rt-from-frequency-dynamics VS CovidColoradoCharts

Although several XBB.1.5 descendants are growing marginally faster, all variants that aren't XBB+486P are on the decline nationwide. This includes CH.1.1, which has been the clear #2 VOC behind XBB.1 since ~October and which has gotten some press recently.

State or national variant numbers can currently be best accessed at cov-spectrum; in this link by clicking on "show regions". Covariants provides an even easier-to-read version, but it only separates a few classes of variants and in 2-week intervals. And Bedford labs variant tracker provides a unique perspective of different variant growth, but currently does not separate by states and has irregular updating.

With variant tracking the main place to follow is the pango-designation project on github (but please do not post there if you go). Cov-spectrum is great for following the epidemiology of variants, as well as bedford labs. And before that covariants, but they haven't kept up with cov-spectrum's development. All of it's very simple from a modelling perspective, but applying the gisaid data to the case data is massively enlightening.

Though they update even more infrequently, I really like the new Bedford labs graphs. These make it very clear just how complicated trying to model the next several months would be.

For a national perspective you can see Trevor Bedford’s GitHub. https://github.com/blab/rt-from-frequency-dynamics/tree/master/results/pango-countries

Bedford labs splits it up into specific case counts, and shows the same thing. Here we see that BA.5 cases just stopped declining and started rising. Unfortunately they update irregularly. I'm looking specifically at the third graph here, and what makes me not want to be certain is the dark gray dots that are rising rapidly at the right. These aren't on the legend but I believe mean "other".

Bedford labs this week has added several new graphs. This looks in detail at just the US at a lot of different lineages. This looks at many countries with fewer lineages. Note that they do not have XBB.1 (everything's grouped as XBB giving variable results) and their updates are very irregular.

4 Bedford labs variant tracker is following XBB now, but not XBB.1. This isn't very useful to see growth in many countries where XBB has been spreading for a bit and XBB.1 has just been introduced, though the r~0.19 (3.8-fold weekly growth) is probably pretty accurate for Singapore.

covariants, cov-spectrum, outbreak.info, blab

But my best source, when it's updated, is Bedford labs variant tracking project.

By onset and reported day, the case drop has stopped. It doesn't seem certain if we are in a plateau for the last few weeks, or if cases are rising, or if it's just a back-to-school plateau followed by further drops. It is not driven by new variants though; BA.4.6 is still <10% of nationwide infections and all the next-gen variants combined are <1%.

By reported and onset date, cases may have flattened. This wasn't visible in the reported-day numbers until today, but it's been showing in the onset-day numbers for a week or so. It could be variance, but the flattening of positivity at the same time makes it more likely to be real. If true though this isn't caused by a new variant; everything is still BA.5 for now.

By reported and onset date, cases are still falling.

By reported and onset date cases keep slowly dropping. Smoothed growth rate is down to -10-15% per week. Super approximate and I should probably make an actual model here, but if there were 1500 cases/day on July 20th and -2% per day, that puts us down to 1000-1100 new cases with onset day of today, which would be the lowest since the start of May.

Cases by reported and onset date.

By onset and reported date, cases seem to be curving up a bit, but with no rapid rise from BA.5 so far. The normal surge pattern is that the upward curving will continue until an inflection point, then it'll curve back down to and beyond the surge peak. But whether this happens soon, or if upward acceleration goes on for a while, or even if we're already past the inflection point, is likely impossible to know.

By reported or onset date, there's still no definitive case rise from BA.5.

By onset and reported date, cases are doing some weird up and down thing. It's probably a plateau and the variance just random, as BA.2.12.1 drops and BA.5 continues to take over. This likely won't last long; within a few weeks BA.5 will be the large majority of cases and for it to stay flat at R(t)~1 isn't unlikely.

By reported and onset date, cases have firmly dropped over the last week. This represents BA.2.12.1 peaking and subsiding.

There's several indications that BA.2.12.1 may have now peaked. This is based first on state case numbers by reported and onset date, which have flattened completely by onset date and have been declining slightly by reported date. And secondly, on Bedford labs' per-variant case estimates, which although they do have high error show BA.2.12.1 declining in the most recent sample collection days available.