Thread by @trvrb

The impact of Omicron on case counts in the US is now abundantly obvious with 885k reported cases yesterday alone. Figure using @CDCgov data and showing daily reported cases with 7-day smoothing on log scale. 1/12

We can partition state-level cases between Delta and Omicron using sequence data from @GISAID. This is made possible by 1.6 million genomes from the US from viruses sampled since Oct 15. This remarkable dataset is thanks to a large number of labs throughout the country. 2/12

Genomic data will necessarily be lagged by about 10 days, but we can use modeling framework from @marlinfiggins to project forward variant-specific frequencies and epidemic growth rates to the present. 3/12

https://twitter.com/trvrb/status/1471651831923150848

Doing so across 27 states with sufficient data gives the following picture for frequency of Omicron across states, where a large number of states are >99% Omicron at this point and lagging states are still >80% Omicron. 4/12

We can partition Delta vs Omicron cases using this approach to get the following picture where Omicron cases were steadily doubling in early Dec before surpassing Delta in mid to late Dec. Log plot makes it clear that doubling of Omicron slowed as prevalence increased. 5/12

This decline is clear from looking at epidemic growth rate r across states through time, where once Omicron epidemics "saturate", little r begins to decline. This has also coincided with a decline in Delta. 6/12

This decline in Delta is likely due to a combination in behavior changes decreasing contact rates compared to a month ago and direct interaction between variants with Omicron infections "blocking" potential Delta infections as they spread through the community. 7/12

With these "phase diagram" plots, I'm attempting to show how case load is related to epidemic growth rate across states. The trajectory of Omicron is shown as a red line with per-capita case count on the x axis and epidemic growth rate on the y axis. 8/12

Early in the epidemic growth rate is roughly constant and you get a line moving from left to right. But as epidemic increases in size and begins to saturate, epidemic growth rate declines and the trajectory curves downwards. 9/12

Apparent large decline in epidemic growth rate in California is likely due to holiday reporting issue, but gradual decline across most states is likely real. 10/12

So far, the overall trajectory of Omicron across states is remarkably consistent with similar initial epidemic growth rates. Some states are just more ahead and others more behind on the same curve based on initial seeding patterns near the beginning of Dec. 11/12

Model details can be found in the following preprint (medrxiv.org/content/10.110…) and we'll update results ~daily at github.com/blab/rt-from-f…. 12/12

Follow up #1: The "1.6 million genomes" shared to @GISAID since Oct 15 in the US referenced in tweet 2/12 is incorrect. There has been 1.6M genomes shared globally from samples collected since Oct 15, and of these 508k have been from the US.