Links: Mostly COVID-19, less comforting

[sonia]

PSA: Coronavirus antibody tests aren’t as accurate as they seem by Amanda Shendruk & Tim McDonnell. Low overall prevalence makes a positive antibody test result only about 50% reliable. *sigh* ETA: See additional info from rosefox in comments about more accurate tests becoming available.


Not COVID-19 related, possibly encouraging.
How to Oust an Autocrat, Lessons from Chile by Teri Kanefield. "I’m using Chile as an example because my husband’s family experienced the Pinochet dictatorship, so I have first hand accounts. Bonus: I’ll also explain why I get so testy with the doomsayers."

Interesting perspective. Stockholm Syndrome – Not What You’ve Been Told by Jess Hill.

In 2008, a review of the literature on Stockholm syndrome found that most diagnoses were made by the media, not psychologists or psychiatrists; that it was poorly researched, and the scant academic research on it could not even agree of what the syndrome was, let alone how to diagnose it. Allan Wade, who has consulted closely with Enmark, says Stockholm syndrome is ‘a myth invented to discredit women victims of violence’ by a psychiatrist with an obvious conflict of interest, whose first instinct was to silence the woman questioning his authority.

COVID-19 related.
Why Days 5 to 10 Are So Important When You Have Coronavirus by Tara Parker-Pope. Description of COVID-19 symptoms and when to seek help. Take notes about when symptoms start.

Why the Coronavirus Is So Confusing by Ed Yong. A lot about the political clusterf*ck in the US. via amaebi.

What the Coronavirus Models Can’t See by David Wallace-Wells. Relaxing social distancing without effective vaccine or treatment invalidates our models.

These Are the States Opening Back Up for Business by Adam K. Raymond. A dispiritingly long list.

Comments

[rosefox]

Low overall prevalence makes a positive antibody test result only about 50% reliable. *sigh*

That's using the numbers in their simulation, which are artificial (90% sensitivity and 95% specificity doesn't correspond to any particular test on the market, and 5% disease prevalence is inaccurate for many locales). Please don't confuse that with real numbers in real situations! If you click the option to adjust the simulation, one of the options is to use the stats for Roche's brand new antibody test, which reportedly has a sensitivity of 100% and a specificity of 99.8%. That's very different from the simulation's default numbers, and it means that even in populations with a low incidence of infection, your odds of getting an accurate test result are effectively 100%. (Assuming Roche's numbers are accurate.)

Here's a chart comparing FDA-approved antibody tests released between April 1 and May 3. With the caveat that all the numbers there were provided by the companies making the tests, it's pretty astonishing to see how the accuracy of tests has improved over just one month.

So the correct statement is that low prevalence compounds existing inaccuracies in early tests, which are now apparently being supplanted by better ones.

Please, please don't discourage people from getting antibody tests—the aggregate data is extremely valuable for public health purposes. Instead, encourage people to ask their doctors how to be certain they're getting the most accurate tests available. (Our allergist, who's an immunologist, wrote us prescriptions specifically for whole-blood antibody tests rather than the much less accurate finger-stick ones.) Or wait a couple of weeks and see whether the FDA takes a bunch of tests off the market because the manufacturers can't demonstrate their accuracy. But don't let one simulator put you off the idea of antibody tests altogether, or give you the idea that those percentages are fixed and unchanging.

[sonia]

Thanks for the additional info. Great to hear there are more accurate tests coming out.

PS: Please don't tell me what to encourage/discourage in my own journal.

[rosefox]

Fair enough. I'm sorry for overstepping.

[sonia]

Thanks.