact because many of the tests are coming back from the labs so late (more than ½ of tests are still coming back more than 5 days after being collected).
Method 3: Backfill the Epi Curve & Exclude Results
The ADHS uses this method to calculate the percent positivity numbers. This method is similar to Method 2 and is also a retrospective look at what the percent positivity was a couple of weeks ago. They backfill new daily test reports to the date of specimen collection.
However, they only include the data that is submitted electronically for the calculations. The Department’s rationale for excluding the data that isn’t reported electronically is that some labs aren’t turning in their negative results (even though Executive Order 13 requires them to do so).
By excluding the non-electronically reported results, they avoid artificially inflating the percent positive figure, but they also don’t get the full picture because so many of the results are not used, and are left to defend why so many of the test results aren’t being used.
Which Method is ADHS Using?
ADHS is using Method 3 to calculate the percent positive numbers for each county. It is those numbers that you will see on the School and Business Dashboards.
Because they calculate percent positives differently than well-recognized national sources, the percent positive numbers that they report will always be different (lower) than the numbers that are tracked and displayed by The COVID Tracking Project, ASU Biodesign Institute and Johns Hopkins University.
It will take some nuanced messaging to explain the differences to the public. It will be confusing because ADHS will state that Arizona’s percent positive rate is significantly lower those well recognized national sources.
This will lead to the impression that ADHS has intentionally selected a method of calculating the percent positive that could lead to premature opening of businesses like bars, nightclubs and gyms. Their method has good merit, but excluding results from labs that don’t report electronically will build suspicion that something is amiss.
A sticky wicket indeed.
Editorial Note: ADHS’ decision to exclude all data that’s not reported electronically is puzzling and unnecessary. Executive Order 2020-13 requires all laboratories to report all their results – both positive and negative. ADHS has conceded that they are not requiring all laboratories to comply with EO-13 (signed on March 23).
The failure to enforce that Executive Order is the reason they have excluded much of the data set. This is impairing their ability to get an accurate estimate of percent positivity. Having an inaccurate measure of percent positivity means that important decisions like when in-person school is safe to begin and when bars, nightclubs, water parks and gyms can reopen.
A simple solution is to enforce the reporting provisions of EO-13. They could then most accurately measure percent positivity.
Let’s fix this, shall we?
New Editorial Note August 19, 2020: The ADHS now states that they indeed are getting all of the data from all of the labs, but that some labs aren’t submitting the data in the format that they would like. Their current position is that they will continue to exclude those data that aren’t in the format that they would like (about 30% of the data set). I expect this position to change and for them to begin inputting the data and using it as questions about their practice of excluding data are debated in the public square.