I, & many others, have been looking for quasi-experimental evidence in the U.S. on the contribution of schools to COVID spread

The @EPICedpolicy team & coauthors have moved incredibly fast to share that evidence, based on data from Michigan & Washington https://twitter.com/EPICedpolicy/status/1341758767621435399?s=20

TL;DR: In-person schooling in low and medium community spread areas does not contribute to community spread; in-person schooling in high community spread areas does contribute to increased community spread

Also, wow, be sure to read their intro and background sections for an amazing overview of what we know, and what we don’t know, about COVID and schools, alongside their findings!

The data is not perfect -- in a perfect world, they’d have exact opening/closing dates, teacher, student, and family specific COVID outcomes, no selection into who gets tested, the ability to do an event study -- but no single piece of evidence ever is

Other findings to highlight:

1) Without accounting for PREEXISTING community spread, it appears like in-person school contributes to community spread: this is the correlation NOT causation thing people keep talking about

Big caution for eyeballing exercises

2) Mitigation efforts are important

Reduced student density makes a difference, especially at highest community case levels

Unfortunately no data on other mitigation efforts

I’d be most interested in mask-wearing -- @ProfEmilyOster’s Dashboard shows big differences by masking: https://statsiq.co1.qualtrics.com/public-dashboard/v0/dashboard/5f78e5d4de521a001036f78e#/dashboard/5f78e5d4de521a001036f78e?pageId=Page_ffb4dc52-5543-46b2-8126-2b7229fd1b17

3) A bounding exercise implies that, if anything, these estimates are biased upward and that the actual contribution to spread by schools is lower

Some important caveats:

1) Data is from this fall (November and prior) before the recent rampant rise in COVID

2) Hard to define “high” spread (seems like it occurs at different points in MI and WA)

3) MI & WA are not representative of the nation

Most importantly, we want evidence from the low-income & minoritized communities hit hardest by the pandemic

But in MI/WA, only a small # of counties have high shares of minority pop'n & thus authors can’t show direct evidence

Notably, we have some previous evidence, mostly from Germany and elsewhere in Europe that generally came to similar conclusions, but impt to see direct U.S. evidence given the different mitigation efforts, pandemic response, and social safety net policies

I know @KatharineStrunk and coauthors are looking feedback and are interested in pushing this analysis further, updating with new data as it becomes available, and investigating additional angles

Stay tuned for more from @douglasharris99 @REACHcentered who will soon release nationwide findings using similar methods but with COVID hospitalizations as an outcome https://twitter.com/douglasharris99/status/1340054884079558657?s=20

Final take-homes:

1) Community rates matter: AS ALWAYS, the best strategy for reopening in-person schooling is to mitigate spread -- that means consistent federal-led pandemic response (and appropriate funding for business closures as needed)

2) Mitigation works

3) Ultimately, the choice to reopen in-person schooling is not about the evidence on spread, but how policymakers and the public weigh the tradeoffs and whether they can manage the logistics

4) In order for there to be more widespread reopening (while acknowledging that much of the country already in-person!), teachers & families, particularly families of color, must trust the safety of schools...

4 cont) ... be it through mitigation, lowered community spread, appropriate funding, better communication from leaders, or vaccine prevalence (or all of the above)