A new study shows rapid tests detect >85% of probable infectious cases (cases with high viral counts).

The ability to detect these cases was just as good in symptomatic and asymptomatic people. This is great news!

But that's not how the New York Times reports it.
(1/11)

Instead of focusing on sensitivity of probable infectious cases (cases with high viral count), the NYT leads with crude sensitivity in *all cases* – a much less relevant metric – which shows only 32% sensitivity for asymptomatic people.

https://www.nytimes.com/2020/11/02/health/coronavirus-testing-quidel-sofia.html
(2/11)

Crude sensitivity is simply not the relevant metric. Rapid tests are meant to be a public health tool, not a diagnostic tool. When used in the asymptomatic population, what matters is how well the test can detect people who are *infectious*.
(3/11)

The NYT article eventually reports the much more relevant metric... 22 paragraphs in!

(Note: “Ct values” move in the opposite direction of viral count. Ct values below 30 correspond to high viral counts. Ct above 30 indicates trace amount of virus.)

Rapid tests detect >85% of asymptomatic cases with high viral load.

Why did the NYT bury this relevant metric?

The answer is that there is no clear cutoff for when infectiousness starts, and we admittedly don't know with exact precision when it becomes practically relevant.

Ability to detect trace amounts of virus misses the point. Over beers, even skeptics agree that higher viral counts are much more likely to be infectious. If a test is especially good at detecting infectious cases, that's useful!
(6/11)

When it comes to deciding whether rapid tests are useful, it doesn't matter if infectiousness is unproven at Cts of, say, 31 vs. 27. We don’t fully understand why certain people are infectious. But we do know it typically corresponds to a high viral counts.
(7/11)

No reasonable person thinks we should not wear masks because we don't know whether they block precisely 50% or 80% of infections.

Similarly, we shouldn't oppose rapid tests because we don't know whether they detect precisely 70% or 90% of infectious people.
(8/11)

NYT journalist, @KatherineJWu, acknowledges that rapid testing is useful, and that the public must be educated about its limits. We agree!
(9/11)

But this is a world where Scott Atlas and the current administration want to reduce testing. To support public health, it is critical to lead with the test’s usefulness, before the limitations, especially if those limitations come from irrelevant metrics.
(10/11)

Good framing: Explain how rapid testing is useful but also advise caution about the unknowns.

Bad framing: Lead with an irrelevant crude sensitivity metric, and then 22 paragraphs later report the much more practically relevant metric.

(11/11)

One final point: There is a quote in the article that is deeply confused about how probabilities work. It should not be in the article.