Sunday 24 May 2020

It's on his head

I last wrote about Prof Neil Ferguson (the Imperial College modelling dude) back in April (see University Challenged - it's all in his head,  8 April) when I expressed shock that the computer model he used to influence government policy is entirely undocumented. Since then Ferguson has resigned from the the government's Scientific Advisory Group for Emergencies (SAGE) after, as some newspapers put it, he "broke his own lockdown rules" having a lady friend round to his house. (Perhaps I should have ended that sentence before "at his house").  His work prompted the lockdown by modelling options rather than setting the rules but never mind as said lady friend, Antonia Staats, added to the amusement value by telling the Guardian that she did not believe their actions to be hypocritical as she regards their households "to be one".  The Staats have two children and an open marriage. According to the Sun at the time of her second visit to Ferguson she suspected her husband had symptoms of coronavirus.

Rather belatedly a number of journalists including Vanessa Chalmers and Matthew Syed have joined me in questioning Ferguson's flaky software. They both report a flurry of criticism against Ferguson's modelling which predicted up to half a million deaths if the UK didn't change track. (Actually I think he predicted half a million in a "do nothing" scenario and a quarter of a million if the government maintained the mid March social distancing protocols without further measures).

What's happened since I inferred that Ferguson's code was crap because he couldn't be arsed to document it properly is that he's published it and others have now decided that it is indeed crap, or at least "unreliable". The program has been called a 13 year old "buggy mess". A group at Edinburgh Uni found that the model could give different outputs when fed the same inputs and gave different results when run on different computers, both of which are absolutely standard checks for software integrity. One source said that this quality of work would have got Ferguson sacked in private sector, which was broadly what I was implying in my article of 8 April. (Yes, of course I am also implying that it's taken the country's journalists only 4 to 6 weeks to catch up with me).

But wow. What if Ferguson's modelling really isn't fit for purpose? While the UK government had signposted that it might move to stricter restrictions, and potentially lockdown, there was an almost audible crashing of gears when Ferguson's body count forecasts were released. What if they were overstated? I wouldn't put it past Ferguson to have tweaked his inputs to get the biggest numbers, as he does seem to be an attention seeker. So was lockdown an over-reaction? After all some countries have never gone for a full lockdown, notably Sweden and South Korea.

The costs are already coming through - borrowing at the highest monthly figure on record in April and all those deferred operations and treatments storing up health impacts for starters. And not just at hospitals - a relative of ours can't get an essential B12 injection because the GP's aren't providing the service and is having to pay for less effective prescription for an oral alternative. Multiply such things by millions of people and thee will be plenty of non-covid health effects from the lockdown. Plus jobs destroyed and schools closed. And all that mental health anguish for families unable to see each other and, of course, the fans of Liverpool FC.......

With regard to schooling, I'm sure the children from middle class backgrounds will soon catch up and might even get some benefit from the change in routine. But the reason the government is desperate to get year 1 and year 6 children back is obvious: so many studies show that children from deprived backgrounds suffer disproportionately. The start they get off to matters (hence the need to get year 1 back) and the start 11 year olds make at high school on completion of their final primary year 6 can also have a huge effect. I think the government already has its eyes on the next election and is worried that the current school closure, along with the inevitable economic hardship for vulnerable communities, could be damaging enough to increase inequality by 2024, fatally impacting its levelling up agenda and aspirations to keep the red wall flat. It may be self interest but who says these things don't matter to Tory governments? They obviously do to this one.

Matthew Syed makes the interesting assertion that the government and its advisers have been guilty of groupthink and attention blindness. He has ploughed through the minutes from scientific advisory groups which were rather belatedly published recently. He notes the documents are impressive and most folk would conclude are underpinned by cutting edge science. It's only when you look at it in the round that you notice something he calls "curious". They are all specifically about flu.

There is the "scientific pandemic influenza group on modelling". The key planning document that has informed government policy for a decade is called the "UK Influenza Pandemic Preparedness Survey". The review of whether to close educational facilities is called the "Impact of School Closures on on an Influenza Pandemic". There is a paper called "Impact of Mass Gatherings on an Influenza Pandemic".

OK, they couldn't look in detail at coronavirus in advance, SARS-COV-2 didn't exist - though you might think they would have at least considered whether the experience with SARS and MERS - and indeed ebola - was relevant and different. There is a reason this is important, indeed critical: flu has long been considered impractical to contain, it spreads too efficiently. But covid-19 can be contained: it has been in South Korea, for example, where there hasn't been a single day with deaths in double figures even though it has kept offices, restaurants and shops open.

The logic of the UK's "contain, delay, research, mitigate" strategy is based on flu. Syed suggests it was the mindset of flu, not experience emerging from Asia that led our Chief Medical Officer, Chris Whitty to assert on 12 March that containment was pointless. (Actually by then I suspect he was right, but not a couple of weeks earlier). Perhaps it was also why his deputy, Jenny Harries - admired in our household for her performances at the Downing Street briefings - said on the same day that "community testing was not appropriate". (Though maybe it was not appropriate because it was not feasible with the capacity then available, which I don't myself think is blameworthy though we might otherwise have started to push for more capacity sooner).

Syed thinks the government is  not culpable for following the advice of its experts during this period. Ministers aren't modellers or health experts and can't be expected to second guess the scientists they have decided to trust. But if the scientific papers had been published sooner the blind spot might have been spotted. There was no downside to publishing earlier: the government was following the advice and and, as Jeremy Hunt has said, it's not as if transparency would have handed our enemy, the virus, some secret information it could use against us. If there was groupthink it was encouraged by the fact that SAGE is full of modellers and clinical academics but has no experts with an explicitly public health or infectious diseases background.

The Imperial College model was, of course, built for flu. And it is possible that not only the UK has been fighting the wrong war: the policies of the US, French and German governments were also influenced by Ferguson's work.

You might by now be thinking that I've got it in for Ferguson. Not particularly, but here comes the cheap shot. He's got form. His work has advised the actions of five prime ministers, on foot and mouth (2001), SARS (2003), swine flu (2011), MERS (2012) and ebola and zika (2014). In the foot and mouth outbreak nearly 6 million cattle, sheep and pigs were culled, partly because Ferguson's modelling had suggested animals on farms neighbouring an outbreak should be included even if there was no evidence of infection. The Times says that "experts later claimed the work was seriously flawed". In 2005 he said that up to 200 million people could be killed by bird flu. Now he did say "up to" but by 2009 the toll was 282. (I remember my company buying a huge job lot of tissues and hand sanitiser, see my post of 15 April). In 2009 a government estimate based on Ferguson's advice said that a "reasonable worst case scenario" was that swine flu would lead to 65,000 UK deaths. OK "worst case" but the total was actually 457.

Maybe Ferguson is like one of those doomsters who can say they correctly predicted the latest stock market crash, having predicted one every year for the last 20 years......

I don't know Ferguson's politics but as he is 1) an academic and 2) his floozy is a left wing campaigner I'll bet he reads the Guardian and not the Telegraph. Which is just as well as as the Telegraph ran a piece entitled "Neil Ferguson's Imperial model could be the most devastating software mistake of all time", quoting a the boss of a software firm asking why the government failed to get a second opinion.

Advisers advise, ministers decide. But first the ministers have decided which advisers to appoint. I said on 8 April that I hoped they had chosen the right ones. I'm not convinced there is evidence that they have had "lucky generals" in this war.

I am not (yet) claiming that lockdown in the UK was unnecessary. After all, the loading on intensive care facilities was ramping up alarmingly until it mercifully peaked leaving some spare capacity, much of it created by repurposing NHS facilities in great haste. Indeed I suspect the news coverage from Italy was as important as Ferguson's suspiciously round numbers of a quarter and half a million. Which is why I suspect he frigged the input data, though by the sound of it he could have just re-run the model time after time with the same data until it randomly gave him the large enough numbers he was seeking. But I am wondering why the curve appears to be slowly flattening everywhere despite differences in policies between countries. David Smith noted a possible explanation given in a paper published by Prof David Miles and Oscar Dimdore-Miles. Yes, related, father and son. David is a professor of financial economics and was once a member of the Bank of England monetary policy committee. His son is an "Oxford scientist". I looked him up and he's a graduate physics student specialising in environmental science so I'm not sure how qualified this makes them to use a so-called SIR (susceptible-infected-recovered) disease spreading model, which they coupled with actual data on numbers of cases and presumably parameterisation (otherwise known as guesswork) on how many asymptomatic cases there might be.

I looked up the Miles duo's paper. It is pretty technical with lots of equations I might once have understood better than the words. It outlines a 'simple' method for estimating the spread of covid-19 in the absence of large scale random testing.They applied the model to the UK and other countries and found that, to get the results to match actual data on daily new cases of the virus, they needed to have high numbers of infected, asymptomatic people. The results were very sensitive to whether the transmission rate of the virus is different for symptomatic and asymptomatic cases, which is uncertain. Nevertheless they postulate that the infection may have spread far enough to mean that the trajectory of falling new cases could be maintained with some easing of restrictions. In other words there may have already been enough people infected for herd immunity to be a factor, which implies a very much higher number of people infected than most experts think.

Now I think that sounds most unlikely. And yet the virus seems to have peaked and be tailing off just about everywhere, independent of the specific policies adopted by individual governments (though I have seen that there has been an uptick in cases in Florida since policies were relaxed). But if that was the case it would also mean that the precise distancing and lockdown strategies didn't matter that much. So we could have stayed with something like the pre-lockdown social distancing exhortations, or gone for a shorter lockdown, with more people working. This would have avoided the restart issues we now have with the population's poor instinctive grasp of risk causing much concern. If so we've just caused enormous health, financial and emotional damage to the country mainly because of one man and his dodgy model.

I said Ferguson's problem was that it was all in his head. But it might be on his head as well.

Sources included:

Neil Ferguson: UK coronavirus adviser resigns after breaking lockdown rules, The Guardian 5 May 2020 included the Antonia Staats comment about "same household" .

Vanessa Chalmers Mail Online 17 May https://www.dailymail.co.uk/news/article-8327641/Coronavirus-modelling-Professor-Neil-Ferguson-branded-mess-experts.html

Matthew Syed, Fixated on the flu and shrouded in secrecy, Britain's scientists picked the wrong remedy. Sunday Times 17 May

Neil Ferguson profile: the professor who turned the UK's coronavirus response, The Times 6 May

David Smith's Economic Outlook column in the Sunday Times, 17 May.

Assessing the spread of the novel coronavirus in the absence of mass testing. Oscar Dimdore-Miles and David Miles. Covid Economics: vetted and real time papers. Issue 16, 11 May 2020, page 161. Centre for Economic Policy Research, CEPR Press.

No comments:

Post a Comment