The 78 per cent solution

Keep me searching for a heart of gold

Following the recent cardiac arrest suffered by Bronny James, Sports Illustrated discussed messages spread by anti-vaxxers which implied that the COVID vaccines were to blame. I agree that it’s not sensible to draw that conclusion, not least because COVID infections themselves can cause heart problems.

However, the article went on to make a startling claim regarding this level of risk:

A recent JAMA Cardiology study of 100 people folks who had recovered from the coronavirus within 2-3 months of the research being compiled revealed heart abnormalities in 78 recovered patients.

This seems pretty terrifying! It’s likely that most of the world’s population has suffered at least one COVID infection by now, so if over three quarters of them resulted in heart problems, this really would qualify as the “mass disabling event” that Zero COVID proponents regularly warn us of.

On the other hand it’s worth at least thinking a little bit about the claim. Luckily, Sports Illustrated provided a link to the paper, which is published in the journal JAMA Cardiology (and has over a million views and a thousand citations). So the paper exists, it is peer reviewed, it’s in a reputable journal.

But the first clue that something is up should be the use of the word “recent” in the Sports Illustrated quote above. In fact the paper itself was published in July 2020, meaning that it pre-dates vaccines and the arrival of omicron, and hence the estimates that it provides may not tell us much about current risks.

But more than this, we have to think about the people who appear in the sample of COVID infected patients that it studies, and consider whether they are representative of the population at large. While the paper is in theory based around the study of people who were infected in the spring of 2020, instead (for obvious reasons) in practice its sample is restricted to people with a positive COVID test at that time.

We need to remember what the situation was back then: PCR tests were hard to come by, and were generally administered to only the sickest patients (those being admitted to hospital or showing severe symptoms).

I can illustrate this using UK data - by the end of May 2020, the UK COVID dashboard tells us that around 220k people in England had tested positive. However, we know that this grossly underestimates the true number of infections. The MRC nowcasting site is unfortunately not working for me, but by eye (see previous post) it suggests that around 4 million people had been infected with COVID by the same time.

So, in other words only around 5-6% of people with a COVID infection in England would qualify to be in the sample of the JAMA paper. If this were a random subset of the infected people, that would be fine, but in reality we know that this is very unlikely to be the case. In fact, as I have described, it would make much more sense to think about this as something like the 5-6% worst-affected people.

This is borne out by the JAMA paper itself. According to them, a third of their sample had been infected seriously enough to require hospitalization. However, obviously a third of all infections did not lead to hospitalization: again, the dashboard tells us that 65k people had been admitted to hospital by the end of May 2020. Interestingly, a third of 220k is pretty close to 65k (and a third of 4 million is nowhere near that), confirming my sense that this is not a representative sample of infected people.

So, given what we know about Long COVID in the pre-vaccine era (as I have described here, and here an uncomfortably large number of infections led to ongoing symptoms), it’s perhaps not a surprise that the worst-affected COVID sufferers at that stage went on to face serious consequences.

But on the other hand, I don’t believe that it is sensible for Sports Illustrated to quote the 78% figure without any kind of context, or to imply that it is relevant now. Personally, despite the issues of self-reported data, I would be much more comfortable quoting the percentages given by the UK Office for National Statistics to describe risks in the omicron era:

• 4% Long COVID of any severity from first infections (1% for children), falling to 2.4% and 0.6% respectively for re-infections,

• 2.8% Long COVID “limiting daily activities” from first infections (0.6% for children), falling to 1.6% and 0.4% for reinfections.

Of course, this isn’t nothing, but it’s lower than the 10-20% figure quoted by the WHO that many people seem to take as the standard still, and certainly a lot lower than the 78% from the original Sports Illustrated quote. So overall, it feels like a cautionary tale about the difficulties in trying to draw inferences about the general population from an unrepresentative sample - which is something I talk more about in my book Numbercrunch.