The Homa Files

… and, her answer made me very nervous
==============

Yesterday the WSJ reported that:

“Health experts  now believe nearly one in three patients who are infected are nevertheless getting a negative test result.”

For details and links, see our post:
WSJ Shocker: 1 in 3 infected patients gets a ’false negative’ test result.

I expected that the story would send shockwaves around the DC science community and, for sure, be a reporter’s question at the daily Coronavirus Task Force press conference.

I was half-right … Dr. Birx was asked the question: “What about the report of 1/3 false negatives?”

Paraphrasing her answer:

1. We have to look into that – probably anomalies since….

2. Test results across sites are fairly consistent … and about what we’d expect.

3. If true, 1/3 false negatives would  mathematically give an unrealistic incidence rate.

Let’s drill down on each of those answers:

=============

1. We have to look into that – probably anomalies. 

I was expecting her to say something like:

My team is looking into it right now. We’re taking it seriously. If the report is true, it would be a very concerning situation.  For now, we still have confidence in our tests and procedures.

My opinion: Dr. Birx basically blew the question off as no big deal.

Not what I wanted to hear.

==============

2. Test results across sites are fairly consistent … and about what we’d expect.

Measurement “quality” is a function of two measures: reliability and accuracy.

Reliability means giving the same answer when given the same inputs.

For example, if you weigh yourself now and then again in, say, 15 minutes … you expect the scale to give you the same weight.

That’s reliability.

But, your scale may weigh heavy (or light) by, say, 5 pounds.

That’s inaccuracy.

Your scale is giving results that are reliable, but not accurate.

So, what Dr. Birx was saying is that the tests appear to be reliable … giving the same answer when measuring the same inputs.

But, that says nothing about accuracy.

All of the test kits may have a systematic error that results in approximately the same level of false negatives.

If so, there is an across-the-board problem … not an localized anomaly.

===============

3. If true, 1/3 false negatives would  mathematically give an unrealistic incidence rate.

Birx reported that she’s getting positive test results in the range of 8% to 30% … centering around 20%.

She says that if there were 1/3 false negatives, the incidence rate would so high that it’s not realistic … just not possible.

Hmmm.

Let’s start with the math.

To make the math easy, assume that the reported positive rate is 25% and that 1/3 of negatives (75%) are “false negatives”.

OK, if 75% of the test results are negative … and 1/3 of them are false readings … then true negatives are only 50% … and true positives are 50% … the original 25% positive plus 1/3 of the negatives.

OK, so is 50% an unrealistic, impossible number?

Hardly.

Keep in mind that only people with coronavirus symptoms (high fever, cough, breathing difficulty) are being tested.

That’s called adverse selection … it’s not representative of the general population.

Given the adverse selection — people with coronavirus symptoms — it wouldn’t surprise me if half tested positive.

In fact. I’m skeptical of the low rates.

If the symptomatic people being tested don’t have coronavirus, what do they have?

==============

Bottom line: After hearing Dr. Birx’s answers, I still think the WSJ report is a big deal.

==============

Follow on Twitter @KenHoma

 

>> Latest Posts

#HomaFiles

This entry was posted on April 3, 2020 at 10:25 am and is filed under Coronavirus - Covid 19. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.