Thread by @JamesWard73

It’s been a while since we’ve had a set of SPI-M models and it’s taken me a few days to read through them properly. Having done so, I have a few thoughts – as usual I’ve found 10 key messages to take away. For reference, the papers are all here: gov.uk/government/col… 🧵

In terms of overall messages, I would agree with everything that @BristOliver said here:

https://twitter.com/BristOliver/status/1451835747607142407?s=20

. So I’ll focus my own comments more in the detail of how the modelling has been done, and its implications.

1) In general, the models are fairly optimistic about the course of the next 2 months, but after that they vary a lot – with some scenarios showing significant peaks in 2022 (at any time from January through to June) and others with no big resurgence

2) Still, unless boosters wane and there is a fast reversion to pre-pandemic contact levels (or a new, immunity-escaping variant), it is hard to get the models to project hospital admissions above 2k per day in England.

3) We can also see from the papers that the most impactful uncertainties are around behaviour (mixing) and waning of immunity – including waning of boosters. Seasonality plays a smaller, but still significant, role.

4) As before, the models have slightly different approaches to modelling mixing and contacts. With a bit of detective work, it appears that Warwick are assuming the risk level is currently around 60% of pre-pandemic normal, Imperial are at 70%, and LSHTM at 80%.

This explains, in part, why the LSHTM model is more optimistic than Warwick’s – because a reversion to pre-pandemic contact levels has less potential to increase transmission. (Note my model is around 70% on this measure, so in the middle of the SPI-M range)

5) There are also different approaches to waning. Imperial use an exponential model, similar to mine. But they appear to have a higher rate of waning for vaccine immunity (around 20% per month if I’ve done my sums right), and a lower rate for natural immunity (10-20% pa).

Meanwhile LSHTM and Warwick have different waning models, where effectiveness is assumed to decrease to a set level (which is different in different scenarios). Warwick do not include any waning of natural immunity, and LSHTM have it waning at 15% per year.

6) For boosters, Imperial assume no waning; Warwick run two scenarios, one where boosters wane similarly to 2nd doses, and one where there is no waning. I’m not 100% clear on LSHTM but I *think* they have no waning of booster protection.

7) For the booster rollout, both Imperial and Warwick use a Cabinet Office scenario showing 1.3m boosters being delivered per week. As @PaulMainwood said here

https://twitter.com/PaulMainwood/status/1451573992255627292?s=20

, this is nowhere near fast enough. (and we’re ahead of that level now – but still too slow).

8) LSHTM did a sensitivity analysis to the effect of extending boosters to under-50s. This shows much lower prevalence in 2022, with c. 20k fewer deaths. If that’s true, it will be hard for the government to resist public pressure to extend the booster rollout.

9) LSHTM also modelled the impact of Plan B. assets.publishing.service.gov.uk/government/upl… Of the 3 planned interventions, they estimate encouraging WFH would be the most impactful, and vaccine certification the least. In their model, Plan B is enough to squash an emerging peak in Spring 2022...

…but the “pendulum” effect is also visible – the peak is either flattened so that it extends for longer at a lower level (helpful for NHS capacity) or, with stronger suppression, the peak disappears but re-emerges at a similar level later in the year (not so helpful).

10) Finally, it’s interesting that SPI-M took a whole page of their report (p2) to set out the case for continuing the ONS survey through 2022, and to support daily testing of contacts of cases with LFDs as an alternative to quarantining contacts. /end assets.publishing.service.gov.uk/government/upl…

PS sorry, I meant to say: overall this is a highly impressive set of papers and I can find very little to disagree with. Big credit and thanks to the teams @GrahamMedley @DrLouiseDyson @EdMHill @mjtildesley @BarnardResearch @_nickdavies @lilithwhittles @erikmvolz & many others!