Make America well again

Introduction

Every year, at about this time, some of us look forward to the NFL (National Football League) playoffs, and the culmination of the NFL season at the Super Bowl in the early New Year, this time on February 7th, 2021.

As it does every four years, the playoff season coincides with the change of Presidential term in the USA, as, this time, Donald Trump (eventually) hands over (or, if he refuses to concede, has it handed over for him) to Joe Biden.

I will look at the impact of a postulated change in the timing of further measures to control the Covid-19 epidemic in the USA, against a current background of rapidly increasing daily case numbers and deaths.

This is also happening elsewhere, including in the UK, and I will show a forward look of my model for the UK too; but for the USA I will run two comparative projections.

I will also show the current Worldometers forecasts for the two countries, which provide useful internal comparisons as well as with my model outcomes.

Modelling tools

My model was originally developed by Prof. Alex de Visscher at Concordia University in Montreal, and since I have started using it, it has been updated a) to handle interventions to the epidemic on specific dates, and b) as I reported in my previous post on October 29th, I have also updated it to handle adaptively triggered interventions, setting upper and lower thresholds for the numbers of Seriously Sick people in the model, to turn interventions on and off, respectively. In this particular post I have set those automatic interventions to zero, since I have postulated specific interventions already.

In my blog post on May 14th, in the first phase of the pandemic, I looked back to March in the UK, and pondered (theoretically) what might have happened if UK lockdown had been two weeks earlier, on March 9th instead of March 23rd. This gives me some useful indicators for handling the two US options below.

I had configured my Coronavirus model to reflect that UK March 9th timing (which was when Italy locked down), and as others reported (including Professor Rowland Kao and his team at Edinburgh University, as well as Imperial College’s Prof. Neil Ferguson in his reported evidence to the UK Science and Technology Parliamentary Select Committee) many lives could and probably would have been saved if that had happened.

There was a degree of hindsight involved, of course, and these were assessments strictly in epidemiological terms (as opposed to consideration of the wider health or economic concerns), but we can use the same approach to ponder the immediate future for the US under two assumptions.

The USA

This time I am interested to explore what the impact of the 70 days or so between now and the inauguration of Joe Biden might be. The two options to model are:

a) Joe Biden’s presumed actions once he takes power, on January 20th, following 70 days of no further measures by the incumbent, in the intervening period, to deal with rapidly increasing numbers of daily infections and deaths in the USA; and

b) an immediate such action (unlikely) in contrast to Donald Trump’s previous policies (or lack of them) to date towards the SARS-Cov-2 epidemic

Joe Biden has already announced his initial Covid-19 advisory team, and one expects that this will mark the end of attempts by some academics, such as the group led by Stanford’s John Ioannidis, and some others to influence Trump policy against various types of interventions; and the reported “Great Barrington Declaration” (GBD), supporting the discredited herd-immunity approach, might also become a casualty; both of these initiatives have been roundly condemned by many other scientists, which I also covered in my blog post on October 29th.

I have adjusted my USA model (which had not been my primary focus before) with an intervention pattern to match the reported daily deaths since early March, as published on Worldometers’ USA page, which my model interrogates each time it is run for any particular country – up to November 10th in this version. Here are my two charts – NB I have made the vertical scales on the charts the same for easy comparison.

In both charts, we see the modelled deaths (in orange) and Seriously Sick (in blue) curves out to Day 800. On the left, we see the outcome for a delay in further action until Presidential inauguration day, January 20th, and on the right, the outcome for (otherwise the same) immediate action.

Modelled deaths match the reported data (the heavier orange curve) up to November 10th in both charts, and you can see that reported and modelled curves are virtually coincident up to that date. This is achieved a) by the model differential equations describing the behaviour of the epidemic, and b) the application of intervention settings to mirror the changes in the reported case and death rates.

The charts project to 11th April 2022 (with 437,649 deaths on the left, and 304,032 deaths on the right). But taking February 1st 2021 (Day 365 for the Model which started on Feb 1st 2020) as a comparison date (to match the Worldometers projections below) for the outcome for deaths, we see deaths at a little under 300,000 (actually 291,950) on the right, where an 8% additional intervention effectiveness is added on November 8th (Day 281 of the model).

But on the left, where action is delayed until the Presidential inauguration day, January 20th (model Day 355), the outcome is much higher at a little under 350,000 (actually 347,843).

On the left we see also the resulting second peak in Seriously Sick numbers compared with the chart on the right.

This difference in deaths is about 56,000 (actually 55,893) by February 1st 2021, which is the immediate (modelled) human cost of failing to take such action in that 70/80 day period, the kind of lesson we learned from the UK situation that I reported above.

Although the charts project to 11th April 2022, I would expect that we would be building into the model the effects of an available vaccine well before then, and so this aspect of the forecast will need modification for that later period. But that is unlikely to change the shorter term comparisons substantially.

The curves for deaths become essentially flat in these two modelled cases by about Day 500 or so, with Seriously Sick numbers very low by then. If no vaccine approach is effective by then, the delay between the two models shows a cumulative difference in deaths by Day 500 (June 21st 2021) of just over 130,000. I am sure this wouldn’t be allowed to happen, and that further actions would be taken by a new President more concerned about his nation’s health than his own interests.

What a difference we have seen in the last four years from the historic Kennedy exhortation: “Think not what your country can do for you, but what you can do for your country”. One would think that a President, above all others, would have that in mind.

In the growth situation we see for the epidemic (both now and, of course, also in the Spring 2020 first phase of the outbreak) we see that decisive, but also early action is necessary; otherwise early small differences become very large ones very quickly in an exponential (or even in a somewhat sub-exponential) situation.

Worldometers forecast for the USA

Worldometers US page links to the University of Washington’s site for their US projections, and here is the November 8th projection to February 1st 2021 for the USA:

As we can see, their forecast is a little gloomier than mine. At best, they are saying that if mask-wearing became universal (unlikely in the USA in the interim) then there would be 337,670 deaths by February 1st 2021, with a projection of c. 400,000 for the continuation of current policies, whereas my modelled assumptions (on the basis of no further intervention actions until January 20th, and then an increase in intervention effectiveness of 8%) lead to 330,000. But we aren’t too far apart – unfortunately.

United Kingdom model update & projection

I have been reporting on the UK situation for seven months, and so I’ll be brief here with just my own current forecast, and then the Worldometers equivalent.

In the case of my own forecast, I have built in the latest UK intervention effectiveness on the basis of (broadly) partial lockdown from November 6th to December 2nd, as announced for England, paralleled, in different ways, by measures in the other three home countries – a multi-tiered approach in Scotland, a fire-break in Wales (just finishing but with some follow-on actions) and a circuit-breaker in Northern Ireland.

My model doesn’t reflect such fine-grained differences, but employs an overall UK intervention effectiveness percentage to represent them. These intervention effectiveness changes and timings are tabulated in the body of the chart below.

This chart is in the same format as I have shown recently. The interventions listed in the chart show a very good match for my model against reported deaths data until November 8th (again with coincident curves up until the current date). I then add a further 9% intervention effectiveness on November 6th, relaxed by 6% on December 2nd (I don’t think there is likely to be a total removal of extra measures at that time) in order to project forwards.

We see the second peak in the Seriously Sick numbers (reflecting reality) and then declining more slowly, while cumulative deaths grow, but at a declining rate.

It may well be that the UK Government will wish to encourage further measures after December 2nd, but this will, no doubt, be assessed in the light of the effect on other aspects of health (both mental and physical) and also on the economy, together with the assessed endurance of the population to continue to comply with the measures requested, especially in what (somewhat hopefully) we might call the festive season.

Worldometers forecast for the UK

Such hopes might be optimistic, and indeed my model’s projected outcomes are more optimistic than the University of Washington projections for the UK, which are shown below, and, as for my charts, draw reported data from the Worldometers UK page.

Here we see that the February 1st deaths projection for universal mask wearing (which is probably a more realistic expectation for the UK public) is c. 105,000, with the forecast for current policies (but, I suspect, not including current lockdowns) at 140,000.

My own forecast for that date is nearer to 75,000 deaths (actually 74,275), but, as stated, I have built in the additional month of partial lockdown measures announced from November 6th until December 2nd, with an assumption that there will be a somewhat reduced level of continuation after that, but still above what we have been used to until November 6th.

I expect these new measures to have a beneficial effect.

Summary thoughts

This post has been mainly about the US situation, with so much desired change to what has been disastrous management of the epidemic in the US.

In that sense this post is somewhat polemical, but after 7 months of modelling the pandemic in the UK, monitoring and learning from some excellent scientists – Adam Kucharski, Neil Ferguson and some others – and being aware of lesser and somewhat contrary contributions on Twitter and elsewhere, I presume to have a useful opinion informed by that experience.

Scientists do differ in the detail, and in some cases in some of the fundamentals of analysis of the epidemic. But, unfortunately, some are coming at this with other, less scientific motivations, whatever they are – fame, fortune, political influence or simply hubris.

In my case, I know that I am probably writing for an audience of one (me), and that my errors will be simply rooted in limitations of my understanding or intellectual capacity. I have no axe to grind.