Thread by @MichaelSFuhrer

It's early days, but it seems that something quite unexpected is happening here. Many restrictions have been lifted in Vic and NSW, and by and large nothing has happened.

1/🧵

https://twitter.com/MichaelSFuhrer/status/1455671311053688838

By "nothing", I mean that (apart from a bump in cases in NSW largely outside greater Sydney), the trend of the effective reproductive number (R_eff) vs. vaccination has followed the same track as during restrictions.

2/🧵

The sloping lines on this plot are the expectations for vaccination effect on R_eff. Their y-intercepts are the R_eff expected with zero vaccination, which should vary with the level of public health and social measures (PHSMs).

3/🧵

The x-intercept is the inverse of the vaccine effectiveness against transmission (VET); a VET of 90% gives x-intercept of 111%.

4/🧵

The different solid lines are expectations from the Doherty Institute modeling for re-opening. "High" PHSM is roughly Stage 4 lockdown, "Medium" is Stage 3, and "low" is roughly Stage 2 (non-lockdown).

5/🧵

It seems Vic and NSW are meeting expectations for "high" PHSM and partial test/trace/isolate/quarantine (TTIQ; "partial" expected for high cases), but without lockdown.

6/🧵

Many possible reasons for this, all interesting:

1) There's something wrong with the data. I can't rule this out. One possible factor which is skewing the data is that cases in vaxxed are more likely to be asymptomatic and not reported.

7/🧵

This is a linear effect on cases, and R_eff measures the exponential slope of cases vs. time. However, when the exponential slope is small (near R_eff = 1) corrections could matter a lot.

8/🧵

But, if (1) is the *entire* reason we see an apparent R_eff<1 when it is really >1, that means cases are growing (exponentially!) and we don't know. Seems unlikely, especially as vax is saturating and the ratio of asymptomatic/symptomatic stops changing.

9/🧵

2) Lockdowns are simply *much* less effective than thought.

This seems also seems unlikely; it is counter to the experience of Vic 2020 and many jurisdictions.

10/🧵

3) Restrictions can be removed on vaxxed with little consequence. (I.e. lockdowns are effective for unvaxxed, but don't matter for vaxxed.)

Surprising if true. VET is accounted for in the plot, and isn't perfect; why does removing restrictions on vaxxed not matter?

11/🧵

4) PHSMs and vaccination aren't independent multiplicative factors. (Similar to #3.)

Again, it's hard to understand exactly how this works. We think we roughly trust the endpoints R_eff(0 vax) and VET, implying that something odd happens in the middle.

12/🧵

It's a riddle.
Covid is full of surprises.
The Doherty Institute model predicted well (particularly for NSW) the point at which vaccination would bring R_eff = 1 in lockdown. How can it predict so poorly the effect of lifting restrictions?

13/13🧵