Recently, two papers on the death toll from COVID have gotten widespread attention. The first was a CDC update on their “excess death” analysis that found there were about 300,000 more fatalities in 2020 through October 3 than expected, but at that time only about two-thirds of that number had been attributed to COVID. The second was a paper and podcast by Genevieve Briand, a Johns Hopkins professor, in which she argued that the effect of COVID on overall fatalities had been exaggerated. Johns Hopkins later repudiated her study and took it off their website. Nonetheless, her views have been widely circulated in the blogosphere.
In early September, I wrote on the excess death issue in this blog post. I thought in the wake of these reports, it would be a good time to update the excess death analysis.
How Many Excess Deaths in 2020?
First, I agree with the CDC’s basic numbers. The US is probably going to have about 350,000 more fatalities in 2020 than 2019. That will be about a 12% increase. The average annual increase in recent years has been about 1.5%, due primarily to population growth. The highest annual increase in recent years was 3% in 2017.
I am basing this estimate on the mortality statistics compiled by the National Center for Health Statistics (NCHS). The NCHS totals fatalities from death certificates it receives from states for each week. However, there is a significant lag time, sometimes months, between the date a person dies and the date that fatality shows up in the NCHS database. The NCHS has a column which shows the percentage of expected fatalities it has received. For example, in the most recent report its database shows fatalities through Week 47 (ending November 21). The database indicates that NCHS has received 100% of expected death certificates through Week 45 (ending November 7). However, in observing the database over the last year, I have seen very substantial upward revisions well after the data NCHS shows a particular week as complete. This reporting delay has led some who have looked at NCHS’s website to incorrectly conclude that fatalities for 2020 have not increased.
The data that NCHS publishes on its website goes back to 2013. The earliest full year’s data is 2014. I have entered the data since 2014 in this spreadsheet to facilitate the comparison of this year’s fatalities with previous years.1 Through Week 45 (the most recent week for which NCHS shows complete), there had been 2,767,341 fatalities this year, compared to 2,445,000 last year – an increase of 323,000.
How many more will be added between now and when we have a final count for 2020 is dependent on how bad the current surge gets and how far behind the states are in getting their death certificates to the NCHS. One silver lining is that the number of excess deaths has been trending down, reaching single digits in recent weeks.
Have COVID Fatalities Been Under or Over Reported?
Through Week 45, the CDC’s official count for COVID fatalities was 236,000. So, the CDC study found, only about 70% of the excess fatalities in 2020 have thus far been attributed to COVID. The CDC study raises the possibility that COVID fatalities have been under reported and implies that most of the excess deaths are “from” COVID.
While there were undoubtedly some undiagnosed cases in the early days, that seems much less likely recently with the current level of testing. And we know that there have been some collateral fatalities from our reaction to the pandemic.
I think it would be a serious error to lump deaths medically related to COVID to such things as an increase in suicides. We need to be able to attempt to sort out fatalities primarily caused by our policy reaction to the pandemic rather than the disease itself. Otherwise, there will be no way to do any kind of cost-benefit analysis of those policy responses. But in any event, it will be a long time before we can sort out whether COVID deaths have been under or over reported. In fact, we may never be able to do that with any exactness.
Professor Briand’s Presentation
This bring us to the kerfuffle over Professor Briand’s presentation. Briand basically makes two points in her presentation. First, the historical distribution of fatalities across age groups has not changed significantly in 2020. This is an important insight because there has been so much media attention that seniors make up a disproportionate share of COVID fatalities. However, that is the case with most causes of death. Older individuals are similarly at an increased risk for cancer, heart disease, etc. Fatalities in the COVID pandemic have generally followed the same age distribution as we saw pre-pandemic.
Second, Briand notes there has been a reduction in other causes of death, such as heart attacks, in 2020. She concludes that COVID fatalities have been overstated because people did not stop dying from other causes. Again, it is a fair point and an important one to consider as we try to understand the interaction of COVID with comorbidities.
But unfortunately, Briand went on to make some general statements in which she seemed to say that there had been no increase to total fatalities in 2020. That is clearly not accurate. English does not appear to be her native language and I think there may have been some confusion about what she intended to say. In any event, while her work provides some interesting insights, it does not disprove the basic proposition that fatalities are going to be up by about 10-15% this year.
“Years of Life Lost” & “Mortality Displacement”
There are two factors which will mitigate COVID’s ultimate fatality toll.
The first involves the concept of “mortality displacement.” Mortality displacement is an epidemiological/demographic concept that describes the phenomenon where a particular event (e.g., a pandemic, heat wave, natural disaster, etc.) moves forward fatalities that would otherwise have occurred later in time. For example, suppose a terminal cancer patient had six months to live but became infected with COVID and died in three months instead. Absent COVID, the fatality that would have been recorded in month 6 is instead recorded in month 3 and thereby reduces the fatalities reported in month 6.
Because the vast majority of COVID victims have been elderly and ill, there will likely be a significant displacement of fatalities, including many that would have occurred in 2021, to 2020. This is particularly likely to be true with many of the fatalities in nursing homes where the average life expectancy is less than two years.
It can also refer to the circumstance when fatalities have been unusually low and as a result there is a larger than normal portion of the population that is at risk of dying. This may also be a factor in COVID fatalities because the increase in fatalities in 2019 was only .3%. That was the smallest in recent years and well below the average of 1.5%.
Second, epidemiologists and actuaries distinguish between “lives lost” and “years of life lost.” Years of life lost is a calculation which compares the age of the decedent to their life expectancy. Again, because COVID fatalities have been very concentrated among the elderly with comorbidities, its toll in terms of years of life lost will probably be relatively less than other pandemics. Most pandemics, and especially flu pandemics, tend to have high fatality rate among children, as well as the elderly.2
Unfortunately, we are not going to have a good handle on the actual count of excess fatalities for 2020 until well into next year because of the reporting delays. And until we get the final numbers for 2021, we will not know the degree to which mortality displacement might mitigate the ultimate mortality of COVID over the entire pandemic. But there is no question that there has been a significant increase in mortality associated with COVID.
We will be sifting through the data for years trying to understand questions like the interaction between COVID and comorbidities, how the “years of life lost” will compare to other pandemics, and the long-term effects of our policy responses on things like including public health, educational attainment and families’ financial well-being. It may be emotionally satisfying to reach for simple explanations on the impact of COVID, especially as confirmation bias for our preconceived notions. But this virus is anything but simple. Rather it is mind-bogglingly complex, and we will be sorting out its impact out for many years.
Note 1 – In the spreadsheet, I used the weekly fatality totals from the CDC’s excess death analysis for 2020 and not those currently reflected on NCHS system because they are slightly more current.
Note 2 – There have been some studies that have estimated the average years of life lost in the range of 12 years per COVID fatality. Those studies are clearly flawed. They used life insurance mortality tables to calculate the COVID victims remaining life expectancy. However, those tables are based on the average for the entire population. The COVID fatalities are concentrated in the portion of the population with comorbidities. For example, the average life expectancy for someone who is currently 80 years old is about 9 years. But if that person is a resident at a skilled nursing facility it is only about 2 years. The fact over 40% of the COVID fatalities have been nursing homes residents and 94% of COVID victims have had one or more comorbidity are indications of the degree to which COVID has mostly preyed on people that were already in overall very poor health.