Thread by @JamesWard73

I’ve been promising to update my comparison of how the July model projections are going. So here it is. As you can see, the model (orange line) did quite well in predicting admissions for the first couple of weeks (in mid/late July), but missed the sudden decline in cases… 🧵

…that happened after the Euros, and so the actuals fell below the projection (good news). More recently, the gap has been closing, but that’s for the (bad) reason that hospitalisations have started to rise again, while my model was projecting a gentle decline.

A couple of notes on the other models shown here:
1. I think we may have done LSHTM a slight favour by selecting their non-waning model for comparison, when their (more conservative) waning model was probably more their real central case.

2. Conversely, we did Warwick a disservice by selecting the scenario that had an immediate step-up in risky behaviour. They also produced an array of scenarios with gradual increases in contacts, which I believe are included in @GrahamMedley’s charts.

https://twitter.com/GrahamMedley/status/1428762394919059461?s=20

(it’s definitely worth following Graham for those updates, by the way, and they’ll give you a better sense than my chart does of how the actuals are comparing to the full set of SPI-M model projections over time.

The actuals aren’t following any of the models particularly closely right now, but we don’t know what will happen in the next 2-3 months, and it may yet be that one of the Warwick scenarios with a later peak will turn out to look fairly accurate).

So, the big question is, what happens in September and October? I’m genuinely not sure, because I can point to several reasons why cases should increase e.g. schools and unis going back, cooler weather, waning immunity, and gradually relaxing behaviour (e.g. less WFH)…

…and also several reasons why cases should decrease e.g. more second doses coming into effect, plus vaccinations of 16-17yos (and maybe younger), as well as likely booster vaccinations for older/vulnerable groups. Plus increasing immunity from infection, of course.

So which team will win the “tug of war”? I don’t know, and note that some of the factors (e.g. immunity waning, and behavioural adjustments) are not that well modelled – if at all – at present. I need to do some work on my model to handle waning and booster vaccinations,

…and even then I’m not confident that it will produce a clear answer, given the wide ranges of uncertainty around some of the key assumptions. Just to take one example, we don’t really know by how much voluntary cautious behaviour is reducing R right now. It might be a lot,

…or it might be just a little, and it makes a big difference to your expectations of what could happen this winter (and next spring). So I’ll let you know if/when I have some clearer thoughts on this, but it might take a while, given the complexity of the modelling.

Also to set your expectations, my plan is to spend less time over the next few weeks on daily data analysis – since there are many people better able to do this than me – and focus my limited time on medium-term modelling, since there are few others working in that space.

(of course, I spent a good chunk of last week doing short-term data analysis on Boardmasters – it’s hard to resist when you can see a puzzle in the data and want to find the answer. So feel free to point out if I seem to be getting distracted from the main task at hand!) /end