Read on Twitter
Remember the preprint that suggested that #covid19 herd immunity thresholds (HITs) were around 6-20% across Europe, and was part of the evidence used to justify the Great Barrington Declaration? @meyerslab re-evaluated the work over the past few months - a thread...
tldr: when we alter what we believe to be flawed assumptions in the model, estimated HITs increase anywhere from 3 to 10-fold across all countries, and now fall well in-line with published estimates and expert consensus for HITs.
The authors of the preprint attempted to identify the cause of the Summer pandemic slowdown across 4 countries (Belgium, England, Portugal, and Spain) by creating a model to tease out the relative impact of heterogeneity and community interventions https://www.medrxiv.org/content/10.1101/2020.07.23.20160762v3
Heterogeneity can lead to epidemic slowdown, because heterogeneous populations (depending on the structure) can hit herd immunity thresholds much earlier than the naive 1-1/R0 assumption suggests. See @meyerslab pub ( https://doi.org/10.1098/rspb.2006.3636) or https://www.quantamagazine.org/the-tricky-math-of-covid-19-herd-immunity-20200630/
To carry out the estimation, the group made strong assumptions about the shape of community mitigation (based on mobility data at the time). Mitigation reached a maximum by early April and is completely removed by September: (Values of 0 mean no community mitigation measures).
However, mobility doesn't capture all mitigation, as we know people can take other precautions like social distance and mask wearing. Here I'm plotting the government response index from Oxford that summarizes 18 policies to reduce community transmission https://www.bsg.ox.ac.uk/research/research-projects/coronavirus-government-response-tracker
We decided to fit a mitigation curve to these policies in an attempt to create a more data-driven mitigation scenario that is at least if not more plausible than the original assumed curve
We then ran their model and estimated the HIT under both scenarios to get new HIT estimates and also estimate mortality if a 'herd immunity' policy was followed (we included the US here as well). You can see the large impact of the different mitigation shape...
Most importantly, we saw that these HIT estimates were highly sensitive to assumed mitigation curve shape in general, and there are a wide-range of mitigation shapes that yield high HITs
This is important, bcuz we don't think our herd immunity estimates are exactly correct... In fact we show that you can only estimate the combined impact of herd immunity and mitigation, so it's statistically unidentifiable (and why you need data-driven mitigation curves).
We demonstrate this by doing the inverse estimation: directly estimating mitigation curves for assumed herd immunity thresholds:
Overall, we believe an abundance of evidence is pointing to higher HITs of 50-80%, based on current pandemic surges around the world, high attack rates in congregate living situations, and our re-evaluation of this unpublished work. and the preprint link: https://www.medrxiv.org/content/10.1101/2020.12.01.20242289v1
This is not to say that heterogeneity in transmission isn't important for understanding spread, I think it's been well documented that it is -- but it is to say that the trends can be accurately explained without invoking the original high levels of heterogeneity
s/o to @meyerslab, @mikha_ehl, @rvsrnvsn, and our extremely talented undergrad Pratyush Potu for all of their hard work to put this together!
