April 27, 2021

If you were walking to work, wearing an expensive suit and shoes, and you saw a child drowning in a shallow pond, and you were easily able to reach them and save them — are you morally obliged to do so, even if it would ruin your outfit?

The answer, most of us would say, is “obviously yes”. The value of your nice clothes is trivial compared to the value of a child’s life.

But as the moral philosopher Peter Singer would argue: we are in this situation every day. We could save children’s lives at relatively trivial cost to ourselves, by donating a few hundred dollars to pay for antimalarial treatments or other low-cost interventions. We would be unable to buy as many nice shoes, but we would save lives, and we have just agreed that that is a morally obligatory tradeoff. The only difference is that in the real, existing scenario, we can’t see the children in sub-Saharan Africa or south Asia; and they’re starving, or dying of malaria, rather than drowning. But they’re still real.

Scott Aaronson, a quantum computing scientist at the University of Texas at Austin, extended that metaphor over the weekend. What if you didn’t even have to ruin your shoes? What if you could just throw a lifebelt which you were carrying but had no intention of using yourself?

That is, he said, what is going on now with Covid. The US has millions of doses of vaccines which it’s simply not using. I’ve seen the exact number estimated as high as 100 million, but I think it’s widely accepted that there are 30 million AstraZeneca vaccine doses sitting, bottled but unused, in a warehouse in Ohio. There seems to be no short-term likelihood that the US FDA will approve it, over (to my mind misguided) fears about blood clotting, so they’re just gathering dust.

Since Aaronson wrote his post, there has been some apparent good news: the US says it will start releasing AstraZeneca doses, a total of 60 million. But there’s no sense of urgency. It’s waiting for the Food and Drug Administration (FDA) to complete a “safety review”, which could take weeks, before releasing the first 10 million. Meanwhile, in India, at the very least 2,500 people are dying every day from Covid in an unthinkable, ongoing humanitarian catastrophe. 

I wanted to remind people of the urgency. Imagine that there are 30 million doses sitting in Ohio; how much good could they do if we could get them into Indian arms straight away? So I thought it would be interesting to attempt a Fermi estimate of that, a sort of first approximation. It won’t be exact, but it might get us to within an order of magnitude, and give us a sense of how much using these vaccines matters.

First: every day in India there are about 300,000 confirmed new cases; that number is growing every day. It does not seem to have ended its exponential growth, despite increasingly stringent social distancing measures.

It is also an enormous undercount. A week ago, Max Roser, of Our World in Data, noted that the Institute for Health Metrics and Evaluation (IHME) model estimated that only about one in every 29 cases in India is confirmed with a PCR test.

Back when he pointed this out, there were about 230,000 confirmed daily cases, which he extrapolated to about 6.7 million actual daily cases. Now that it’s well past 300,000, the number is probably more like 10 million. Since then, the IHME has updated its model suggesting that it could be double that, but let’s stick with the 10 million figure.

To reiterate: that’s not the total number of people with the disease. That’s a plausible number of new cases every day.

How many of those people are going to die? Well, the official death toll, as I said, is about 2,500 a day at the moment. But just as with the cases, that is likely to be a severe undercount. As Al Jazeera reports, crematoria seem to be burning far more bodies than the official death toll would suggest: they mention one city that reported 20 deaths, but its Covid-only crematorium has processed 63; another that reported 25 but had cremated 100. The Guardian tells a similar story.

But that’s just the start. Murad Banaji, an Indian mathematician, reports that there were 476 funerals in Kanpur on one day — normally you’d expect about 100 — and yet only three reported deaths from Covid. John Burn-Murdoch of the Financial Times has others: cities are cremating many, many times as many people as have apparently died from Covid (some of his data is taken from Banaji). The IHME model estimates that the number of people dead is about double the reported numbers; Burn-Murdoch thinks it could be much higher, as high as 10 times.

India has a relatively young population, and the risk of death from Covid is hugely affected by age, so the infection fatality rate – the risk of death for someone infected by the virus – was probably quite low, comparatively speaking: perhaps about 0.3%. (This Nature paper says that the IFR in Bangladesh is about that, and since the two countries have a similar age profile, I’m assuming it’s about correct for India.)

But that will all have changed now. The Indian healthcare system is overwhelmed; there isn’t any oxygen to keep patients alive. This news report suggests about 40% of ICU patients in India were dying back in September: if we imagine that all of the 0.3% of infected people who died went through ICU first (which isn’t true, but let’s imagine it), then that means about 0.75% of Indian people infected with Covid ended up in ICU (0.3 is 40% of 0.75).

But if you end up needing ICU in India now, you’ll almost certainly die: the things you’d have gone into ICU for — mechanical ventilation, oxygen, close medical attention — aren’t available. There just isn’t the capacity to treat you. The healthcare system has collapsed. ICU treatment was, in most cases, what was keeping the 60% who didn’t die alive. It’s a good bet that the IFR now is up much closer to 0.75%.

Let’s imagine, now, that we can get those 30 million vaccines into people’s arms as soon as possible. What good might they do 1?

First, to keep things simple, I’m going to assume that the social distancing measures stabilise the Indian outbreak at its current level of about 10 million new cases a day. There was a plausible estimate that about 30% of Indians – 300 million people! – had already had the disease in February, and about 10% have had a vaccine, so it’s reasonable to think that about 40% have some sort of immunity now. At some point that will start slowing down the outbreak significantly, but I’m going to ignore that, again to keep things simple.

Second, I’m going to assume that the vaccine prevents about 90% of deaths. It’s usually more than that, but there are new variants going around India, against which the Ox/AZ vaccine may be a bit less efficacious. Prof Rupert Beale, the head of the Cell Biology of Infection lab at the Crick Institute, thinks 90% is probably a reasonable guess, and Banaji thinks that vaccines are already having an effect.

And I’m going to assume that it takes about three weeks after the first dose for the vaccine to have a full effect.

So if you gave 30 million immunologically naive people the vaccine tomorrow, what would happen? Over the next three weeks, given that 1% of the population is being infected every day, about 20% of them will probably get infected, and 0.75% of them will die. That’s 45,000 deaths.

But after that, things will start to improve. Over the following two months, given our simplified numbers, about 13 million more of them would have caught Covid, and about 100,000 would have died. But, because they’re all vaccinated, all but 10,000 will, in fact, live.

So given these very plausible, or even conservative, estimates, the 30 million AstraZeneca vaccines sitting in an Ohio warehouse could save 90,000 lives in the next three months. Of course, it’s not plausible that they could be put in arms tomorrow – but, then, they also wouldn’t be distributed at random; they would, you’d hope, be given to the most vulnerable people, people with an expected infection fatality rate much higher than 0.75%. The Indian vaccination programme may well not be as well-targeted to at-risk groups as the JCVI-led UK programme, but if it were given entirely to, say, the over-65s, then it might save five times as many people. All of these are estimates, of course, but I think that it’s extremely reasonable to think that it’d be something like this.

And there’s something else to consider. As well as directly protecting people, it will break chains of infection; each infection prevented by vaccination will prevent some number of future cases. Working out even roughly how many would need proper scientific modelling, but as a sort of thought experiment, I checked and saw that the Indian vaccination programme is currently managing about two million doses a day. It was higher, before — about three million — so let’s go with that.

If we naively assume that the 30 million doses simply push the programme on by a proportionate amount, it would mean that the fight against the epidemic is accelerated by 10 days. Again, if I naively think “10 fewer days like today”, then that’s about 25,000 fewer deaths. That’s probably not a very sensible way to think about it — God only knows what the Indian epidemic will be doing in a month or two months’ time — but different models have the daily deaths peaking from 6,000 to 40,000 some time between now and the summer. I don’t think that ballpark figure is unrealistic.

Maybe I’ve got my maths wrong in all of this; maybe I’m wrong by an order of magnitude. Maybe they’ll only save about 10,000 lives. You could still save 10,000 lives for the cost of sending some vaccines you aren’t going to use to another country.

As I said above, there’s been some good news. The USA is going to release 60 million doses as they become available. But it’s not clear that they’re going to go to India – some, at least, are earmarked for Mexico and Canada – and, bafflingly, they have to undergo safety review. Every hour counts, but it will take weeks, at least, for the FDA to carry out this review; weeks in which thousands of Indian people die every day. India has already approved the Ox/AZ vaccine; it has its own experts, and they have assessed the risks and benefits, and they don’t need the FDA to check it for them. This maddening, patronising, counterproductive safety-first approach will cost thousands of lives. Simply get every dose you have on refrigerated aircraft right now and ship them to India as fast as you can.

And it’s not just the USA. Other countries have supplies of vaccines that could do good. Denmark, I think, has about 50,000 unused doses of Ox/AZ which it won’t use. I imagine that lots of other European countries have an oversupply. But the US millions are the overwhelming bulk of the issue.

When Aaronson wrote his blog post, the USA was standing, feet dry, by the pond, refusing to throw their life-belt to save thousands of drowning people. Now, at least, it looks as though they might throw it, after a suitable period of checking that it is adequately buoyant and is painted a regulation shade of orange. But the complete lack of urgency is probably going to cost thousands of lives. For pity’s sake, if you’re not going to use them, get them on a boat and send them somewhere that will.

  1.  Thanks to Adam Kucharski, Rupert Beale, Max Roser and Saloni Dattani for help with figures