Thread by @DiseaseEcology

Thx Marc for reply & 2nd look. Still several misunderstandings - I've tried to clarify below but doubt that will convince you.
My assessment after re-reading orig papers, your paper & seeing recent school outbreaks is v. different.
Much more cautious conclusions seem warranted.

https://twitter.com/mlipsitch/status/1287368183201124352

I strongly acknowledge importance of re-opening schools, especially for younger children. But inability to determine whether contact rates or susceptibility or data biases explain lower infection in young kids suggests a need for cautious conclusions. Yours are very strong:

Misunderstandings
-testing all in household does not remove bias b/c of timing of test. If child is asympt index & infects parent who is symptomatic & that leads to testing kid; test can miss fading virus. See same point made by Christian Drosten:

https://twitter.com/c_drosten/status/1274714443008573442

(cont)

B/c asymp infection more likely in kids than parents, this bias lowers measured infection prev of kids.
Testing all household members/contacts by swab does NOT address this bias. You'd need either stool or Ab test to pickup missed infections.

Size of this effect depends on how often adults are mis-IDd as index case which is obviously unknowable, but seems like it warrants caution in interpretation of relative infection rates especially w/ next issue:

-infection is combination of contact + susceptibility. Lower infection could be lower susc OR less contact (type, distance, duration). None of the studies address this & contact b/w ages well known to differ (fig); inc "shielding" w/in households
journals.plos.org/plosmedicine/a…

e.g. your ref (16) assumes same contact rate for b/w & among all ages (beta_aa) & thus ascribes any diff in infection to susc. but diffs in contact rates would give same pattern
medrxiv.org/content/10.110…

You explicitly consider diffs in contact rates (quote) & present strong support for them (2nd quote). But you discount them for 0-9 b/c of comparisons of 0-9 vs 15-19 & non-spouse. Why couldn't contacts of adults & 15-19 differ from adults & 0-10 & from adults->non-spouses?

Uncertainty could easily be addressed by more cautious wording: "data show differences in relative infection rates of kids & adults that could be due to contact rates or susceptibility".
How can one can rule out diffs in contact rates which will inc. a bunch w/ school re-opening?

-viral loads of kids - your text focuses on subset of data (Ref 24 Table 2B) that orig authors explicitly advise against b/c it is biased towards later period when loads decline. Data authors recommend (Table 2A) shows equal loads of kids & adults as orig authors clearly argue.

This makes it seem like you are selectively reporting data to support a conclusion rather than taking data at face value or reading details conveyed by authors of paper. Probably not your intent but it reads that way.

Discussion of recent outbreaks also gives impression of having prior belief that is robust to data. Primary school outbreaks are written off as due to crowding (that apparently overcomes huge proposed diff in suscept) while lack of transmission from tiny datasets are solid.

Total number of infected children in 3 papers cited supporting quote below is N=9. and in ref 35, w/ N=3, each child had 1 contact.
This seems like a small dataset to make strong conclusions.

Having primary schools open w/out kids playing very closely with each other seems at odds w/ most parents' experiences. Thus, suggesting that we can keep transmission low in primary schools by avoiding crowding seems unrealistic.

You (Marc) have much stronger voice in community & policy & if you feel confident in your conclusion:

"opening of primary schools may have limited effect on
SARS-CoV-2 transmission in the community"

then so be it.
I very respectfully differ & have tried to explain why.