Excess Mortality Studies

As Dave Dixon’s work on this topic proceeds, also informative to look at the general research.  Here is yet another study talking about methods in excess death research.  The determination that there even are “excess” deaths can be tricky.   For example, imagine that every flu death was converted to a CV-19 death during the epidemic, and those were the only CV-19 deaths.  We would have lots of CV-19 deaths, but no excess deaths, at least due to CV-19. At any point of time, a lot of things are happening that affect the absolute number of deaths.  The population grows, or shrinks, the age structure of the population changes, the treatments available for some diseases changes, social issues, like drug and alcohol use, suicides, homicides, change.  So many factors can influence death numbers and death rates.  It is complex to say the least, to identify a trend over multiple years.   (Medrxiv Paper)

In this paper from Germany, the researchers used a very long look back at deaths to ensure that they were creating an appropriate baseline.  As we have noted before and these researchers point out, determining what would have happened with deaths in a future particular time period is difficult, particularly when a large event is currently affecting those death numbers.  The best we can probably do is look at a previous time periods and variability to project forward and say here is our best guess of where deaths would be in this future time period based on those previous time periods.  The epidemic roughly began to cause deaths beginning in 2020, so the specific form of the question here is “based on what was happening with deaths in the X years before 2020, what would we have expected deaths to be in 2020, 2021 and 2022.”  Then we can look at the actual number and see if there was an “excess” or potentially a shortfall.  And because a significant event–the CV-19 epidemic–is prompting this examination, we need to see how much of the excess is due to that cause, and what may have happened to other causes in this period.

Whatever method is used to determine projected mortality and then compare it to actual, can be tested by doing an analysis of the method in past periods.  For example, you could take mortality from 2000 to 2010 as a baseline and see how your projection of deaths in 2011 and 2012 compared to reality.  This sort of methods testing is obviously important, as it may alert you to a flaw in the method which can potentially be fixed.  These authors express surprise at how seldom this reality testing has been done in all the current excess deaths research.

For European countries, which is the focus of this paper, five years of pre-epidemic death data is often used to establish the baseline.  The authors point out that this uses 2015, a year of significant flu deaths, and 2019, a year of very low mortality, and in total results in a baseline that is probably too low, meaning excess deaths are calculated too high.  So these authors looked at mortality in 31 European countries for the period from 1965 to 2019 to establish their baseline and average annual variation.  They projected forward to 2020 and 2021 and compared that projection to actual deaths.

The big picture finding is that most estimates of excess mortality during the epidemic greatly overstate its presence.  According to these researchers’ analysis, in ten of the 31 countries mortality in 2020 and 2021 was within the range of past fluctuations and in 21 it exceeded those past fluctuations, indicating actual excess deaths.  Deaths inherently fluctuate more on a percent basis in a country with a smaller population than a larger one, so that should be taken into account as well.  Minnesota has a relatively small population so we should expect somewhat large annual fluctuations. An interesting side finding is that higher excess mortality is clearly correlated with the wealth of a country, likely because wealthier countries spend more on health care resources which are then available to treat CV-19 and on health care, meaning their populations are in generally better health.

A final word on this lengthy and somewhat technical post.  Going forward for at least the next five years, I think it will extremely important to monitor death trends by age and by cause to see what lasting impacts may have occurred.  Were there substantial pull-forward deaths?  Did CV-19 cause lasting health problems which may result in an earlier death?  And of course my favorite, for which I confess to a bias, did the epidemic response cause excess deaths in certain age groups and for certain causes.