After I noted a couple of days ago that age-adjusted all cause mortality in the USA rose quite a bit more in the first year of the COVID pandemic than it did in the first — and by far the worst — year of the Spanish flu pandemic in 1918, Andrew Gelman pointed me to this post on his excellent stats blog.
The post is from April, and in it Gelman notes that while age adjusted all-cause mortality rose by a higher percentage in 2020 than it did in 1918, this is a statistical artifact of the way that number is calculated. In short, age adjustment works by layering mortality statistics from a particular time and place onto some standard population model, in order to take into account the fact that populations vary enormously in terms of their demographic contours.
For example, the crude mortality rate in Italy and Japan — that is, the percentage of a population that dies in a year — is nearly double that in Bangladesh, even though people in the former countries have much better life expectancies and overall health than those in the latter. A much higher percentage of Italians and Japanese than Bangladeshis die in any year, however, because the populations of the former countries are much older than that of the latter, and of course age is almost always the most important variable in population-level mortality statistics. Adjusting crude mortality statistics for age differences in populations attempts to correct for this factor.
In the case of the COVID and Spanish flu pandemics in the USA, two facts produce a much smaller rise in age adjusted mortality than in crude mortality in 1918: The 1918 population was much younger, and the Spanish flu epidemic killed teenagers and people in their 20s and 30s far more often than any other part of the population. (In fact in 1918 mortality rates actually dropped among everyone 55 and up, and barely rose among 45 to 54 year olds).
37.6% more people died in the USA in 1918 relative to 1917, while the official crude mortality rate rose by 29.3% — both far higher numbers than the comparable statistic for 2020 relative to 2019, which was a still plenty appalling 17.8%. Nevertheless, the 2020 number is more than four times higher than any other increase in the annual American mortality rate in the last 100 years, with second place going to 1943, which featured the peak of US combat deaths in WWII. Note that the crude mortality rate in 1918 was affected only a little by the roughly 40,000 American combat deaths that year, which represented only 2.7% of the total deaths in the country’s population that year. More American soldiers died from Spanish flu than in combat. (Side point: I don’t understand why the official crude mortality rate for 1918 is lower than the percentage increase in total deaths, given that the nation’s official population for statistical purposes seems to have been almost identical in both years, in part of course because all the excess deaths suppressed the population growth rate.)
So Gelman makes a good point that describing the COVID epidemic in 2020 as worse than the Spanish flu epidemic in 1918 because of the relative effect of each on age adjusted all cause mortality is by itself misleading.
But . . . there’s now a huge caveat to all this, which Gelman couldn’t have anticipated when he posted on this subject last April.
As I’ve noted before, a very curious feature of the 1918-1919 pandemic in the USA is that, even though around a quarter of the estimated total Spanish flu deaths of 675,000 during those two years took place in 1919, all-cause mortality actually declined sharply in the latter year in comparison to the pre-pandemic baseline.
Crude mortality rate per 1000:
There’s surprisingly little — which is to say almost no — discussion in the medical literature of this extremely striking and counter-intuitive outcome. All other things being equal, the crude death rate in the USA in 1919 should have been about 12% to 15% higher than it was in 1916 and 1917, given all the Spanish flu deaths that took place in 1919. Instead it was 7% lower. The net result of this is that the estimated total deaths from the pandemic are almost double the collective total of “excess” deaths in America in 1918 and 1919, if we calculate excess deaths as deaths above and beyond those we would have expected to see if mortality rates had remained the same during those years as the rates during the two years immediately preceding the pandemic.
It seems there are three possible explanations for this result: Either estimates of total Spanish flu deaths are far too high, or the secondary effects of pandemic in regard to public health were so positive that they produced a massive reduction in other causes of death in 1919 in particular, or some combination of the two former explanations.
Which brings us to the truly astounding difference between the Spanish flu and COVID epidemics to this point. What Gelman couldn’t have realistically anticipated in April was that the right wing response to COVID would be so utterly perverse that the total number of excess deaths in the American population is almost certainly going to be higher in 2021 than it was in 2020, even though a completely safe and almost completely effective vaccine began to be distributed just two weeks prior to the end of the latter year, and has been universally available for the majority of 2021!
The USA has seen 830,400 excess deaths since February of last year — and nearly half of those excess deaths have happened in 2021, with more than a quarter of the year remaining. We will almost certainly hit one million excess deaths during the epidemic before 2022 rolls around — a number which represents nearly three times the total excess deaths recorded during the Spanish flu epidemic, i.e, just about as many in strictly per capita terms (The US population is three times larger today than it was 100 years ago).
And of course there’s no realistic prospect that 2021 will be the last year of the COVID epidemic.
It’s true that the 1918-1919 epidemic was especially horrible because it killed the young rather than the old. But, given that we’ve managed to so far produce a similar per capita excess death total as that incurred during the entire Spanish flu pandemic — and we’re still very much counting the bodies piling up during this one — even though we now possess medical-technological weapons that Americans a century ago could have barely imagined, I’m going to give the grisly palm of victory for worst American public health disaster since at least the Civil War to COVID over the Spanish flu.
References: For mortality statistics just before and during the Spanish flu, see here.