I’ve been cautious about saying much about omicron but I think it’s getting into the space where I can add something. And what I’d like to say is: despite the encouraging signs from South Africa that the omicron wave is milder than those that came before, I’m still worried. 🧵
The reason for this is essentially our old favourite: multiplying a very large number by a small percentage can still give you a pretty large number. There’s a view that omicron won’t be a big problem because severe disease protection will hold up well,
...and the vast majority of reinfections and breakthroughs will be mild. But I’ve run a few numbers, and *even with that assumption of maintained severe disease protection*, a wave of hospitalisations with a peak similar to wave 2 (in Jan 2021) is quite possible.
I should be clear, I’m not using my forecasting model for this – it’s currently broken, and at some point I’ll give it a decent burial and a review of its highs and lows over the last year. All I’m doing here is multiplying a few numbers together. And I’d like to emphasise:
- This is all highly uncertain
- There is no guarantee of anything
- The reality could be a lot worse, or a lot better, than I’m suggesting here
- This is not a forecast, just a scenario from a few assumptions with very wide error bars.
So, with those huge caveats ringing in our ears, let’s suspend disbelief and dive in. Firstly, let’s assume that omicron is showing a growth advantage of about 4x in South Africa – I think that is consistent with the most recent views e.g. here
My guess is that this is mostly likely composed of a 3x advantage derived from a higher reinfection/breakthrough rate, and either a small increase in intrinsic transmissibility (R0) or a shorter generation period. I’m not an expert on this, and others will have better guesses...
…than me. But my guess (shown by the red cross imposed on this picture, originally from @trvrb) is at the optimistic end. If omicron has a higher level of immune escape, and reduced transmissibility, that will turn out much worse for the UK and other high-immunity countries.
Now, just because omicron has an advantage of 4x in SA doesn’t mean it will do here. There are reasons to think that vaccine-based immunity vs. infection might hold up better than prior-infection-based immunity is doing, particularly for those with 3 doses or 2 doses+ infection.
So let’s suppose that, rather than a 3x increase in reinfection/breakthrough rates, in the UK we only get a 2x increase. This is consistent with @andrew_lilico ‘s assumptions here: which I think are reasonable (except for his view on omicron's R0).
If this was the case, we’d expect to see omicron with a growth advantage of around 2.5x in the UK, and a doubling time of around 3 days. This appears consistent with the latest SGTF data in the UK – although there is much potential for distortion in this.
If I now make the optimistic assumption that omicron’s generation time is reduced a bit (e.g. to 3 days from the ~4 days that delta seemed to be working on), then that would correspond to a Rt for omicron in the UK of about 2. Unmitigated, that would give us a sizeable wave.
But how many cases and hospitalisations would there be? Don’t we need the (broken) model to tell us that? Fortunately, no. We can work it out from the immunity calculations, with a bit of effort. My model and @andrew_lilico’s were in close agreement that the current...
…effective level of population immunity in the UK (vs delta) is a shade under 80%. If this drops to around 60% (as assumed above, causing the jump in Rt from about 1 to about 2), then we need a lot more infections to get back to a herd-immunity threshold where Rt = 1.
Assuming (as above) that omicron’s R0 is the same as delta’s, we need to get back to about 80% overall immunity. (and if you’re thinking, hang on, wasn’t delta’s R0 around 6.5 so that should be 85% not 80% shouldn’t it? Yes, but we’re not quite at normal behaviour in the UK...
…and we’re still isolating people with symptoms, which probably brings R down by about 25%, so that’s why the HIT is a bit lower). But an unmitigated wave which starts at 60% immunity and reaches HIT at ~80% immunity isn’t going to stop there – it will overshoot.
So in total I’d estimate (there is a bit of toy modelling in the background here) a bit over 30% of the population would end up being infected in an unmitigated omicron wave. That’s a lot of people i.e. about 20 million of the 67 million in the UK.
And it could be worse: my toy model uses an “all or nothing” immunity model. With a “leaky” immunity model (see thread below) you’d need a lot more people getting infected, because they don’t start at 0% immunity or get to 100% immunity after infection.
So my total number of infections could be underestimated by a factor of 2 or more. But 20 million is quite scary enough thank you, so I’m going to optimistically stick with that number for now. Now take a breath, because here comes the difficult bit: modelling severity.
What we want to know is the Infection Hospitalisation Rate (IHR) i.e. what % of people who are infected by omicron need hospital treatment. We can see that the Case Hospitalisation Rate (CHR) for delta in the UK is currently around 1.6%.
And assuming that we’re catching around half of infections as cases, that implies the IHR for delta is around 0.8%. What could it be for omicron? I’ve looked at this a few different ways, and my best guess is that it reduces by a further factor of 2, to around 0.4%.
Note that since we’ve assumed the number of susceptibles was doubled for omicron, this is basically assuming that those extra new susceptibles (people being reinfected or vaccine break-throughs) are all guaranteed to be mild cases, not needing hospital treatment.
They still contribute to the wave of infections, and so create higher risks of *other people* needing hospital treatment, but they’re not going near hospital themselves. This sounds a bit optimistic, because not all reinfections or breakthroughs are mild.
But on the other hand I haven’t allowed any benefit from additional boosters that aren’t already in the stats today, which is a bit pessimistic. So let’s assume for now that those missing effects net out, and move onwards (we’ll come back to this assumption later).
So, with a new omicron IHR of 0.4%, and 20 million infections, we can work out that there would be about 80,000 hospitalisations in the omicron wave. Is that good or bad news? Well it certainly doesn’t sound great for those 80,000 people. But on the other hand, we’ve had over
…600,000 covid hospitalisations in the UK so far, so it’s only another 15% addition to that total. And if you spread it out evenly over (say) 3 months, it wouldn’t be a much higher run rate of admissions than we have today. Phew. Panic over. Except….
…with a starting Rt of 2, the omicron wave wouldn’t be spreading itself evenly over the next 3 months. Instead you’re going to get a sharply defined peak, similar to epidemic waves that we saw much earlier in the pandemic. Again using my toy model, I can estimate that…
… nearly a third of the total cases would come in a single week, at the peak of the wave. (I’m not trying to predict when, but if I look at @BristOliver’s graph below, mid-January looks like a reasonable bet).
Of course, that means that a third of the hospitalisations would also come in a single week, so that peak week would have around 25-30k hospitalisations in it. Add in a bit of continuing strain from delta, and you’ve got roughly the same peak (30k) as wave 2 in January 2021.
Now, there will be some people who think: ok, that’s not brilliant, but we’ve survived it before and we’ll do so again. And maybe they’re right. And I’m certainly not rushing to re-introduce restrictions that would devastate people’s liberty, livelihoods, and mental health.
But we know the NHS is already in a perilous state, and I do think it’s maybe worth a debate as to whether we want to let a wave of that size happen again. The “good” news (in so far as anything in this is good) is that if 80k hospitalisations were the size of the problem,
…and if the UK omicron wave started with an Rt of 2, that’s in the range where NPIs some way short of a full lockdown might be enough to flatten the wave sufficiently to make things more manageable. Still not good, but maybe better than a fast wave that stretches NHS capacity.
Others on the other side of the argument will make the case for full suppression – at least until we’ve completed the booster programme, and perhaps even until we have better, omicron-specific vaccines. I’m not convinced we can or should wait for the latter,…
…but slowing things down to get more boosters in arms could make some sense. Ideally, we need more data on the effect of boosters vs. omicron, and on the likely size and shape of the unmitigated omicron wave, in order to inform those debates and judgements.
Which brings me back to what I said earlier (see below). This is all highly speculative at this stage. My numbers could easily be out by a factor of 5 IN EITHER DIRECTION. On the downside, any of the following things could make things worse:
- Omicron having more immunity escape, and lower transmissibility (for the same overall growth advantage)
- Omicron having the same generation time as delta, and so slightly higher transmissibility (for the same immune escape and growth advantage)
- A “leaky” immunity model which would require more people to get infected
- A less optimistic severity assumption for omicron, which allows for some extra reinfections / breakthroughs to need hospital treatment.
What could make things better?
- If the early estimates of growth advantage for omicron in the UK are over-estimated, and vaccine immunity holds up even more strongly than we’ve assumed here, OR
- If we get a lot more boosters into arms in the next month, before the omicron wave really gets going. Those boosters will reduce the total number of infections (by increasing VE vs. infection) and also reduce the severity of some of the breakthrough infections that occur.
So I’m not saying that disaster is certain, or unavoidable. In fact I think we might well have levers we could pull to make it less horrible – but those levers have pain and costs attached, which will make the political debates even more intense than a disaster we can’t avoid.
And it may turn out that things aren’t that bad. I sincerely hope I look back at this thread in a week or two’s time and laugh at how unreasonably pessimistic it seems, once we got better data. But I worry that it could just as easily be the other way.
And with that range of uncertainty, I don’t think speculation is pointless or irresponsible. In fact I think it’s highly responsible: we need to understand the range of scenarios we might face. I’ve given you one scenario: I’d welcome views and suggestions on others. /end
PS in response to comments, I would like to add two more potential upsides:
- the impact of antivirals (although I expect this to be marginal)
- the potential for omicron hospital stays to be shorter on average than for delta, so relieving some of the pressure on the NHS.
PPS I realise that I also forgot to mention the possibility of omicron being intrinsically less severe or more severe than delta - so this needs to go in both as a possible upside and as a possible downside

• • •

Missing some Tweet in this thread? You can try to force a refresh

Keep Current with James Ward

James Ward Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!


Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @JamesWard73

30 Nov
it's a while since I've given you the age breakdown of English cases (sorry, it's been a busy couple of weeks at work and at home). and it's looking OK I think: overall cases are pretty flat, maybe even slightly falling. and the falls continue where it matters most (in 60+) 🧵 Image
digging into the detail, starting with the younger age groups: these now all look fairly stable, maybe just slightly over the peak. (and note I'm being cautious by truncating 3 days on the specimen date series - the next day looks slightly better again, on a sneak peek). Image
there's still a little bit of growth in the 20-40s, mostly from the 30-40s: Image
Read 9 tweets
18 Nov
only three graphs today: one bad, and two good (which is probably a fair summary of how I'm feeling). firstly the bad: cases in 5-9s appear to have hit an all-time high, which isn't ideal: 1/4 Image
the better news is that growth seems to be fading in the younger age groups (but with cases still growing, for now at least), and cases are already falling in the older (60-80 and 80+) age groups: 2/4 Image
even better to see this feeding through into hospitalisations: we can see the orange line (65-84s) clearly trending down. I'd like the blue line (85+) to have a steeper downward trend over the next week 🤞: 3/4 Image
Read 4 tweets
13 Nov
Honestly not a big fan of the case trends in the last couple of days. Here’s the overview: with the kids still driving the growth, but other groups also starting to pick up. (and I wouldn’t take too much cope from the flattening on the last day – that will get revised up) 🧵
Looking in more detail at the under-20s, we can see the power coming from the 5-9s, but with 0-4 and 10-14 not far behind, and even 15-19s now back into positive territory (i.e. cases growing)
In the 20-40s, cases now also (gently) growing:
Read 9 tweets
11 Nov
The headlines from today’s case data are concerning, and it’s certainly not great news. But there are some glimmers of hope in the detail by age, so let’s have a dig in. The overview doesn’t look too bad, with case rates stable in the under-40s, and still falling in the 40+ 🧵 Image
But that is a slightly lagged view of what’s going on, and obscures some useful detail. So I’ve built some new graphs which break down each of those lines into 5-year age groups, with no averaging, and using data up to specimen date 9th Nov – NB this date will get revised up.
Here’s the under-20s, to start with – we can see growth re-emerging in the 5-9s and 10-14s, but not so much in the 15-19s. The 0-4s were already bouncing around, close to stability – and that looks to be continuing here. Image
Read 12 tweets
3 Nov
The case trends remain hard to read, but with a bit of age-group case ratio analysis, I think we can see a bit further into the fog. Let’s start with the overview: cases still falling in the under-20s and over-60s, but not the 20-60s. But that hides some important detail. 🧵 Image
Looking first at the older (60+) age groups., we can see some very steep (and welcome) declines in the 75+, a gentle decline in the 70-74s, and gentle increases in the 60-69s. Exactly what you might expect if we were giving boosters to the most elderly first, for example. Image
So – good news overall. But also a reminder that if we leave the 60-69s unboosted for too long, we might see those case rates drift upwards, which would be very unhelpful.
Read 15 tweets
30 Oct
There been something bugging me for a while about how my model (and others) works. And I think I’ve finally pinned it down. It’s subtle and technical, but I think might turn out to be surprisingly important. And it’s all to do with the “leakiness” of the immunity model. 🧵
The difference between a “leaky” vaccine and an “all-or-nothing” vaccine has been well-covered by others – my favourite explanation is in this thread:
As it happens I’m not particularly happy with the “leaky” terminology – it could mean a number of different things, and so could be misinterpreted. I prefer to think about this in terms of the variability of the vaccine effectiveness (VE) at individual level.
Read 25 tweets

Did Thread Reader help you today?

Support us! We are indie developers!

This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal

Or Donate anonymously using crypto!


0xfe58350B80634f60Fa6Dc149a72b4DFbc17D341E copy


3ATGMxNzCUFzxpMCHL5sWSt4DVtS8UqXpi copy

Thank you for your support!

Follow Us on Twitter!