Thread by @jadenozzz

Starting now:

House of Lords #ScienceofCOVID19 inquiry, including modelling experts discussing the science behind the epidemiological models which informed the UK response to #Covid19UK

parliamentlive.tv/Event/Index/c3…

Neil Ferguson now on stressing that many different groups are involved in modelling.

Another witness discusses uncertainty around when transmission occurs (e.g., how long before symptoms)

Peak of cases can be due to either suppression measures or natural course of disease (by "use of susceptibles")

Better understanding now of hospitalisations and deaths in the UK have helped refine models. Previously, data from Wuhan and cruise ships was being used to make predictions.

Italy indicated a "potentially large outbreak in the UK".

Estimated that 20% of hospitalisations in the UK would have acute respiratory distress syndrome, with 15% needing ventilation.

Originally fatality rate predicted to be 0.5-1%. Narrowing this range with more data now.

Neil Ferguson says "it will be a learning experience as to how we allow society to resume whilst we maintain control of the virus"

@AlistairHaimes

Now discussing effect on specific groups.

Neil Ferguson says it's unclear whether BAME is an independent risk factor, or is associated with socio-economic status, lower ability to shield (eg. needing to go to work), and high prevalence of other risk factors in this group.

Neil Ferguson says public's adherence to 'lockdown' was better than expected.

Another witness (not labelled in the video) says it's difficult to understand how adherence to restrictions affects R change (ie. epidemiological change in transmission)

Early predictions could have been too pessimistic.

Lockdown may not have had an effect as R was already declining. Thomas Mernier finds no evidence that lockdown may not have saved any lives.

Witnesses disagree and say that just moving the curve along by a few weeks isn't the only measure of lockdown effectiveness (but no clear answer as to what was a good measure of their effectiveness)

Here is the article:

medrxiv.org/content/10.110…

Now discussing Sweden and policy differences. Neil Ferguson says they used the same science but came to a different policy conclusions.

Ferguson "[Sweden] make the argument that countries will find it very hard to stop second waves, &their policy in... front loading the health impacts will allow them to leave lockdown earlier,I don't agree... but scientifically they're not that far from scientists in any country"

Deaths are a lot lower in Sweden than would have been predicted.

Neil Ferguson says social distancing in Sweden has led to a R0 of ~1.

Now discussing R0 in different contexts, with discussion of nosocomial transmission.

Question is why are we not getting separate R values for different contexts?

Expert witness says a lot more transmission is happening in hospitals and care homes than community

Neil Ferguson says he agrees and expects this to be a distinct possibility.

"If we'd done a better job of reducing transmission in hospitals and care homes we would have a little bit more wiggle room in [line broke]"

Q is asked: Worrying assumptions were made about death rates early on. Looking back, how much did we get right and not?

Witness: Alarmist values early on were about worst case scenario, if we did nothing at all.

"If we'd have thought more about what was happening in hospitals and care homes, as you've heard, early on perhaps that's one of the areas where modellers have dropped the ball, y'know maybe we should have thought... 1/2

"... with hindsight it's very easy to say care homes and hospitals are these huge collections of vulnerable individuals so with hindsight we could have modelled those early on and thought about the impact there" 2/2

Ferguson: "Most chains of transmission in the UK came from Spain, and to some extent Italy. There were 100s if not 1000s of infected people coming in in late Feb and March, epidemic was further ahead than we thought."

He says this also explains higher mortality figures

(Bear with me, I will resume this after I come back from a meeting!)

Typo, sorry. Found no evidence that lockdown saved any lives

Neil Ferguson: relaxing controls won't automatically result in increased R, might get a small increase.

Transmission will remain flat between now and September.

Moving in to autumn winter is less certain.

Prof Matt Keeling: Could see more infections rising if certain parts of community stop adhering

Questions now on test and trace:

Will this improve accuracy of transmission data.

Dr Adam Kucharski: less disruptive control measures can exist with better testing and tracing. Can better understand settings which drive transmission

Dr Adam Kucharski: widespread screening in hospitals can show asymptomatic transmission in such high risk environments.

Ferguson: testing and tracing is not a panacea. Asking a lot of people to isolate for 14 days. If it works it could reduce the R by 22.25 (%?) but unsure

Imperial model written in Fortran?

Ferguson: No, C. File has been released so people can check.

Now hearing from new witnesses:

Dr Paul Birrell, Medical Research Council (MRC) Biostatistics Unit

Professor Mark Woolhouse OBE, Professor of Infectious Disease Epidemiology, University of Edinburgh

Prof Woolhouse: models looked at granularity of the effect of schools on transmission, because these have their ancestry in flu models, but not care homes, and where we really needed it was in care homes.

Prof Woolhouse: another good example of the limits of granularity in modelling are in contacts - based on self-reports of who people see, but doesn't consider whether they see them indoors or outdoors, "there's a MASSIVE difference in this virus between indoor & outdoor contexts"

Question: models seem to have considered hospitals but not care homes, even though Italy and Spain report problems in care homes.

Prof Woolhouse: during lockdown, deaths in Scotland more than half >80 years old, more than 40% live in care homes

Question: are we watching data from other countries as they loosen restrictions?

Witness: Initially we were reliant on data from China, but observing Italy may be more useful now.

Question: social distancing was introduced, and then shortly after, lockdown. What do the data show about these steps?

Prof Woolhouse: we were initially considering an emergency lockdown of 6 weeks. We would have given a different answer if we'd been told the lockdown was for 8 weeks or 12 weeks or 6 months or whatever it's going to be

Dr Birrell: releasing measures at the wrong time could cause second wave during flu season. Mid March the tide turned, shift to 'we need to lockdown now' to [stop] the overwhelming of health services. And that was bourne out.

Could modelling have supported different interventions or different timings?

Woolhouse: if lockdown was going to be 6 months it should have started earlier

Question about public understanding of the models and data, including use of R. Are we getting it right?

Woolhouse: focus on single R is getting it wrong, especially to drive policy. This is the impression, that R is a critical number...

Woolhouse continued: Policy goals I considered were to save lives, protect NHS staff and reduce burden on NHS as a whole. R does not necessarily directly translate into these goals. What's happened over lockdown is R has come down significantly in the community...

... but it has not come down and may even for a spell have been above 1 in care homes, "well that immediately tells you R isn't the right thing to be looking at, it's the wrong way around! If I wanted R to be below 1 anywhere I'd want it to be below 1 in care homes"

Woolhouse: this focus on R has kept our eye of the ball, rather than where the focus should be - [protecting] the population over 70, that's where the risks lie. And R does not capture that.

Dr Birrell: Interpretations of R and usefulness have changed over time. More useful in growth stage. "R loses all utility in the care home epidemic, we should be using different metrics"

Woolhouse: go back to old fashioned public health and go back to risk. Individuals can make informed choices about how they behave. R number does not help us know how to behave.

Woolhouse: models not the only thing needed to guide future response. Old fashioned epidemiology such as contact tracing is more important. Focus shifts from models to data.

Woolhouse: there was not enough in the models about care homes. And we still aren't getting the right data we need from care homes.

Dr Birrell: antibody testing will also be required and should be considered, but used carefully due to sampling biases.

Dr Birrell: a single epidemic curve produced in a model is not useful, because it's going to be wrong. You're setting yourself up to fail. It's been a battle to communicate that we should be providing a range of possible values which includes the value that eventually transpires.

Prof Woolhouse: single R and new cases need to be broken up by geographical location, age, risk group, etc. We need to concentrate where we're most concerned, need data from hospitals &care homes, it's starting, but the data management systems are "very cumbersome indeed"

Prof Woolhouse: we can predict things for about 2 weeks ahead, but not really any longer than that.

Dr Birrell: data should have primacy over any model going forward

Dr Birrell: we can't predict behaviour of public going forward.

Prof Woolhouse: modelling community is now concerned about is that R could get to e.g. 1.1, we're concerned about a second bump, but not about a second wave

Dr Birrell: R of 1.1 is difficult to detect

End!

You can watch here: parliamentlive.tv/Event/Index/c3…

@threadreaderapp unroll