Saturday, July 03, 2021

Is the Infection Rate in Santa Clara County Zero?

According to the Santa Clara County Covid-19 Cases Dashboard, new infections are running about 25 to 30 a day. According to the testing dashboard, the test positivity rate is 0.5%. Checking online for the false positive rate from the most reliable of the tests, I get a range of estimates which includes 0.5%. So we don't actually know from the test results whether anyone at all is currently infected.

When I first noticed that, I checked the death rate from the same source. It was running at about 17 deaths from Covid a week, which suggested that there were still infections. This morning, however, when I looked at the same dashboard, the death rate (7 day average) was shown as about 3. It wasn't a drop over time but a sharp change in the number given for past weeks, the same weeks that had shown a death rate of about 17 as of two days ago. 

My first guess was that they had for some reason replaced per week figures with per day figures and not bothered to tell anyone. But then someone pointed me at a recent news story reporting that the county had concluded it had been over reporting deaths and had just reduced its estimate for total Covid deaths by 22%. The reduction from 17 to 3 is a lot more than 22%, but the rate of actual Covid deaths is much lower than it was a few months ago so the ratio of mistakenly identified Covid deaths to actual deaths might be much higher — if, for instance, they tested everyone who died, using a test that had a significant false positive rate, and counted anyone who died with Covid as having died of Covid — as, according to the article, they had been doing.

An infection rate near zero with apparent deaths and infections almost all false positives fits the pattern of a rapid decrease in both deaths and infections down to a low level, followed by a roughly constant level of both thereafter. The rapid decrease implies a reproduction rate below 1, which should have meant a continued fall. 

So perhaps it stopped falling because it got to zero.


Tim Lambert said...

The latest figures from Victoria show zero cases from 24,247 tests, which would imply that the false positive rate is much much less than 0.5%.

DinoNerd said...

Does it matter? Santa Clara is surrounded by counties where rates have been going up rather than down. It's easy for people - and their infections - to cross county borders. Even if the rate in Santa Clara is currently precisely zero, it's unlikely to stay that way for long.

It may have hit the point where infections that come in don't spread, or don't spread far (r < 1), which would be wonderful. In fact, I think it probably has. But a rate of zero (rather than one that rounds down to zero) can only be a temporary thing, as long as the rest of the area still has cases.

OTOH, as a person living in Santa Clara, I recently ate out for the first time in 16 months. The covid situation here and now is indeed very good, and I hope it stays that way.

David Friedman said...

I don't think your figure can be correct — I found two different studies which estimated false positive rates from the most reliable of the tests, and they are way too high for that.

The article gives a number of tests given and a number of cases acquired locally, interstate or overseas, all zero. But it doesn't follow that no tests were positive — they may have retested people who came up positive in order to eliminate false positives.

Tim Lambert said...

Victoria does not not retest to check before reporting positive tests. That would make no sense. The point is report the positive test as quickly as possible so potential contacts can go get tests. If they retest and find a false positive they then can correct the count and adjust the list of exposure sites. See here.

Here's something from Public Health Ontario:

As of May 23, 2020, PHO Laboratory has detected false positive SARS-CoV-2 results on approximately 20
occasions among over 228,000 specimens tested to date for COVID-19, with ~11,000 specimens testing
positive. This represents a false positivity rate of less than 0.01% (specificity of >99.99%), which is well
beyond performance targets for a laboratory test, even acknowledging there may be some false positive
tests that are not detected.

David Friedman said...


If they don't routinely retest all positive tests, how can they know what the false positive rate is? That "even acknowledging" at the end tells us that they do not know what the actual false positive rate is, even if they claim to.

I am reminded of the definition of a Supreme Court justice as someone who grades his own exams.

Here is one of the pieces I found:

Tim Lambert said...

David, the false positive rate cannot be more than the positive rate.
Here is another source showing that currently in Australia there are 3000 tests per positive result, a positive rate of 0.03%. The false positive rate cannot be higher than that and is certainly less.

Your source (from Nov 2020) says that checking real lab results gave FPR of 0.2% to 0.9%. Again, since real tests in Australia have a positive rate of 0.03% and less, those numbers are no longer correct if they ever were.

David Friedman said...

Here is a more recent discussion:

It also has:

"Following analysis by an expert review panel, and retesting through the Victorian Infectious Diseases Reference Laboratory, two cases linked to this outbreak have been declared false positives. "

Which appears to show that Victoria retests. And that it gets false positives.