May 28 covid-19 update

As I'm sure every American knows by now, the U.S. passed 100,000 deaths from covid-19 this week. It took two months and a day to get from 1,000 deaths (on March 25) to 100,000 deaths (on May 26). That's more than American casualties in the Korean and Vietnam wars (combined) and almost exactly the same number of Americans who died of cancer in the same period. (Approximately 600,000 Americans die of cancer every year.) Like covid-19, cancer can kill young, healthy people but primarily kills the elderly (70% of cancer-related deaths, and 80% of covid-19 deaths, are among people over 65). So maybe it's a useful perspective to think about covid-19 in relationship to cancer. What limitations and inconveniences would we be willing to tolerate if we could stop cancer dead in its tracks? How much would we want our government to spend on eradicating cancer if we knew it was entirely preventable? The U.S. spends more than $5 billion per year on cancer research, although this is dwarfed by the trillions the government is spending in response to the pandemic.

As the U.S. has reopened over the past few weeks (at different speeds and in different ways in different states), we have not yet seen a massive increase in infections or deaths, but neither do we appear to be winning against this virus. I was struck again this week by the disparity in the epidemic across the U.S. I heard from a friend in New York City who knows multiple people who have died from covid-19, including a close relative and a neighbor who left behind a pregnant wife and small child. I also talked to a relative in Montana and a friend in Alaska, two states that have seen minuscule numbers of cases and deaths. As the debate over how and when to re-open continues to rage, we really are living in separate realities. This analysis shows that Democrats and Republicans don't just see the epidemic differently, but are actually encountering different epidemics. As always, the point is that we urgently need more and better data to support strategies that are appropriate to local epidemics. No country has a "national epidemic"; every country is a patchwork of vastly different epidemics and we need targeted, data-driven approaches.

That said, here are some national-level visualizations that show how differently the epidemic is unfolding in Italy, the United States, and South Africa. The red bars shows confirmed cases by day, the blue bars deaths by day, and the orange bars cumulative active cases. (All from https://nssac.bii.virginia.edu/covid-19/dashboard/, and all graphs start from 1000 confirmed cases. I apologize that the labels are cut off, and pay no attention to the dates on the x-axis, as each graph shows data through May 28 despite what the labels say.) Italy is a good example of a severe epidemic that is now being relatively well controlled. Cases per day and deaths per day are down sharply (to less than 1000 cases per day and approximately 100 deaths per day), and active cases have also declined. In the second row, we see that in the U.S. daily deaths and cases have declined somewhat, but active cases are still climbing. Finally, we see that in South Africa all three indicators are still increasing quickly and that the epidemic is just taking off.

Let's take a look at my graphs of cumulative confirmed cases and deaths (plotted on a linear scale, log scale, and by 100,000 population). South Africa has the highest rate of growth of both cases and deaths right now, with both doubling in less than 2 weeks (compared to doubling of cases in approximately 5 weeks in the U.S. and in 8 weeks in Italy). You can especially see the difference between epidemic curves in the log scale graphs (middle row). I've also added case fatality rates (CFRs) back into the graphs (top right graph). I'm calculating CFRs with a 2-week lag (dividing deaths to date by cases 2 weeks ago to account for the lag between diagnosis and death).

Keep in mind that these are fatality rates only among those with confirmed infections, and that CFRs will thus vary with testing rates. Yet I think the quite different CFRs have to do with more than just different testing rates in the three countries. Italy has one of the oldest populations in the world, which is widely recognized as contributing to Italy's high CFR. The U.S. has a younger population and South Africa has a younger population still, which may be why CFR in South Africa is lower than the U.S. or Italy (at 4.7%). This could change, and in fact South Africa's CFR has been increasing steadily over the weeks. But so far South Africans with known infections are at less risk of dying than Americans or Italians, which is good news. Put another way, in terms of fatality from the virus, South Africa's young population may be enough to counteract the fact that so many South Africans have risk factors such as infection with HIV and TB.

Let's talk about Sweden again. About a week ago, Sweden was getting a lot of press for having the highest covid-19 deaths (per capita) in Europe. This was not nearly as alarming to me as the fact that Sweden discovered (through antibody testing) that fewer people were infected than they had thought. This means that Sweden is much further from herd immunity than Swedish experts had assumed, and that claims that Sweden would be approaching herd immunity by June are not accurate. Sadly, there seems to be no charmed middle ground of voluntary social distancing and a relatively intact economy in which a population quickly gets to herd immunity with a minimum of deaths. We don't seem to be close to herd immunity anywhere in the world. In Stockholm, in late April, ~7% of people had been infected with SARS-CoV-2, and in other parts of the country it was only ~4%. (I have not been able to find national-level data.)

Let's do some rough math and estimate that Sweden is only 10% of the way to herd immunity, although it's probably less than that. (Herd immunity requires that ~70% of a population be infected or vaccinated, and it's unlikely that the country as a whole had Stockholm's level of 7% of the population showing antibodies.) I can't find exact dates for when the antibody testing was completed, only that it was in late April, but let's take the cumulative covid-19 death count as of May 10 to account for the lag between infection and death (i.e. give people who tested positive for the virus by late April time to either die or recover). That number of deaths is 3,225, and dividing it by the population of Sweden (10.23 million), gives us the result that 0.032% of Sweden's population had died of covid-19 as of May 10. If we multiply that number by 10, it gives us the likely death rate to reach herd immunity: 0.32% of a population dead.

In the U.S., 0.32% of 331 million people is 1.04 million dead, which is consistent with what I've estimated before (and with other far more sophisticated modeling). In South Africa, 0.32% of 59 million people is 184,000 dead. Of course, this is a rough calculation and there are many things we don't know. Sweden was likely not yet to 10% of herd immunity by the end of April. South Africa may have lower fatality rates as it has a younger populations, although in other countries of the global South young people are dying of covid-19 at much higher rates than in Europe (for reasons as yet unknown), which doesn't bode well for South Africa. South Africa might also have heightened fatality rates among populations with other risk factors (e.g. HIV and TB) and because of fewer healthcare resources. Analysis of data from New York City has estimated the infection fatality rate (IFR) to be 0.86% to 0.93%, which extrapolated to 70% of the population would be nearly twice the estimates we just produced. IFR estimates from Spain, Italy, and France are higher still. (All data from the last two sentenced linked to from this article, which also addresses the CDC's recent low estimate of IFR.) There's so much we don't know - hence the title of this blog, Coronavirus Conundrums.

But I think it's important to bear in mind that until we have a vaccine, the only thing that will prevent us from reaching ~70% of our population infected is the painful, disruptive lockdown and social distancing measures we are going through now. I am not trying to offer answers on exactly what those measures should be. I am trying to remind us what is at stake. If over 1 million Americans dead, or hundreds of thousands South Africans dead, is a greater death toll that we can accept, we have no choice but to slow the virus down until we have a vaccine. We may see in hindsight that many experts and modelers weren't that accurate in their predictions of how the epidemic would unfold. But I don't think they're wrong regarding the bottom line, that ~70% of us will get infected before the epidemic will naturally burn itself out, and that 0.5-1.0% of those of us who become infected will die. That's one in every 200 or 300 people dead. You might do the math regarding how many lives are likely at stake in your town or city, as a reminder of the importance of our actions - and collective action.