During this pandemic, a key factor in the rules surrounding what is permitted in our area, as well as the policies that Kendal adopts for its residents, has been Chester County’s “test positivity rate”. I have been noticing that the rate reported by the County differs from the rate reported by the New York Times (which provides rates for every US county). That difference caused me to wonder: which rate should we pay attention to? And a related question: why focus on the positivity rate at all? I’ll return to the various reported rates below, but first a word about positivity.
“Test positivity”—an imperfect metric. The definition of test positivity is simple: you take the total number of positive tests occurring in a given period (say, a week) and divide it by the total number of tests given during that period. The result is the percentage of positive tests.
It is not immediately obvious why this percentage is useful. It does not provide insight into the two key questions local health authorities would want answers to: what is the prevalence of Covid in my area, and how fast is it spreading. The problem is that measuring those two factors requires widespread and frequent testing, including testing of people with no symptoms, and that just isn’t happening in Chester County (or virtually anywhere else).
Test positivity is relatively easy to measure, because all test results (positive or negative) have to be reported to the county Health Department, and in turn to the state Department of Health. And test positivity does provide some useful information: if it is low, that means most people who are tested because they have symptoms that might be Covid do not actually have the disease. The chances of missing a significant outbreak are low. If the percentage is high, it suggests there may be an outbreak and an immediate increase in testing is warranted.
How high is too high? Guidelines from the World Health Organization recommend that restrictions should be in place if the positivity rate rises above 5%. They recommend that governments should not relax restrictions until the rate is below 5% for two consecutive weeks. With the exception of one late-summer period, Chester County’s positivity rate has been above 5% for a full year, and it is currently well above that level.
So what is the current positivity rate in Chester County? There are multiple answers to that question. When I checked the rates on April 6, Chester County’s Covid website listed it as 9.05%. The New York Times Covid county page had it at 7%. The CDC website said 8.78%.
What should we make of these differing values? Shouldn’t they match? I decided to try doing my own calculation. I went to the Covid “dashboard” maintained by the county and downloaded the data there that gives counts of positive and negative tests. Here’s what the dashboard data looks like:
Just by checking a few of these daily numbers, you will quickly see that the positives are frequently around 20% of the total (positives plus negatives). I calculated a 7-day average, because that’s what the county says its rate is based on. The result: an astonishingly high rate of 22.3%. Hmm. I knew that couldn’t be right, but I checked my work and I couldn’t find an error.
The county’s official positivity rate for the period, on the other hand, was 9.04%. Here’s a screen capture of that page:
I sent an email to the county Health Department, explaining my calculation and asking why my result was so different from theirs. I got a nice response from a person named Emma Crowley, who is the Health Department’s Surveillance Coordinator. She explained that the numbers I had gotten from the dashboard were inflated because they include “the number of positive PCR tests AND positive antigen/antibody/etc tests.” (PCR tests are the nasal-swab tests we get here at Kendal, and they are the “gold standard” for Covid testing.)
This suggests that a large number of non-PCR tests are being conducted in the county, and a much higher percentage of them are positive, thus inflating the numbers on the dashboard. Why these non-PCR tests should be positive so often is a mystery to me. But I am glad to have an explanation of the odd data.
So which number is right? As for the differences between the other positivity figures, I was able to find information that explained at least some of the discrepancy: the county and CDC data is based on a 7-day average, whereas the New York Times is a 14-day average. During a period of rising positives (which we are currently experiencing) you would expect at 14-day average to be lower, since it contains 7 additional days of values from a period when positives were lower. The difference between the CDC and the county values was only a quarter of a percent, which I would consider not worth worrying about. It could be something as simple as a difference in the timing of the postings. In other words, for practical purposes all three numbers are probably correct. The choice between a 7-day or a 14-day average is arbitrary—there is no single correct answer.
The inflated positivity data on the county dashboard (which suggests a positivity rate over 20%) really is a problem, though. It may be useful for looking at trends, but it gives a seriously flawed impression of how many people test positive for Covid in our county. Ms. Crowley writes in her email response: “We’re aware of the ‘Positive’ label confusion [on the dashboard] and are working to fix it.”
Addendum 4-9-21 p.m.: Just before publishing this post, I checked the websites mentioned above one last time. For what it’s worth, the Chester County website has just done its weekly update and is now showing a positivity rate of 7.96%., more than a percentage point below last week. The CDC site is showing 8.8%, and the New York Times is still at 7%. And the county’s dashboard is still showing daily numbers that work out to something above 20% when averaged on a weekly or bi-weekly basis.