Thread by @DiseaseEcology

How much faster can we detect Omicron by sequencing more cases?

As many countries try to determine if Omicron is present, one simple approach is to sequence more. How much time does this buy us? A simple calc puts things in perspective.
tl;dr sequencing 5x more buys ~13-40d

The math:
The probability of detecting something by sequencing N samples given it is present at a prevalence P is simply:
1-(1-P)^N
Graphically, this relationship for two lowish prevalence values is:

So, the more the better, initially ~linearly when variants are rare, but w/ diminishing returns as prevalence increases & as N samples sequenced increases.

If we sequence more, how much more quickly can we detect Omicron, given that it is present and growing at a rate Rt over a serial interval of ~5.5 days? If we sequence X-fold more & variant is rare, this will allow us to have approx same detection prob w/ X-fold lower prevalence.

Since variant is growing at Rt, we can calculate roughly how much faster, on average we'd detect Omicron, by specifying Rt, and the increase in sequencing effort X, and the serial interval (time b/w generations of cases).
Advance warning = 5.5*log(X)/log(Rt)

We don't yet know what Rt will be in each place (it will depend on vaccination & past infection and contact rates/behavior), so let's use 3 values, 1.25, 1.5, and 2. For these values, and an arbitrary X=5-fold increase in sequencing, we buy 40, 22 & 13 days advance warning.

That's great, but what matters is what we do with this time. Same argument is true for stopping travel from potential sources. It makes little sense to restrict travel if we aren't going to do anything different. It buys a little bit of time, that's all (& possibly very little).

A few possible options of what to do:

https://twitter.com/DiseaseEcology/status/1464692441471078402

If we get excited about searching for Omicron but don't do anything to stop it from establishing then all we'll do is have a more precise description of it's invasion & impact.