NEJM editorial argues that low sensitivity tests are problematic
nejm.org/doi/full/10.10…
If the alternative is a perfect test, sure. But tests are still too scarce. We need more & faster tests; can sacrifice Se w/ clear communication
@michaelmina_lab @DanLarremore @RWalensky 🧵
nejm.org/doi/full/10.10…
If the alternative is a perfect test, sure. But tests are still too scarce. We need more & faster tests; can sacrifice Se w/ clear communication
@michaelmina_lab @DanLarremore @RWalensky 🧵
2/8
From the editorial: "A big concern has been test availability, but test accuracy may prove a larger long-term problem."
I disagree w this framing: the two features are in direct tension. If we hold out for "perfectly" sensitive tests, we resign ourselves to less testing.
From the editorial: "A big concern has been test availability, but test accuracy may prove a larger long-term problem."
I disagree w this framing: the two features are in direct tension. If we hold out for "perfectly" sensitive tests, we resign ourselves to less testing.
3/8
Great modeling studies from folks I tagged shows frequency/TAT are MUCH more impt than sensitivity:
medrxiv.org/content/10.110…
jamanetwork.com/journals/jaman…
Even w/o repeat tests, even one 70% Se test catches far more than no test at all. (& those it misses probably shed less virus)
Great modeling studies from folks I tagged shows frequency/TAT are MUCH more impt than sensitivity:
medrxiv.org/content/10.110…
jamanetwork.com/journals/jaman…
Even w/o repeat tests, even one 70% Se test catches far more than no test at all. (& those it misses probably shed less virus)
4/8
Others have argued this more eloquently than I. But the key to me is that w/ clear communication abt risks (which is hard!), tests don't have to be perfect to be helpful. (No test ever is.)
Slow, scarce testing remains a crisis. We need more and different solutions. Badly.
Others have argued this more eloquently than I. But the key to me is that w/ clear communication abt risks (which is hard!), tests don't have to be perfect to be helpful. (No test ever is.)
Slow, scarce testing remains a crisis. We need more and different solutions. Badly.
5/8
I disagree w/ second part of this: "Diagnostic testing will help in safely opening the country, but only if the tests are highly sensitive".
I think the framing in the article that leads to this conclusion is too much abt patient care use; for inpts, you want high Se & Sp...
I disagree w/ second part of this: "Diagnostic testing will help in safely opening the country, but only if the tests are highly sensitive".
I think the framing in the article that leads to this conclusion is too much abt patient care use; for inpts, you want high Se & Sp...
6/8
...but most spread is in outpatient setting; testing there is more to mitigate spread than diagnose illness. If we are clear that tests are imperfect and thus can't replace masking and distancing, a lot of imperfect testing can absolutely help reduce (NOT eliminate) spread.
...but most spread is in outpatient setting; testing there is more to mitigate spread than diagnose illness. If we are clear that tests are imperfect and thus can't replace masking and distancing, a lot of imperfect testing can absolutely help reduce (NOT eliminate) spread.
7/8
I do agree fully with this: "the FDA should ensure that manufacturers provide details of tests’ clinical sensitivity and specificity at the time of market authorization" as the key to not letting perfect be the enemy of good.
We'd have to get a lot better at SciComm, though.
I do agree fully with this: "the FDA should ensure that manufacturers provide details of tests’ clinical sensitivity and specificity at the time of market authorization" as the key to not letting perfect be the enemy of good.
We'd have to get a lot better at SciComm, though.
8/8
Could infinite perfect tests + perfect behavioral responses to them eliminate this and allow mostly normal life? No, b/c neither is possible.
We should find folks at highest risk of transmitting (likely those w/ high viral loads = less likely missed by low Se tests).
Could infinite perfect tests + perfect behavioral responses to them eliminate this and allow mostly normal life? No, b/c neither is possible.
We should find folks at highest risk of transmitting (likely those w/ high viral loads = less likely missed by low Se tests).
9/fin
(And we should mask and distance, even if we test negative. For so many reasons.)
(And we should mask and distance, even if we test negative. For so many reasons.)
