Using this bit of information from @HannayRyan

https://twitter.com/HannahD15/status/1478135176920137729?s=20

, the COVID dashboard from NSW health: aci.health.nsw.gov.au/__data/assets/…,
and a clever suggestion from my cousin (Matt Lilley), we can derive the case ICU rate of omicron in NSW: between 0.17% to 0.19%. We have 5 pieces of info to do this:
*There were 26 people in ICU on Dec 16
*195 ICU admits from Dec 19 to Jan 3; ie ~208 from Dec 16 to Jan 3.
*74% of ICU patients from Dec 16 had Delta
*The case ICU rate of Delta was a steady 1.25% in late Nov/early Dec
*Daily case numbers

I get a lot of Qs re: how my estimates of the case hospitalization and ICU rates seem a lot worse than other graphs on Twitter, which show we won’t get anywhere near the October peak in hospitalization and ICU (which I think we do in ten days).
The two graphs you’ve probably seen The first one is comparing a “stock” (the number of people in hosp) with a “flow” (new daily cases). This is like expecting the water level in a large bucket to immediately double as soon you open the tap one more turn.

The most important question for capacity to handle the Omicron wave is the progression of cases from hospitalised to ICU and death.

In South Africa the progression from hospitalised to ICU and death for Omicron was 20%-25% of the rate of Delta.
Preprint: papers.ssrn.com/sol3/papers.cf… (This is the best table to use because using the first n cases limits the biases that are introduced by only using admissions that have a resolved outcome - different outcomes have different lags).

We still don't know what share of that reduction comes from disease v immunity.

The NSW case hospitalisation ratio (CHR) for Omicron stabilised at around 40% of the CHR for Delta (i.e. a -60% reduction in the CHR), controlling for indiv characteristics. At the population level, since Omicron cases show more selectivity to vaccinated persons it's -70% lower. The CHR has fallen from ~7.5% to ~3.2%, and it has now steadied there. Using the newest NSW weekly epidemiology report, we know that in the week of
Dec 4:66% of adult cases were fully vaccd
Dec 11:75% of adult cases were fully vaccd
and this composition chg lowered the CHR by 10%

Update on NSW case hospitalisation rate (CHR) for Omicron with three more days of data: basically the same. Many people asked for a (longer) distribution of hospitalisation times rather than using 5 days, and I have added that, but the final estimate is still around the same. The top panel of the graph here just uses a simple 5 day lag between cases and hospitalizations. The bottom panel assumes confirmed cases (ie +1day post-test) take another 4-9 days to turn into hospitalisations. I think that's probably a bit too long but it's by popular demand.

NSW is the ideal setting to measure the severity of Omicron, and it is showing a CFR for Omicron that is <1/2 of Delta
Five reasons it's ideal:
*Prior infection is irrelevant (~2-3% of NSW is prior infected)
*Delta cases were steady through Nov, making a stable baseline CHR

1/15 *Omicron became dominant extremely rapidly, so a change in the CHR will be noticeable
*Accurate case surveillance w/ age breakdown and <1 day lag in test to case count.
*Vaccination was largely complete by Omicron's arrival (at ~94% of adults), boosting was minimal.

2/15

I get people are hesitant to conclude omicron is less virulent by looking at hospitalisation/death rates from SA, since they have 2w+ lags, and hospital cases are still backfilling from 10d ago. But there's a great reason which doesn't suffer these issues.

https://twitter.com/andrewlilley_au/status/1467228063863762944?t=_xoHR2r-iTM1q9qh8LouCQ&s=19

The incidental hospitalisation rate is a way of inferring the true infection hospitalisation rate without needing to guess the total number of cases, or deal with long lags (it's subject to only a ~2day lag which is immaterial).

New favourite example for why synthetic matching is near impossible: CDC paper linking vaccination status with deaths not related to COVID finds a 2/3 reduction in other deaths. (Controls for age, sex, race, and whether you previously had flu shots as a proxy for health coverage) Of course, it's not that the vaccine reduces other deaths, it has to be selection, but the magnitudes are enormous. Most surprisingly, this is common across all age groups including 18-44, and all races. The only subgroup which shows a lower measure is teenagers.

When will NSW gain its 70% vaccinated freedoms?

This is a tough nut to crack, because you need to forecast a) the average time it takes between doses and b) the % of people who miss their second dose appointments.

My best guess is October 8.

1/6 (a) and (b) combined will determine the time it takes for each age group to go from a first dose % threshold to reach that same point in second doses.

If you try to do that from the population as a whole, it's a bit of a mess - it looks like it's coming down over time because

Update from Israel comparing all forms of protection including new data on boosters given and breakthrough infections (natural immunity from infection after vaccination).

Shows that natural immunity will almost surely be part of reaching herd immunity. The graph uses data from people 60+ only who were fully vaccinated as early as January through March. It shows the rel risk of infection vs being fully vaccinated in Jan (i.e. 10 is 10 times lower risk.) They don't provide a reference % protection here so I've taken the value for

This is definitely the best vaccine efficacy paper released:
*Randomly selected HHs tested on a predetermined schedule (no voluntary testing and no selection issue)
*Includes CT<30 only (high viral load)
*By days since vaccination, not month of infection
*Splits AZ/Pfizer x gap Firstly, as many suspected, adenovirus doesn't wane as quickly as mRNA (if it does at all). They're equal after 3 months and the authors suggest AZ will likely be better after 4 months.