A @CDCgov report in Jan concluded that "university counties with in-person instruction experienced a 56% increase in incidence".
They only examined 21 days before/after classes start.
Since then, those counties saw a much lower incidence vs counties w/remote instruction.
🧵
You can see above that counties with in-person instruction had a ~50% higher incidence in the weeks after classes start than remote counties, consistent with the CDC report.
But during the peak in Dec/Jan, counties w/in-person instruction actually had a ~50% *lower* incidence.
Contrary to what many believe, remote instruction did not decrease county-level incidence during the fall surge, when compared to in-person instruction.
Below is a breakdown based on total cases per capita.
Below is the CDC report with their original results. It was published in January but only uses data up until September.
Due to possible geographical bias, the CDC added a "matched analysis" where in-person university counties were matched with non-university counties based on geographic proximity & population.
Here, in-person counties also saw a slightly higher initial incidence that tapers off.
There are a lot of data quality issues with the original unmatched analysis, which is why I believe the matched analysis is the best method to view these results.
For ex, the unmatched analysis had 70% of remote learning counties located in a single state, skewing the results.
The CDC did not do publish their analysis on deaths, so we will take a look.
In both the matched and unmatched analysis, university counties with in-person instruction had a lower peak than university counties with remote instruction or non-university counties.
If you look at total deaths, the pattern also holds. Counties with in-person instruction currently have the lowest total death rate.
Between deaths and cases, deaths is probably the more important metric, since it's more correlated with the strain on our healthcare systems.
Spikes in cases, especially those concentrated on young adults on college campus bubbles, should not be an automatic cause for concern.
What happened here is similar to the CDC study on mask mandates. Short-term correlations do not necessarily lead to long-term causal relationships.
They should not be used to set/influence policy, especially one as important as education.
Unfortunately, I've seen too many people cite data and sources that support their personal beliefs on school openings, and ignore data and sources that counter their beliefs.
Whether this is done intentionally or not, it is a dangerous precedent.
Of course, this analysis is difficult to repeat for primary and secondary schools.
But given what we know about transmissibility by age, the data is promising that in-person schooling, when done correctly, is not a primary driver of community transmission.
I want to thank the CDC for graciously providing me the county groupings needed to replicate and extend this analysis.
I really appreciate their commitment to open data sharing.
To conclude, the data does not support the claim that in-person instruction at colleges and universities lead to increased COVID-19 incidence and deaths in the long run.
Existing debates about schools seem to be more driven by politics than science, unfortunately.
Update: Some individuals have responded that most schools were closed after Thanksgiving, which is valid. But that still doesn't explain why counties with remote instruction had a higher wave.
When can we return to normal? Forget about "herd immunity".
Below is my estimate for the number of susceptible individuals over time, as a proportion of the US population.
Looking at this graph, what is the best point to go back to normal? Christmas? Fall? Or Summer?
🧵
By summer, everybody who wants a vaccine will be able to get one. The vulnerable population will long have been able to receive their shots. Hospitalizations & deaths will be at negligible levels.
Normality will happen... with or without herd immunity.
Our country currently has no concrete guidelines for when to expect a return to normal. We seem to be more concerned about a theoretical threshold than setting realistic goals about when restrictions can be dropped.
Rather than clearly highlighting this bold assumption in the article, it's buried at the bottom as a footnote.
At what point do unrealistic model assumptions become misleading? Does the science support immunity lasting indefinitely? Or 80% of Americans getting fully vaccinated?
If the model used more realistic assumptions based on real-world data, they would probably conclude that we won't reach herd immunity. But that would likely break the premise of the model, so I can see why they chose not to explore this.
I've seen many news articles cite that "the UK variant could be the dominant strain by March". This is emphasized by @CDCDirector.
While this will likely to be the case, this should not be an automatic cause for concern. Cases could still remain contained.
Here's how: 🧵
One of @CDCgov's own models has tracked the true decline in cases quite accurately thus far.
Their projection shows that the B.1.1.7 variant will become the dominant variant in March. But interestingly... there's no fourth wave. Cases simply level out:
This conclusion is based on the following new developments over the past month:
- Remained high levels of vaccine hesitancy
- New variants that may lower vaccine efficacy
- Rollout of the J&J vaccine (efficacy ~70%)
- Delayed arrival of the children vaccine
That said, herd immunity does not have a hard threshold, and being close to herd immunity may be sufficient to prevent large outbreaks.
Our goal should not be to reach "herd immunity", but to reduce COVID-19 deaths & hospitalizations so that life can return to normal.