4. Professor Michael Gravenor

PROFESSOR MICHAEL GRAVENOR (affirmed).

Questions From Lead Counsel to the Inquiry for Module 2B

Lady Hallett: I hope we haven’t kept you waiting,

Professor.

Mr Poole: Please take a seat, Professor. If you could start with giving us your full name, please.

Professor Michael Gravenor: My name is Michael Brynley Gravenor.

Lead 2B: Professor, whilst giving your evidence, if you can please keep your voice up so we can hear you but also so your evidence can be recorded. If I ask you a question you don’t understand, please do ask me to rephrase it; and thank you again for coming and assisting the Inquiry this afternoon.

The witness statement you’ve provided for this module is at INQ000347979. We don’t need to go to it but at page 35 you signed and dated that on 28 October of last year. Are the contents of that statement true to the best of your knowledge and belief?

Professor Michael Gravenor: Yes.

Lead 2B: Professor, by way of overview of your career, then, to date, you’re a professor of epidemiology and biostatistics at Swansea University; correct?

Professor Michael Gravenor: That’s right.

Lead 2B: Your academic career has been focused on infectious disease epidemiology and public health data analysis through mathematical modelling and statistics; is that also right?

Professor Michael Gravenor: That’s right.

Lead 2B: Although you have not previously worked with coronaviruses I think I’m right in saying you have professional experience in the practical application of mathematical models of infectious disease for policymakers; is that right?

Professor Michael Gravenor: That’s correct.

Lead 2B: In terms of your role during the pandemic, is it right that in May 2020 you joined TAG and the modelling subgroup which was co-chaired by Brendan Collins and Craiger Solomons?

Professor Michael Gravenor: That’s right.

Lead 2B: From April to May 2020 you and some colleagues from Swansea University, which we will call the Swansea modelling team, had been commissioned to provide a Welsh-specific epidemiological models to assist the Welsh Government respond to Covid-19; is that right?

Professor Michael Gravenor: That’s right, yes. I maybe wouldn’t say commissioned, but …

Lady Hallett: You volunteered.

Professor Michael Gravenor: Yes.

Mr Poole: Was the work that you did voluntary, unpaid work?

Professor Michael Gravenor: Yes. Up until much later in the day.

Lead 2B: To the best of your knowledge, Professor, were there any epidemiological models being developed and run in Wales for the Welsh population in February and March 2020?

Professor Michael Gravenor: I would say that some of the – or a good few of the SPI-M models would include Wales as an element. I wasn’t aware of any specific focus on Wales from any other models.

Lead 2B: So is it right in this period, February/March, Wales was heavily dependent on the modelling work that was being done by SPI-M, which was also being used to inform SAGE discussions?

Professor Michael Gravenor: Yes, that’s correct.

Lead 2B: Professor, when did you first become aware of Covid-19?

Professor Michael Gravenor: I think it would have been on the news in early January, I expect.

Lead 2B: What were your views about what was happening in Wuhan and the potential pandemic that was unfolding in the January 2020?

Professor Michael Gravenor: I think – I knew as colleagues several of the epidemiologists that were involved in the analysis of the outbreak in Wuhan and I read their reports as they were coming out in January, so I followed it quite closely because of my background and because of my links with some of the people working on it at that time. So I think I was fairly up to date with the evolving situation and the potential concerns, because I could see it being conveyed by colleagues.

Lead 2B: But obviously at this point you had no formal role in advising the Welsh Government, you were entirely independent?

Professor Michael Gravenor: That’s right.

Lead 2B: Did you raise the alarm with your professional colleagues? I think you say in your witness statement that very large numbers of infections in 2020 would be likely.

Professor Michael Gravenor: Well, within the medical school in Swansea, yes, in informal conversations, yes, it looked fairly obvious.

Lead 2B: What kind of interventions did you think would be required?

Professor Michael Gravenor: At that point, I – I really don’t know at that point. I think that’s going back quite a long time and a bit of hindsight involved there. I just expected a large epidemic.

Lead 2B: Could we, please, have INQ000374405 on screen.

This is an email that you sent on 24 February 2020, and it’s right you sent this to an individual who worked at Public Health Wales?

Professor Michael Gravenor: That’s right.

Lead 2B: I’m just going to pick it up from about four lines down:

“The other is corona related. I cornered Brendan the other day …”

Is that Brendan Mason, who worked for Public Health Wales?

Professor Michael Gravenor: That’s right, Brendan and I were lecturing together on that day in the medical school to medical students.

Lead 2B: So you say:

“I cornered Brendan the other day, he was measles outbreak lead, and I was asking him for data and information on MMR vaccination. He was looking pretty pale with 24/7 preparation for corona, and I briefly asked him if he needed any modelling support, might be a bit late in the day but Wales does represent a devolved and small public health response so potentially they can do things differently, university campus closures, that sort of thing, if they wanted. They may make some support on scenarios and if they do I wondered if you were interested.”

By doing things differently, what do you mean? Differently to England, differently to the rest of the UK?

Professor Michael Gravenor: I honestly can’t remember what I meant at that point. I really was at that point just reaching out to some colleagues to see if they would be interested in essentially some modelling analyses, because it might be provided. And it was through my contacts with Brendan that I was introduced to Public Health Wales and Dr Chris Williams and colleagues.

So at that stage I was following the modelling analysis quite closely, and I thought that some help may have been – in terms of interpreting perhaps the modelling output that was coming out. I was aware that these kind of reports might have been news to a lot of people. So I was just really fishing around at that stage for maybe something that we could do to help interpret what was happening in Wales. I was aware that the health response was devolved but I don’t think I was referring to any specific kind of activity at that point.

Lead 2B: No, you were obviously offering modelling support, and in that email we just looked at you said it might be a bit late in the day, so was it your view that, sort of, modelling should have really been put in place already by the – towards the end of February 2020?

Professor Michael Gravenor: Yeah, I guess that’s what I thought, yes. I – or that they would be relying on the, you know, well established groups in SPI-M and the large groups involved there, Imperial and the London School of Hygiene and Tropical Medicine. So I knew that a lot of modelling support was going to come from that area and that’s what might have been relied upon.

Lead 2B: Did you get a reply from that offer, did Public Health Wales take you up on your offer at this stage of providing modelling support?

Professor Michael Gravenor: On that day, no, but that is the – as I said, that was, I think, when – as I recall, that’s when, sort of, leave was cancelled, so to speak, for Public Health Wales and things got very, very busy indeed.

So I didn’t hear back immediately, but that is the route at which I was introduced to Public Health Wales and, ultimately, Welsh Government.

Lead 2B: And individuals like Dr Chris Williams who we heard from earlier?

Professor Michael Gravenor: Yes, so Dr Mason introduced me to Chris Williams and that’s really where the link started.

Lead 2B: On 14 March you sent an email, perhaps we can have it up, INQ000374409. You sent an email, and I think it was – you sent it to someone who I understand is a mathematical biologist, you say:

“All gone a bit bonkers here, control via natural herd immunity? I think they will backtrack on that. No evidence on duration of immunity, no evidence on long-term respiratory complications of survivors by age.

“Don’t understand why. Massive investment in the surveillance and testing of the obvious route into the country via half term ski trips. Shut down and have a substantial in % terms, it has to be (given it is at low end) effect on R0. Chase up cases like hell with all the resources going there, slow everything down in the summer, wait for better treatments for next winter.”

Do you stand by what you said then in this email of 14 March, that by locking down earlier, investing in surveillance and testing of those entering the country and chasing, in your words, cases up like hell would have been the best strategy for Wales in mid-March 2020?

Professor Michael Gravenor: It’s difficult looking back at these now, but parts of it, parts of it.

I think by the 14th – I think – we’d gone past the routes into the country, I think we’d gone well past that. So I think this is more of a comment that – as you know, herd immunity was discussed many, many times and the approach there means slightly different things to different people, but I think it well reflects my concerns that – I mean, there is a little bit of modelling there, which – the first thing that would come to my mind is that the idea that we reach a certain threshold and that’s it, that’s – as an epidemiologist, that never seemed like a very sensible conclusion, because viruses are extremely adept at changing and so there was always going to be concerns over immunity, and I think this is – this is before the term “Long Covid” was mentioned but systemic nasty respiratory viruses cause damage and we don’t know the problems there.

But in terms – and also, in terms of large scale emergency response, then shutting activities down and reducing contacts a lot has always been part of discussions of pandemic response in terms of things like influenza and – in worst-case type scenarios, then shutting down and waiting for vaccine improvements in terms of influenza is a discussed option. So it seemed that we were very much, very much at that point in mid-March.

Lead 2B: At this point in mid-March, I mean, it’s right, isn’t it, to say that there was no massive investment in surveillance and testing of those coming into the country, whether it be ski trips or otherwise. There was no shutdown, we know, for a further week. There was no contact tracing, let alone the rigorous contact tracing that you’re advocating for in this email.

I mean, would it be fair to say that this was a missed opportunity for Wales to have better controlled the first wave?

Professor Michael Gravenor: No, I think at this point there was not really – I – I don’t think this was applicable at that point in time, it just simply wasn’t. So this is not something that could have been done at that point in time, but it’s something that we had to move towards. And so I would slightly separate out the surveillance and the routes into the country: surveillance I meant within Wales, not in terms of international travel, which I think was long gone by then.

But no, I don’t think it’s something that could have been done at that moment. It’s something that would – that it seemed very apparent that we would have to invest in going forward.

Lady Hallett: You’re not saying it should have been done before?

Professor Michael Gravenor: I think there would – I think it’s apparent that there wouldn’t have been much time before to put that full surveillance and testing in place. I think, given the numbers of tests that were available at that time, that’s not something that could have just been done at that point.

So I’m not saying that it was an option that was missed, it’s just a comment that that’s where we’d got to head towards.

Mr Poole: This is mid-March 2020.

Professor Michael Gravenor: Yeah.

Lead 2B: And we’ve heard from various witnesses alarm bells ringing with them in early to mid-January, so it’s a two-month period.

Professor Michael Gravenor: And I think testing, bringing a testing system into place to deal with that is a huge challenge and, as it proved, took considerably more time. So I would not like to represent it as an option that was right there available at that time. That’s not correct.

Lead 2B: We’ve heard from Dr Roland Salmon earlier today about his views on the efficacy of lockdowns, and in your statement to the Inquiry you have said that it was clear that once the situation of late February, early March was reached, a first national lockdown was necessary.

Why do you describe it as being clear by late February, early March, that lockdown was necessary?

Professor Michael Gravenor: Well, there was very little – I think by late February, early March we – the reported seeding throughout the country was suggesting that the kind of things that had evolved around surveillance and testing and – were not sufficient to be able to control it UK-wide, and it was going to spread very, very rapidly.

So all we then have to – is to greatly reduce contacts, greatly reduce mixing between individuals, and at that point we had a pretty good estimate of the transmissibility, dealing with a very, very transmissible virus, and it seemed that there wasn’t really anything else in the short term, other than a substantial reduction in contacts.

Lead 2B: Do you think that the national lockdown should have been implemented earlier than 23 March?

Professor Michael Gravenor: In retrospect, I think it would have been helpful it was, yes.

Lead 2B: I’m right, aren’t I, that your modelling subgroup analysing an earlier lockdown in a July 2020 paper?

If we could, please, have INQ000302585 displayed. Excellent. Page 7, thank you.

I want to look at figure 9. This shows the effect of different timings of lockdown parameters on the potential course of the pandemic in Wales. Under the scenario of no mitigation measures at all, the pandemic would be expected to have reached a very high peak in mid-May. Yes?

Professor Michael Gravenor: Yes.

Lead 2B: If only pre-lockdown reduction levels of contact were maintained, so in other words no full lockdown, a peak of over 250 deaths per day may have been expected near the beginning of June; yes?

Professor Michael Gravenor: Yes.

Lead 2B: And if lockdown had been delayed by only five days the scenarios here suggest an additional 28% of deaths would have occurred. Am I reading that right?

Professor Michael Gravenor: Yes.

Lead 2B: And if lockdown had been introduced only five days earlier than 23 March, an expected 24% of deaths may have been prevented; is that also right?

Professor Michael Gravenor: Yes, that is the output from that model fit, yes. So moving all those reductions in contacts earlier would – can only really have the effect of slowing down the epidemic earlier, with a knock-on effect on the first peak. So I understand that there are debates about how we slowed down contacts prior to the mandated lockdown, and so this modelling exercise is an exercise in moving kind of all of those events earlier in time. But I think it’s – it’s an inevitable part of infectious disease dynamics that if you reduce contacts earlier, you reduce the peak. There are potential consequences of that later on, but in terms of the peak, yes. I think that that is – I think that that’s – there’s a strong, strong case for that.

Lead 2B: I was going to ask you, Professor, would an earlier first peak have led to a greater number of deaths in the second wave?

Professor Michael Gravenor: It is possible you have – it depends what you do in the second wave. There are – by suppressing it so hard, you have fewer people infected, and that means later on, when mixing increases, there are more people that can become infected in the second wave. You would technically have a slightly higher R value when the second wave would be initiated, which was inevitable.

So those things are important considerations, as I say, very, very important consideration, yes, but I guess it depends. I think the question of whether you’d have more in the second wave begs the question of what do you do differently in that second wave.

Lead 2B: Quite.

In your statement you describe other reasons, indirect reasons, why an earlier lockdown would have been preferable, and one of those reasons relates to care homes, and in your evidence you say care homes would have fared better from an earlier lockdown.

Just briefly, why do you say that?

Professor Michael Gravenor: It’s – that’s a tricky question, because we didn’t – we do not model care homes explicitly in our work. When – many of these models work better on a large scale, a large population scale. When it comes down to individual hospitals or individual care homes then there are local level effects that must be taken into account in terms of disease control that are going to be implemented and can never be captured by a broad scale modelling exercise.

I think that that comment would refer to the fact that the late lockdown meant we had a very high prevalence of infection throughout April and early May in the UK, and any effort to keep an infectious disease out of a risky environment, such as a hospital or a care home, is more difficult if the prevalence in the community is higher.

So we’ve since looked at the relationship between the prevalence in the community and risks in care homes, and there is a significant association between the two, in that clearly infection control is likely to be easier if the prevalence in the community is not so high.

So by keeping that prevalence – maybe it’s something we’ll come back to later – but by keeping that prevalence at a lower level, there are consequences of having a high – dealing with a high prevalence.

And I think the infection at the hospitals and the infection in the care homes, which don’t follow this epidemic curve exactly, they show different problems with infection spread in these environments, and I would say that one aspect of that link would be the high community prevalence.

Lead 2B: Professor, I just want to change topics, if I may, and talk to you about the Imperial influenza model.

The evidence heard in Module 2 confirmed that the earliest models created in the UK to deal with the Covid-19 pandemic were created by SPI-M using the Imperial model; that’s right, isn’t it?

Professor Michael Gravenor: Yes, amongst other models, yes.

Lead 2B: And you’ve described in your statement how the Imperial model had been developed for influenza.

It’s right, isn’t it, that there are advantages and disadvantages of relying on a model that has been developed for a different disease? So on the one hand you have the advantage that the model is available for use rapidly; on the other hand, the Imperial model being based on influenza has different epidemiological features that were less relevant to Covid. Is that a fair summary?

Professor Michael Gravenor: Possibly. I would put a lot more emphasis on the former. So I didn’t –

Lead 2B: On the advantages?

Professor Michael Gravenor: Yes.

Lead 2B: The positives?

Professor Michael Gravenor: I didn’t consider it a weakness really at all, because it was certainly very, very convenient that these issues had been thought about at a large scale and geographical scale and a lot of the impacts of disease spread had been built into them, and then changing those models to reflect, for example, the different incubation period or the different infectious period of a different virus is something that can be implemented by an expert team like Imperial very quickly.

Lead 2B: Let me just put to you some of the comments that Professor Mark Woolhouse made about the disadvantages of the Imperial model and see what you agree with and what you disagree with.

Professor Woolhouse said influenza models explicitly represented schools rather than care homes and influenza models tend to focus on social distancing as the preferred method of intervention rather than alternative interventions. And he identified two reasons for that: first, contact tracing is not a useful intervention for influenza due to its short generation time and high numbers of asymptomatic cases, therefore is not incorporated into the models, but obviously is a key intervention for SARS-like infections like Covid; and then second, he makes the point that influenza has a lower R number than Covid, meaning that social distancing measures required to keep an epidemic manageable can be much less drastic than a full lockdown.

The first question is: as a point of principle, do you agree with those observations?

Professor Michael Gravenor: Yes, I think all those are valid observations, yes.

Lead 2B: As a consequence, then, of using the Imperial model, do you think that that adopted a trend or a bias in favour of lockdowns rather than focusing on the contribution of, say, case detection, contact tracing, self-isolation, shielding, and so forth?

Professor Michael Gravenor: I’m not sure how one follows from the other. I mean, the models can’t include all the important factors, they’re always a simplification of reality. So I think the major drawback would be not having explicit care homes and that route.

So that is a – that is a problem. So I agree with those issues. I think that they don’t necessarily flow from choosing that model as the starting point, because, of course, all these models were greatly developed over time, but choosing this model as a starting point I don’t think that that was problematic in that sense.

It’s – it was identified very quickly over time what needs to be changed and added. Contact tracing, for example, was analysed by several modelling groups very, very early and models and papers were published on that in, I think, late January, early February. So some of those issues were being addressed quite early on, some but not all.

Lead 2B: Now, although you were not involved with the Welsh pandemic response at this period of time, are you able to comment on any concerns that the models were London-centric or followed a pattern too closely pegged to London?

Professor Michael Gravenor: At this point I would not – I wasn’t privy to any data or models themselves, so it’s quite difficult to comment on that, I would say. Yeah, I just – I just – at that point I was not actively using the models or building them.

So I think – do you mean London-centric in terms of data and analysing the outbreak in London? Because they weren’t in any way confined to London.

Lead 2B: Let me just put to you some comments from one of your colleagues on TAC and the co-chair of the modelling subgroup, Craiger Solomons, who has commented that the approach of trying to use the material model crudely adjusted for Wales’ population size provided results, in his words, of poor quality. And that – his reasons, I’ll give you the three reasons and then ask for your – whether you agree with them.

He said that because models were seeded to Wales rather than England they could not account for differences in Welsh demographics, differences in Welsh geography, rurality, socioeconomic factors, population movement, and also different timings and durations of NPIs.

Professor Michael Gravenor: Okay, so there’s several points there and I do agree with some of them, I guess.

So I don’t think it’s a problem with the models. It’s more the kind of questions you’re asking from the models, and I think they would be a little bit less Welsh-focused by these groups, which is perhaps not surprising. So I think it’s not the models themselves, no. It would be perhaps your last point, in terms of if slightly different timings are involved, the seeding of the models is a reasonable – reasonable point. Wales getting infections slightly after large parts of England means that at any point in time you might be at a slightly different stage of the epidemic. So having the ability to use those same models but in the Welsh context gives you a little bit more insight, I suspect.

Lead 2B: Now, you say in your statement that it was clear by the end of March that a Wales-specific model would be required, and we’ll explore after the break in a moment the development of the Swansea model in the spring and the summer of 2020.

Did the lack of a Wales-specific model increase planning uncertainty in Wales?

Professor Michael Gravenor: I think what I meant by required was we’d been asked to do it, so I don’t think –

Lead 2B: So in your view not needed?

Professor Michael Gravenor: I don’t think I would have known at that point. So when I said “required” I meant we would – we were – we were required to do it because we’d been asked.

Mr Poole: I understand.

My Lady, if that’s an appropriate point to take a break.

Lady Hallett: Yes, certainly. I shall return at 3 o’clock.

(2.42 pm)

(A short break)

(2.59 pm)

Mr Poole: Professor, I’m going to ask you some questions briefly about the Swansea model next, we know that the Swansea model was not operational or used by policymakers during the first wave of Covid and you very helpfully in your witness statement set out the timeline of its development, which I’m not proposing to take you through now, save to note that modelling work using the Swansea model commenced around May to June 2020; is that right?

Professor Michael Gravenor: That’s right.

Lead 2B: And then modelling results were available around August 2020?

Professor Michael Gravenor: That’s right, yes.

Lead 2B: Now, could you just please provide a brief high level overview of how the Swansea model worked for us?

Professor Michael Gravenor: It’s probably worth saying that there’s not one model. In the intervening time before we developed the Swansea model we provided lots of small modelling analyses and questions and developed lots of different models over the period, probably ten or 15 different models. But the main model we used, which was labelled the “Swansea model”, not by me, it – we – these models, as you’ve mentioned, have been in development – are best if they’ve been in development for a long time, so we took the decision not to build it from scratch. I felt at first it would be too difficult to do that.

But by that time a lot of the SPI-M modelling groups had made modelling frameworks available to the public, and we explored a range of those, and used a framework that was provided by the London School of Hygiene and Tropical Medicine, and then that’s the model that we adapted for Wales.

On a broad level, it describes the transmission of an infectious disease within the 22 local authorities of Wales, so it’s what we call a local authority level model. So the demographics are relevant to those local authorities. And then the results are collated on a Wales level. And that was the level where it was probably most appropriately used.

But at the heart of it it’s a local authority SEIR-type infectious disease model.

Lead 2B: And in terms of the uses of the model, is it right that the Swansea model played a role in modelling a range of key policy decisions, so if I just run through a few of them: firebreak, social distancing, self-isolation requirements, the reasonable worst-case scenario in autumn and winter 2020, the potential effect of the firebreak lockdown in October/November 2020, and also the return of children to school in 2020 and also in 2021; is that right?

Professor Michael Gravenor: Yes, I would say so. I think isolation-type models were probably based around contact tracing models, which were done separately.

Lead 2B: I think you say in your evidence that the Rt estimate alone is not sufficiently robust to inform decisions. What other factors would you be expecting policymakers to consider?

Professor Michael Gravenor: I think I’m referring there to the R value is crucial in terms of the direction that the epidemic is taking and how fast, but it has to be put in the context of a time. I think I would be referring there to the prevalence. So if, for example, there was an R of – an Rt of around about 1.4, you might expect the doubling time over about ten days, which is useful, but the situation there has to be related to the prevalence.

So, for example, if there were 100 cases a day a Wales, which at times would have been a relatively small number, this would indicate that in a week or so you might expect 200 cases per day, but if you were in a situation where the prevalence is already 1,000 cases a day, then in a week or so there’ll be 2,000 cases a day, and in absolute terms the growth rate’s the same but in absolute terms it makes a very big difference to the impact of that. So the impact is not directly from Rt, it is combined with the overall prevalence.

Lead 2B: I understand.

I’m going to ask you about some of the modelling then that the Swansea model was used for. The first major event I want to touch on in the summer of 2020 is Eat Out to Help Out.

Now, we know Eat Out to Help Out, that was introduced between 3 August and 31 August 2020, so the Swansea model was up and running, could have advised policymakers on how Eat Out to Help Out would impact transmission and impact on hospitality and deaths. Were you consulted on the Eat Out to Help Out scheme or asked to model any of its effects?

Professor Michael Gravenor: No.

Lead 2B: Can you help us, what was the community caseload of Covid-19 in Wales immediately prior to 3 August 2020?

Professor Michael Gravenor: I couldn’t tell you exactly. It was very low.

Lead 2B: Very?

Professor Michael Gravenor: It was very low.

Lead 2B: Low.

How would the removal of many NPIs affect the position in June/July 2020, as restrictions eased?

Professor Michael Gravenor: Well, it would increase the Rt value and we would return to an exponential growth of the epidemic.

Lead 2B: In your opinion, did the Eat Out to Help Out scheme accelerate the arrival of the second wave in Wales?

Professor Michael Gravenor: I haven’t seen any analysis of that and we haven’t conducted any analysis of that ourselves, so I can’t really comment on that.

Lady Hallett: I think we –

Professor Michael Gravenor: Anything that – sorry.

Lady Hallett: I’m sorry to interrupt. You carry on.

Professor Michael Gravenor: I would just say that anything that increases the – anything that increases close contacts in a risky situation is going to increase. Mixing – anything that increases mixing is going to increase Rt and accelerate the arrival of the autumn wave. The extent to which it happened, I really don’t know.

Lady Hallett: I think that’s consistent with evidence I’ve heard in a previous module. And I should also say that Rishi Sunak, who introduced the policy, indicated that Eat Out to Help Out was meant to be conducted in a Covid-secure environment. So I don’t know how one can factor that into modelling calculations, but you hadn’t done them anyway, so …

Thank you.

Mr Poole: Moving on to the autumn 2020 and the firebreak, you describe in your statement that it was clear by 11 September 2020 that the R number in Wales was above 1. At this time a TAC report I think referenced a SAGE R number for Wales of between 0.7 and 1, and stated that the current R number was higher than this suggests. Why do you think it was higher than suggested?

Professor Michael Gravenor: Well, one thing to say is that the published R numbers were always lagged by – in the order of two to three weeks, depending on where you really look at it. This is because the signal that we observe for any changes to transmission, say an increase in transmission, are cases and hospitalisations and possibly deaths, and these do not occur at the time of infection, they occur after a delay. And that delay could be in the order of two weeks.

On top of that, you – there is a delay from the last data point that you had before you estimated Rt, which could add a few more days to that as well, and then there’s potentially a delay in communicating that advice. So it all adds up to the most – the most up-to-date R value really reflects the situation a couple of weeks prior.

So if you want to make a comment on today’s R value, then you really have to look at the trends that you’ve seen in the past and any other knowledge you have about mixing.

So if we were in a situation where we see the trends in R increasing, and on top of that perhaps some other areas of the country sitting on SPI-M would be reporting R values above 1, so you can see trends there. But on top of that, if in terms of behaviour the only thing that’s really happening in terms of there’s not controls on – not so many controls on social mixing and schools have gone back and other activity is going on, then it would be a very reasonable conclusion that the R value was above 1 at that point.

Lead 2B: So when SAGE on 11 September were reporting the R number for Wales as between 0.7 and 1, and you say likely to be higher, what do you think the R number was more likely to be at that point?

Professor Michael Gravenor: I possibly could have brought that information with me, but I think we were head – I think it was in the order of 1.2, 1.3.

Lead 2B: I think I’m right in saying, aren’t I, by this point in time you had modelled a new reasonable worst-case scenario which showed a potential for a large second wave?

Professor Michael Gravenor: Yes.

Lead 2B: Now, in the modelling work that you were doing in late August, September, am I right in thinking that while you modelled some NPIs being introduced to reduce contact and bring the R number down, the assumption was that there would not be a repeat of the March 2020 national lockdown? Is that right?

Professor Michael Gravenor: In the reasonable worst case?

Lead 2B: Yes.

Professor Michael Gravenor: Yes.

Lead 2B: Where did that assumption come from, that there wouldn’t be a repeat of a national lockdown?

Professor Michael Gravenor: I think that would come under the remit of the reasonable worst case that we were asked to model. So it would be, the situation in August or September, this amount of behaviour, where is this taking us? So it’s – I mean, there are various different uses of the models, and modelling explicit policy such as introducing restrictions in movement, et cetera, would be part of a scenario modelling. In terms of the reasonable worst case I think the remit would generally be: if things stay as they are where are we heading? And it can possibly take into account some changes over time.

Lead 2B: Modelling was conducted for the 11 September 2020 TAC advice that went to the Welsh Government – I don’t need to display that advice here – it was noted in that advice the pattern of increasing cases is similar to the situation in February, action should be taken to prevent significant harm arising from Covid-19 or another national – sorry, or another full lockdown.

Then again, that was 11 September.

On 18 September a TAC advice, if we could have that, please, displayed, INQ000222823, as we see there it’s 18 September 2020, if we could look at page 2, the first bullet point, please:

“The epidemic is evolving rapidly across Wales and the UK, meaning that estimates become out of date very quickly. There is consensus that the situation continues to be serious. This is highlighted by the sad news that we have begun again this week to have deaths from Covid-19 recorded in Wales.”

Then if we can please have a look at the fourth bullet point on that page:

“A package of … (NPIs) on local and national scale may be needed to bring R back below 1. Some NPIs may need to be in place for a significant length of time, though an earlier and more comprehensive response is likely to reduced the length of time for which they are required.”

What did you envisage by an “earlier and more comprehensive response”, Professor?

Professor Michael Gravenor: So, again, the earlier that you act, you’re acting at a lower prevalence, and the degree which you suppress it then takes you down to an even low prevalence. So in terms of buying time, from that sense, acting earlier suppresses it to a lower level and delays the next action. Waiting longer means you have to either act more severely to bring it down to very low levels or you are acting to bring it down to a somewhat lower prevalence from which it will return as well.

I think the important thing of this point in time is we estimated how many people had been infected in Wales during the first wave, and it’s not a very large number, it’s maybe 6, 7% of the population at most, and it just left a huge potential for growth which is reflected in the reasonable worst cases for the UK as well. The situation is not quite the same because the R value is generally much lower than it was in March, and that’s because of the understanding of isolation and test and trace and just general realisation that you shouldn’t be spreading a virus.

However, the R value is – it doesn’t need to be very far above 1 to be problematic, and I think – you know, this is sometimes difficult to communicate, but the – an R value of 1.2 doesn’t sound much different to an R value of 1.1, but very approximately, in terms of contacts, you have to reduce your contacts by 20% from 1.2 and 10% from 1.1 approximately. So it’s twice as much effort, so twice as much of a reduction in contacts required just for small changes in R. So the potential was very, very much still there.

Lead 2B: When you’re talking about an “earlier and more comprehensive response”, is “comprehensive response” alluding to potential for the need for a lockdown or a firebreak lockdown?

Professor Michael Gravenor: I am – I don’t know. I expect so. I didn’t write this, of course.

Lead 2B: If we could, sticking with the same document, please, page 5 and then the third bullet point on that page, please. Thank you.

“In mid-April mobility of Facebook users in Wales was 50% lower than the baseline, this is 1% lower than the baseline and is up slightly from last week. 22% of Facebook users in Wales are staying put, similar to the previous week. In early April around 45% were staying put – this was around 18% in early March.”

Does this mean that, in addition to the worsening indicators that we’ve just looked at, people in Wales were travelling more in September than they were in March, which was obviously likely to culminate in greater community transmission?

Professor Michael Gravenor: Yes, possibly. I find it difficult to comment on this, I never analysed this data at all.

Lead 2B: I understand.

If we could, please, have a look at some further modelling for a 2 October 2020 TAG advice – thank you, INQ000066408 – we can see that on the screen there.

It’s page 2, please, first bullet point.

It says:

“Some data streams indicate potential slowing in the growth rate of the epidemic, but it remains likely that infection incidence is growing overall in Wales.”

Then, please, the second bullet point:

“The latest estimate of R; from … (SAGE) for Wales is between 1.3 and 1.6.”

Then, please, the fifth bullet point on that page:

“Unless measures bring R back below 1, it is possible that infection incidence and hospital admissions may exceed scenario planning levels.”

So from your point of view, Professor, was it clear from early to mid-September 2020 that significant intervention would be required to reduce transmission, and then by early October, this being dated 2 October, there was concern that hospitals exceeding scenario planning levels, so in other words the NHS in Wales would risk being overwhelmed?

Professor Michael Gravenor: Yes, I would agree with that, yes.

Lead 2B: Do you think the tone of the TAC advice documents that we’ve just seen was sufficient to convey the seriousness of that message to the Welsh Government?

Professor Michael Gravenor: I think I’d have to see the full context for that.

Lead 2B: Well, were there occasions when you thought perhaps stronger warnings, stronger messaging might be warranted?

Professor Michael Gravenor: I don’t think so. I don’t – I don’t think so. I think the – I think there was some uncertainty there, perhaps, in terms of the estimates and perhaps because we’d come from the period where the R value was estimated as being under 1, even though that was out of date, so I think there was a growing realisation here across TAC that we were heading towards interventions.

Lady Hallett: Could you look at the first bullet point, Professor. I mean, if I were a politician reading this, “Some data streams indicate potential slowing in the growth rate …”, I’d think, “Oh, good, we’re going the right way”, and I may not even focus too much on what remains of that sentence.

Don’t you think that should have been much more of an alert rather than, “Oh, we may be getting better”? It doesn’t sound very strong to me.

Professor Michael Gravenor: Yes, I think the second bullet point should have gone first on that, because I think by that point we were fairly clear that R was greater than 1.

It’s hard to remember exactly when this was written. When it was written and when it was dated might be two different things. But … yeah –

Lady Hallett: Do you know who drew up the briefs and the reports?

Professor Michael Gravenor: Sorry?

Lady Hallett: Do you know who drew them up?

Professor Michael Gravenor: No. No. I knew that they could be circulated. Yeah, circulated for comments, yes.

Mr Poole: Professor, you describe in your statement that you were first asked to model a specific firebreak scenario of two to three weeks on 11 October. Were you surprised that that work was commissioned a whole month after you started raising concerns about transmission rates in Wales?

Professor Michael Gravenor: Not really. Once again, I’d say we weren’t commissioned to do it. There was no commissioning involved at this stage at all.

Lead 2B: You were asked to.

Professor Michael Gravenor: And the second thing to say is it’s true that we were asked to do the specific two and three-week model. At that point we’d previously modelled a range of scenarios, so it wasn’t the first time we’d looked at that. So I think we’d been looking at firebreak situations elsewhere, circuit-breakers were being actively discussed in SPI-M and in many places, so that was not the first time we were asked to do it and we had been preparing that before.

Lead 2B: I mean, did the fact that you were carrying out this modelling work entirely pro bono, I think, until August 2021, did that have an impact on the work that you were able to do?

Professor Michael Gravenor: Yes, yes. It – we would have been able to do a lot more. So everything we were doing – the team at this point, myself and Professor Lucini and Dr Dawson and Dr Bennett, we all had full-time commitments to our roles in the university, so if we could have been relieved of that we would have been able to run more scenarios, I think we were quite keen on running additional models alongside – it’s important to have an ensemble if you can, and we could’ve looked at more scenarios. We could’ve done more things. We could’ve possibly brought in more people as well, which would have been very, very useful. So I think a mechanism that could have enabled that would have been quite useful, I think.

I think it’s an important point to make that in an emergency situation there are certain expertise that does not necessarily sit in a standing capacity within government or within health services, and it exists in places like universities, and accessing that expertise is – I think it’s very important, not just from our point of view, but important – there’s expertise in all sorts of the response, the behavioural side, the genetics, and accessing that – a mechanism to access that expertise I think is an important lesson we’ve learnt and – yeah. So this was largely evening work, yeah.

Lead 2B: That 11 October modelling of a two to three-week firebreak, were you surprised that you were asked to model a two to three-week firebreak as opposed to, say, a four-week or even longer firebreak?

Professor Michael Gravenor: Yes, I’ve certainly considered this since. I wouldn’t say I was surprised. We – we just did it. This is late in – not – this is 15 October. We have a question, we do it. And so I think just delivering that to a – to a level of quality that we’d be happy with is what took our focus. I don’t think we – we had already run situations that were longer, and we know that a longer firebreak would have more of an impact. So at that point in time, I can – I can say that we just took it and we did it.

Lead 2B: I think I’m right in saying you did in fact model, though, a four-week firebreak, did you not?

Professor Michael Gravenor: Yes, we’ve looked at – we looked at all ranges, we could –

Lead 2B: What –

Professor Michael Gravenor: – our model and simulations over many different combinations and repeated … all sorts of situations.

Lead 2B: What did the modelling show about the impact of a four-week lockdown?

Professor Michael Gravenor: I think – I think we – well, it would have driven the prevalence lower and it would have given more time before prevalence returned to the pre-firebreak levels. That’s something we reported on in detail for the two and three-week scenarios. For the other scenarios we have – had those numbers, but at this point this was the remit of our – of our investigation, we focused on those two and provided that.

Lead 2B: Having sort of perhaps stepped outside the brief and modelled a four-week lockdown, and having made the findings that you did, did you advise TAG or TAC of those findings at the time?

Professor Michael Gravenor: At the time, no. I think they were shared – I think they were discussed in the modelling subgroup perhaps prior to that, but I think by the time we got to dates going over the school half term, so when we looked prior to that, we just set up a set of scenarios in which you can vary the length of the firebreak, and they may or may not have had school closures and they may have had different effects. So there would have been very many scenarios that we would have, and then we would focus then on the two to three-week – and then we were given a date. So they wouldn’t have been directly comparable anyway because we were running over the half term, and then after the half term of the firebreak there was a mixed school – some – some return to school. So it was a very specific situation that we were modelling in response to that email, which does not directly correspond to the previous kind of circuit-breaker experiments that were done before.

Lead 2B: From what you’ve just said, it sounds as though by the time you were asked to model the firebreak it was pretty advanced thinking on the part of the Welsh Government that there was going to be a firebreak. I mean, did modelling – in your view, did modelling work need to be carried out in order for that decision to be made to impose a firebreak?

Professor Michael Gravenor: I think possibly not, I think it’s going to perhaps October we – there was a fairly well set plan for guide thinking in terms of the impact and maybe being most useful in terms of the return time. So it’s an important decision. So you’ve got to understand the particular circumstances in which it’s going to work and the uncertainty around that. So within the remit of the two and three-week firebreak we considered a combination of what the true value R was when we headed into it. Now, that was unknown, so we considered a range. Then we have to consider a range of the impact, and so we considered a range of those, because it may not reduce contact as much as it happened in March. So we can use that as a yardstick but it may not be as effective. Then finally we have to consider what is the likely R at the end of the firebreak. And there are certain things to consider there, it’s deeper into the winter, there might be a carry-over effect of the firebreak, whether it’s beneficial. Which is what largely transpired. So there are many different potential outcomes.

And I think the modelling is useful to sort of have those explicitly down – we don’t necessarily know which one of these is going to be followed so it’s not necessary to show that, but with that – with that analysis you can say that under the range of the assumptions – under the range of scenarios that we consider to be reasonable then this is going – this is a possible likely effect. And if that is a suppression for some extra time beyond the actual break, then it has – then it will have a big impact on prevalence, as it did.

Lead 2B: Professor, did you have any concerns that the Welsh Government on occasion were perhaps overly eager to obtain modelling outputs at perhaps the expense of timely decision-making?

Professor Michael Gravenor: It’s not my experience at all, no.

Lead 2B: The First Minister’s provided a statement to this module of the Inquiry in which he said that the firebreak produced the gains which had been expected but that the gains were, in his words, much more short-lived than the modelling available to the Welsh Government had anticipated. Now, we know that by early December in Wales indicators for clinical admissions and the Rt rate were nearing pre-firebreak levels, so is the First Minister right in that assessment?

Professor Michael Gravenor: I believe not. I think, as I’ve kind of described previously, the firebreak scenarios covered a very large range, but in reality the time bought was at the upper end of our optimistic scenarios. So while there were some scenarios in which it could possibly be lower, they were the very, very, very most optimistic.

The – what we – what we hoped was a baseline scenario was the current R and then it would have a large impact, because the – I believe the comms and the preparation was done very well and people were aware and they were – it looked as if they would be on board. So we set an impact that would be quite effective, not quite as effective as March, but a very effective one. And following that we assumed that the R would be exactly the same again. Using that model, we projected a return time to the pre-firebreak conditions of 38 days, and the data will show that the return time was in the area of 39, 40 days.

Now, I think that’s coincidentally close, okay, that’s not something you expect from these kind of model exercises, they do not have that kind of precision, but it coincidentally shows that that baseline scenario was almost exactly the time bought. And I would say quite a substantial amount of time, if – so 39 days added on to the time of the firebreak, which is only two weeks long, is quite substantial and it shows the trajectory that the epidemic took within the firebreak.

And that’s been – there’s many independent corroborations of that. You can see in the ONS data that the prevalence in Wales is half that of England around about that time, and you can – and there are, I believe, published estimates from independent modelling groups, the London School published a paper on the circuit-breakers in England which analysed the effectiveness of the Wales firebreak and showed, I think, that it had a 45% reduction in Rt, which is quite considerable.

So I think the time bought is pretty much very, very close to what we suggested in the models.

Lead 2B: So when the First Minister said the gains were much more short-lived than the modelling available to the Welsh Government had anticipated, I understand your evidence to be the modelling was accurate. Would it therefore follow that the duration of effects had not been communicated to the Welsh Government?

Professor Michael Gravenor: I don’t think that’s true, I think we – I think we said for a two-week we would expect a three to five-week reset, and for a three-week we’d expect a five to seven-week reset.

Lead 2B: So the net effect of that is you say that statement from the First Minister, that’s just wrong?

Professor Michael Gravenor: Yes, I think it doesn’t reflect the post-firebreak period.

What we found post-firebreak is that there was a little bit of a period where growth was – one might imagine that the R number declines very, very rapidly, and as soon as everyone goes back to normal it goes right back to normal. The evidence is that there was a period after the firebreak where it was actually growing a little bit more slowly than prior to the firebreak, and then it picked up speed. Which might represent the conditions of transmission and going deeper into the winter, as we approached December, but also the arrival of the Alpha variant, which is circling that time, which is considerably more transmissible.

Lead 2B: Professor, just before we leave this topic of the firebreak, just two short questions, and we’ve still got quite a bit to cover, so if you could try and keep your answer as brief as you can to these questions.

First is: was the timing of the firebreak in your view reasonable or should it have come earlier? My second question: should it have been longer?

Professor Michael Gravenor: I think the timing was – earlier would have helped, as we have talked several times about the prevalence issue. I do believe that the timing including the preparation and the comms was about right. But having it at the lowest end means that it was always going to be the minimum impact in the shortest amount of time. So looking back on it now, we – I do think it should have been longer.

A longer firebreak could have – given how effective it was, given how effective it was in reducing Rt a longer firebreak would have set – if that, if those benefits had continued, it would have set the prevalence down to a very low level, and then we would have headed toward December.

I think a four-week firebreak would have put the reset time deep into December. At this point in time we’ve got a lot of knowledge from – about the transmission conditions deep in the winter, including knowledge of Alpha, the Alpha variant.

So I think a longer firebreak would have put us in a much, a much better position in December and potentially could have avoided some of the worst of that second wave.

Lead 2B: And the four-week firebreak, just to be clear, had been modelled but the results of that modelling exercise had not been passed on to TAG or TAC?

Professor Michael Gravenor: Not as part of the – not as part of that commission, because the instructions came to provide evidence on the two and three-week.

Lead 2B: I understand.

Professor Michael Gravenor: The work has been – the effects have been shared informally in the modelling subgroup and in terms of just general discussions about a longer firebreak has a bigger effect.

Lead 2B: Professor, I want to move on to the winter period 2020. On 2 December TAG published a statement regarding NPIs in the pre-Christmas period.

Could we, please, have INQ000350039. Thank you.

If we could have a look, please, at page 3. If I can go to the second paragraph, in fact, on the third line of that second paragraph, the – where it starts:

“The firebreak had the intended impact of a short sharp early intervention to push back the epidemic by three to four weeks. The benefits of this period of negative growth have nearly been lost, with case numbers and hospital admissions nearly reaching levels seen at the beginning of the firebreak.”

Then, please, if we can go to the same page, page 3, but the fourth paragraph, that starts:

“Deaths are currently as high as May, with the excess death rate in Wales higher than in England and Scotland over recent weeks, and tracking above our reasonable worst case.”

That was obviously a concerning picture epidemiologically; yes?

Professor Michael Gravenor: Yes, yes, I think that’s possibly around the worst point of the epidemic, I would say, in my experience.

Lead 2B: And in the same document, the policy modelling done by your team at Swansea University compared some different NPI interventions over December 2020 against some varying rates of background Rt.

If we can, please, have page 12. Thank you.

So the policy options being compared were: first, no intervention; then entering Tier 2 restriction; and then, the third one, entering Tier 3 restrictions.

Pausing there, why were those the three options that were modelled? Was that the commission that you had that received?

Professor Michael Gravenor: For that particular paper, yes. So they would be directly asked to use those conditions. I think prior to that we’d already started looking at the next point at which changes would need to be made, and we’d already considered various levels including Tier 4.

Lead 2B: Now, the advice of TAG that was summarised back in 2 December 2020 – I don’t need this to be displayed – but that summary was that policy modelling suggests that – I’m grateful – introducing the equivalent of Tier 3 restrictions, for example closure of hospitality and entertainment, reduction in mixing prior to the relaxation of restrictions before Christmas will reduce the number of hospital and ICU beds required for Covid-19 patients and subsequent deaths.

Is it right that the strongest, if I can put it that way, NPI option then being explored in early December were Tier 3 restrictions, or you’ve just alluded to you were in fact looking at Tier 4 or effectively lockdown restrictions as well?

Professor Michael Gravenor: We were looking at them, but that was the question that was being asked here.

Lead 2B: So you were effectively reporting only – and this is no criticism, because of the extent of your commission – but you were reporting only on the strongest NPI option, being Tier 3 restrictions?

Professor Michael Gravenor: Yes.

Lead 2B: Now, we looked previously at a TAG meeting on 13 October. It was noted in those minutes of that TAG meeting that, and I just read this to you:

“Yesterday there was a COBR meeting and announcement around the [three] tiers in England – a SAGE chairs group took place yesterday and no one felt that the highest tier was strong enough to bring R below 1.”

So my question is this: if Tier 3 restrictions had been seen as insufficient before the firebreak lockdown, and Wales was quickly approaching pre-firebreak levels, why was it assumed that Tier 3 restrictions would be sufficient this time around?

Professor Michael Gravenor: Sorry, what was the date of the SAGE?

Lead 2B: 13 October.

Professor Michael Gravenor: So there was a lot of – there was a lot of debate about the R values to use for different tiers and it did change a little bit over time, and I think there was a little bit of a difference between areas and between DAs. So we’d previously used a value where it just pushes it under 1, and I think that was appropriate, I think that was appropriate at the time. Again, we would explore sensitivity to that. But by the time we’ve got into December, I think it was just becoming clear that that was not the case.

So I think by this, by around about this point – which is why I think I mentioned that this point in the epidemic, it became clear that this was definitely not the point in Wales. So shortly after this, we start to realise that Tier 3 is nowhere near enough. But we’re now dealing with a much more transmissible situation than we had previously, and bringing that kind of Alpha advantage in transmissibility, which we didn’t have at this point in time, into it would show that it would not be close to enough with Tier 3.

Lead 2B: So it’s your view that Tier 3 restrictions would not be sufficient to bring the R value down?

Professor Michael Gravenor: Oh, it would bring the R value down.

Lead 2B: Sorry, to bring the R value down below 1?

Professor Michael Gravenor: No.

Lead 2B: Yet we see here in this TAG –

Professor Michael Gravenor: Sorry, I don’t mean – it was around about this time, it was – I couldn’t – I wouldn’t say that it was before this was written, no. No, sorry, I don’t want to give that impression at all. It was right about this time.

Lead 2B: But I would be right in saying that nowhere in this advice does it allude to any concerns that in fact Tier 3 restrictions would not have the effect of bringing the R value below 1?

Professor Michael Gravenor: No, I don’t think it mentions that. It certainly has an impact on R which is going to reduce the number of cases, it’s going to reduce the number of hospitalisations but … yeah, I think that there are – there were points at which that would be a reasonable assumption to make, but those turned out not to be true quite soon after this.

Lead 2B: Before we move off this document, I just want to ask you some final questions about the advice given about pre-Christmas NPIs, and in particular schools.

I’m grateful, page 13 of this document.

We can see there in the table that your team had modelled the difference in hospitalisations and deaths between schools staying open and closing between 14 to 18 December. Schools staying open for that period you estimated would result in between 120 and 150 deaths; that’s right, isn’t it?

Professor Michael Gravenor: Yes.

Lead 2B: I’m asking you this because in a TAG meeting two months earlier the picture looked different.

If I can just have those notes, please, INQ000313192 – I’m grateful – page 4, paragraph 1, right at the top:

“When asking people to self-isolate because they’re Covid positive, there are issues around what other members of the household do and this is a moral and public health discussion and may lie outside the remit of TAG. Worth highlighting the evidence of children transmitting to adults is so limited that it may be worth letting children out anyway.”

And then, reading on, please:

“Concern is around children in schools and what happens there. Agree immune suppressed children are a different situation, but the current evidence does not suggest transmission in the school context. Still quite a lot we don’t know about asymptomatics infection; the modelling originally done was based on flu where children are super spreaders and this has not been replicated with Covid.”

But my question is simply this: why were you then, in December, modelling deaths against school closures if you weren’t satisfied that children were meaningful spreaders of the virus?

Professor Michael Gravenor: I’m sorry, I didn’t. This is not my opinion.

Lead 2B: So is the answer that when we look at the 2 December TAG paper, this was something you had been asked or TAG had been asked –

Professor Michael Gravenor: Can you clarify the date of this?

Lead 2B: This is 9 October 2020.

Professor Michael Gravenor: Okay. So I think the exact role of transmission in schools was pretty much unknown, I think, by this point. So this is all assumption. I think we’d gone through a situation where schools closed pretty quickly in March, prevalence was driven down to very, very low levels and then there was the school holidays and then a couple of weeks back in school. None of that is sufficient to give you a good estimate of the importance of transmission in schools.

Lead 2B: Now, we know schools were closed early on 14 December; hospitality, however, remained open. Do you think you should have been asked to model the effects of keeping schools open and closing hospitality venues?

Professor Michael Gravenor: Sorry, for which dates?

Lead 2B: So 14 December is when schools were closed early in Wales, but hospitality remained open. So my question was simply: do you think you should have been asked to model the effect of in fact keeping schools open but closing hospitality?

Professor Michael Gravenor: Yes, I think we could have done that, it’s a blunt tool, whether you can distinguish things very easily, in terms – in terms of schools we relied heavily on what we would call the contact matrix between different age groups and this is obtained from surveys and empirical epidemiology, surveys such as CoMix, and they basically tell you how often a person of one age is in contact with a person of another age on a typical day.

So within those surveys, the contacts that take place within schools are to some degree recorded, so those are the contacts that we would remove from our model when schools are closed. So it was – it was not based on the observation – so our model was not based on the observational epidemiology of how often transmission occurs within schools, it was based on the typical contacts that are made within schools and between school-aged people and people outside of the schools.

So in some ways it’s a very easy thing to implement in the model, and you can remove those, which simulates school closures. It doesn’t capture all of the nuances of that, but at the end of the day those contacts exist, and reducing any contacts in the model across or within groups and across groups – because not all the contacts can be maintained within children – if you reduce any of those contacts, you are going to reduce infections and cases and hospitalisations, deaths. The extent to which they are changed is very, very difficult to do – but ultimately all we’re doing there is reducing the contacts of that age group – and a little bit harder to do under other circumstances.

Lead 2B: Can we just return to the chronology. We were working our way through December and the advice that was being given by TAC. On 11 December there was a further TAC advice that was published. I don’t need it to be displayed, I’ll just read you a few relevant passages from it:

“Cases of Covid-19 per 100,000 of the population in Wales, have increased by 54% since our last report.

“As of 9th December, test positivity for COVID-19 … is above the red circuit breaker indicator threshold, at 19.4%.

“As of 11th December, the number of people with confirmed COVID-19 in hospital, has increased by 9% since last week, remaining higher than the April peak and above the red circuit breaker indicator threshold.”

So as with the October firebreak, I think you were commissioned to provide some further modelling, but was it not obvious from 11 December that Tier 4 restrictions, namely a lockdown, would be needed to get a grip on transmission rates?

Professor Michael Gravenor: Yes.

Lead 2B: Do you think that a third lockdown should have been implemented earlier?

Professor Michael Gravenor: Yes, in retrospect, yes.

Lead 2B: Now, in December 2020 Professor John Edmunds stated, “this is the worst moment of the whole epidemic”, and in January 2021 that, in his words, “really major additional measures” were needed.

We now know that Wales fared particularly badly in the second wave and in December 2020. Do you think the same comment applies then to Wales? Was Wales caught in a very bad position?

Professor Michael Gravenor: Yes, I would say so. I think it’s – it doesn’t give much solace, but I think there is a – it was somewhat unfortunate to come out of a firebreak into the highest transmission period, December, with the emergence of the Alpha variant which went on to make up a very, very substantial part of the second wave. So whilst there was measures going on in early December, they clearly weren’t enough, and they came at a time of the return period from the firebreak.

So I think it is one of the worst situations in that point, in the sense that we dealt with – we had to deal with a very difficult situation, which was very high transmission rates, from the point of high prevalence.

So it comes back to the point again, is if you had a longer firebreak, for example, you deal with those situations from a lower prevalence, and so by that point you would be able to perhaps respond or consider the response a little bit – a little bit better, and certainly the – you know, the consequences of that period would have been – would have been reduced.

So whilst Alpha can’t be predicted, it kind of illustrates that importance of the overall prevalence, and so dealing with that already high was a – it was rather a sort of perfect storm and I think – I think that could have only been avoided by a longer firebreak before – beforehand.

Lead 2B: Professor, just two short topics before I finish.

Indirect effects of NPIs. Could more have been done by your team on indirect harms over 2020 and 2021, had the resources been made available to you?

Professor Michael Gravenor: Are you referring perhaps to economic type effects or –

Lead 2B: Indirect effects, NPIs, so social, economic, quality of life.

Professor Michael Gravenor: This is not our area of expertise in terms of social effects. No. And I – we set up the model very, very early to build in potential costs so that indirect effects and cost benefits could potentially be weighed up against each other. That was done in September.

So, however, I don’t think that the numbers to bring into the model were ever – ever provided, were ever made available. So I think the framework was there, but if we’d have been given more time, no, because there was nothing to – there was nothing to bring in. So we were, I think, frustrated by that and made a certain amount of progress on the costs sides for the health. That fed into every model run. But in terms of how the knock-on societal or economic effects, I felt that there was no information being provided to us at all that would allow us to bring that into the model.

Lead 2B: Finally – and I think you may have answered this question already in your evidence this afternoon – what, if any, modelling was done to reflect the elderly population in Wales? What specific considerations were given by TAG and TAC on preventing the spread of Covid amongst the elderly, particularly those in care homes?

Professor Michael Gravenor: Well, TAG would have a care home group which worked with the SAGE care home groups for very – mostly, that’s not something I was involved with. In terms of modelling, it is – it is something that was not part of – most, I think, almost all the models didn’t model care homes explicitly. So they were not – they were not tools that were well developed for that question. You can look at the questions of reducing contact in elderly and how much that would improve matters, but that’s – that’s not much help if you really don’t know what’s going on on the ground.

So I think that that is a – not so much of a modelling question, more of a disease management on the ground type of question. And if we understood that much better, then perhaps it could feed back to the models, but it’s not something we were involved with at all.

Mr Poole: Professor, thank you. Those are all my questions, but there are some questions from behind me.

Lady Hallett: Ms Heaven.

Questions From Ms Heaven

Ms Heaven: My Lady.

Good afternoon, Professor Gravenor, just a few short questions. I represent the Covid-19 Bereaved Families for Justice Cymru.

I want to start by asking you – so, sorry, it’s back to the firebreak, please. So this is a comment that was made to the Inquiry in the witness statement of Andrew Nelson – I don’t know if you know who he is or was at the time – chief executive information officer at Cym Taf Morgannwg University Health Board. Just so that we can be clear, part of his role in the pandemic was to model the impact of Covid on hospital flows, resources and healthcare systems.

I think you nodded when I asked you if you knew who he was; is that correct?

Professor Michael Gravenor: Yes, Andrew was a very valuable member of the modelling subgroup, TAG.

Ms Heaven: So he says this at paragraph 250, my Lady:

“In regard to the firebreak, it is apparent from emails and files [which he attaches, I’m not going to show them to you because we don’t have time] that I and others raised concerns that the two-week period was going to prove insufficient to allow Wales to avoid a pre-Christmas lockdown and that it would not reduce the prevalence of Covid to a level in line with the Swansea University model for Q3/4 2020/21 which had formed the basis of Welsh Government’s planning guidance to the NHS at that time.”

So we obviously know – and we can see the email that you received from Rob Orford on that Sunday evening on 11 October requesting you to look at the two to three-week period.

So my question is this: were you then made aware of these concerns that were being raised, we know it was on 16 October, by Andrew Nelson?

Professor Michael Gravenor: I don’t recall that, no, I don’t recall that being a major discussion. We were very – I think it could well be something that would have been commented on, that when the announcement came that it was going to be about two weeks, I think several people would have perhaps made the comments that they would like it to be longer. I don’t remember it being a major point of discussion within our modelling group.

Ms Heaven: Well, I can’t take you to any emails to show that you were aware, so I’ll move on from that point.

He also says this:

“With the benefit of hindsight, the modelling group maybe should have done more analysis of extending the autumn firebreak until the majority of the most vulnerable JCVI groups had been vaccinated.”

And he says 80% of the groups had received their first vaccination by 16 February, and he says “as this may have reduced fatality”.

So before I ask you for your comment on this statement, just so that we can understand your evidence just a moment ago on this issue, you said that if there had been a longer firebreak this would mean very low prevalence into December, you said four weeks would push deep into December, and you say in your statement – and indeed you’ve said it today – that this would have meant that Wales would have faced the period of high winter transmission plus the emergence of the Alpha variant from a starting point of much lower community prevalence than it had to face in December 2020.

So that’s your evidence.

Now, just before I ask you to comment on the vulnerable groups and the vaccine roll-out, presumably the Welsh Government knew at the time, in October 2020 and before the firebreak, what you’ve just said today because, as you confirmed, you had modelled a four-week firebreak and presumably you’d fed the results back into the Welsh Government; is that correct?

Professor Michael Gravenor: Sorry, I described that in a little bit of detail. I wasn’t asked specifically to model for that. We set up the model to contain firebreaks of any duration.

Ms Heaven: Yes. But you say in your witness statement that you modelled two to four weeks.

Professor Michael Gravenor: Yes.

Ms Heaven: That’s what you say in your statement. So presumably you fed the results of a four-week firebreak back into the Welsh Government?

Professor Michael Gravenor: It would have been discussed at some point, at some point perhaps at the policy modelling group, but –

Ms Heaven: So if you –

Professor Michael Gravenor: – I can’t point you to the exact –

Ms Heaven: No.

Professor Michael Gravenor: – exact time. As I say, those scenarios are not directly comparable to the commissioned two/three-week ones because they had specific dates and specific scenarios round school.

So the point being we considered a range of firebreaks and then, by then, it was narrowed down to two to three weeks.

Ms Heaven: Okay. But if three weeks buys a three to five-week delay, does it follow that four weeks buys a four to six-week delay?

Professor Michael Gravenor: So I would say three weeks buys a five to seven-week delay.

Ms Heaven: Okay, so what does four weeks buy?

Professor Michael Gravenor: Seven to nine.

Ms Heaven: Okay. And the Welsh Government knew that, did they?

Professor Michael Gravenor: I wouldn’t say that they knew that, that was not in the report. So the report for the five to seven weeks and the three to five weeks was in the report that went in for the two to three-week firebreaks. That did not include the four-week firebreaks.

Ms Heaven: Why not?

Professor Michael Gravenor: Because that was the report requested at that specific time for the two to three-week firebreaks. It didn’t reflect previous work, it reflected the questions that we were asked at that point in time, and we have to be very focused on those –

Ms Heaven: Okay.

Professor Michael Gravenor: – and all the scenarios – there are, as I mentioned, there are lots of scenarios around those, not – that don’t just involve the time of the firebreak.

Ms Heaven: Okay.

Professor Michael Gravenor: So once we were asked to do two and three weeks, we focus on that.

Ms Heaven: We don’t want to take an unfair point against the Welsh Government; they didn’t know about your results for the four-week modelling?

Professor Michael Gravenor: No, but I think we would – we would all know that we have the evidence from a two-week firebreak in terms of how much of an effect it has, and the three-week firebreak –

Ms Heaven: Yes.

Professor Michael Gravenor: – the extrapolation to a four-week firebreak is –

Ms Heaven: Common sense?

Professor Michael Gravenor: It’s common sense.

Ms Heaven: Okay.

So let me move on to the question, then, on the vulnerable groups and the roll-outs. You said you modelled many scenarios. We can’t see any evidence in the disclosure or in your witness statement to suggest that in autumn 2020 your modelling team was asked to or indeed modelled various scenarios relating to an extended lockdown – so this is the Andrew Nelson point – beyond four weeks, factoring in things like the proposed timing of the vaccine roll-out to vulnerable groups; and of course we know the vaccine arrived in Wales in December 2020 and I think the roll-out started in the January.

So is the Inquiry to understand that you were not asked and hence did not conduct modelling on this issue, so extended lockdown, linked to timing of vaccine roll-out?

Professor Michael Gravenor: Extended firebreak?

Ms Heaven: Yes, firebreak, thank you.

Professor Michael Gravenor: No.

Ms Heaven: No. So you were not asked and you didn’t model it?

Professor Michael Gravenor: I guess other than the four-week, no.

Ms Heaven: No. Okay.

Very finally, if I may, my Lady, 22 December 2021, First Minister for Wales, Mark Drakeford, announced restrictions that would come into force on Boxing Day, and this was obviously in response to Omicron. It includes, as I’m sure you remember, the rule of six, meeting in pubs and restaurants, cinemas, face coverings in restaurants but they could be taken off when you were sitting down, outdoor events limited to 50 with 30 indoors but no restrictions for smaller meetings in private homes.

So it’s just a very short question: were you asked to model these range of proposals that were announced on 22 December and, if not, do you accept you should have been asked? A.

Ms Heaven: 2021.

Professor Michael Gravenor: Erm –

Ms Heaven: So it’s the rule of six again.

Professor Michael Gravenor: So in 2021, at exactly that time, we modelled a lot of scenarios for Omicron. At that point I think we were largely modelling the range of scenarios that reflected the uncertainty regarding the severity of Omicron. So we’d had a very limited number of – a very limited amount of data on the hospitalisations and deaths which indicated the levels of severity. So we had to run sets of scenarios that were all consistent with that and then see what was happening next. So I think – as I remember, we didn’t – we certainly didn’t, in answer to your question, model those very, very specific interventions because we never do.

Ms Heaven: Okay.

Professor Michael Gravenor: It is more the broader reductions in transmission that are accompanied by those interventions, and we certainly modelled lots of scenarios of Tier 1, Tier 2, Tier 3 –

Ms Heaven: Okay.

Professor Michael Gravenor: – type interventions in the period from December and going into January with Omicron, as we gradually learnt more about its severity. But quite a lot of scenarios December …? then, yes.

Ms Heaven: Well, thank you very much, those are my questions.

Thank you, my Lady.

Lady Hallett: Thank you, Ms Heaven.

I think that completes the evidence for this week.

Mr Poole: My Lady, it does.

Lady Hallett: Thank you very much, Professor, and if by the sounds of it you had to fulfil your other full-time commitments as well as doing this work, please accept my gratitude, I’m sure the gratitude of people of Wales, to you and to your colleagues.

The Witness: Croeso.

(The witness withdrew)

Lady Hallett: Thank you. 10 o’clock Monday, please.

(4.05 pm)

(The hearing adjourned until 10 am on Monday, 4 March 2024)