Thread by @statesdj

Epidemiologists and economists say we need millions if not tens of millions of tests/day to convince people it's safe to re-open the economy, e.g. this Harvard study estimates we'll need 20 millions tests per day ethics.harvard.edu/Covid-Roadmap
1/

Nobel Laurette Paul Romer’s @paulmromer simulations say we need to test everyone in the US every two weeks, or about 25 million tests per day paulromer.net/covid-sim-part…
2/

The most sensitive and specific tests for COVID-19 use quantitative RTPCR. We have about 1,000 instruments each able to run about 1,000 RTPCR tests per day for a national capacity of about a million tests per day.
3/

And @ashishkjha finds that testing capacity is not distributed where it is needed leading the inadequate capacity in many areas
4/
npr.org/sections/healt…

So why do we have Governor Cuomo saying “Come on in, we have plenty of testing available.”
5/
governor.ny.gov/news/amid-ongo…

And a New York Times article saying labs are having trouble finding takers for their capacity.
6/
nytimes.com/2020/05/21/hea…

There’s another bottleneck that we need to address. All of the current tests are approved for use with a physician prescription and only for testing of symptoms. We have home pregnancy tests and home HIV tests. Why not home SARS-CoV-2 tests?
7/

To understand why the FDA restricts we need Bayes theorem. The US has a population of 335 million and about 1.6 million confirmed cases (0.5%). If a test has a false positive rate of 1%, that means 2/3rds of the positives will be false positives
8/

While the technical (i.e. ideal lab conditions) false positive rate for RTPCR is close to zero, real world assessments range as high as 8% with a conservative FPR estimate of 0.8% overall
9/
medrxiv.org/content/10.110…

And false negative rates for RTPCR range from 0 to 40%, a significant part of which may be due to inadequate sampling from deep nasal swabs that a difficult and painful to perform well
10/
medrxiv.org/content/medrxi…

If a physician feels that you have symptoms consistent with COVID-19, the prior probability that you have the virus goes way up, and the 0.8% false positive rate becomes much less of a concern. A called positive is likely to be a true positive
11/

The FDA isn’t being obstructionist, they just want the results of the test to be meaningful true positives not confusing false positives. To get there on direct to consumer testing, we need real world false positive rates in the 1 per thousand range
12/

So how do we achieve real world false positive rates of one per thousand? Big centers that do lots of testing so they can do really exhaustive QA/QC testing. If you’re only running a few hundred tests a day, it’s hard to find per thousand errors
13/

There are many reasons we want to seriously ramp up testing. Employers want to keep their job sites safe. Ideally, that means testing every work every few days. To do that we’ll need 50 million test per day
14/

Some of the reluctance to increasing testing may be a sense that our case trackers are overwhelmed already, why add to their work? But testing more means we find new cases earlier before they have lots of contacts
15/

And if we really ramp up testing and systematically test everyone, that means we start to pick up presymptomatic cases which account for 80% of transmission and really cut down on the tracking workload
16/
science.sciencemag.org/content/368/64…

In one context we may be overusing RTPCR. RTPCR is extraordinarily sensitive and can detect a few molecules of viral RNA in a sample. Many patients shed fragments of viral RNA long after they have recovered
17/
ncbi.nlm.nih.gov/pmc/articles/P…

The Korean CDC examined 769 contacts of 265 cases who were initially RTPCR negative and later had a positive RTPCR test. No new infections were detected.
18/
cdc.go.kr/board/board.es…

If people are not contagious after they have clinically recovered and have developed good titer IgG antibodies, they can be safely released from the hospital or quarantine. This would avoid tying up people and resources waiting weeks for viral fragments to clear
19/

So let's get with the program and get serious about ramping up truly massive scale testing so we can catch cases early, making tracking easier, make job sites safe and get the economy back on track
20/fin