The actual number of US deaths caused by severe acute respiratory syndrome coronavirus 2 infection has been investigated and debated since the start of the COVID-19 pandemic. Here, we use machine learning trained on US death certificates from March 2020 to December 2021 to predict 155,536 (95% uncertainty interval: 150,062 to 161,112) unrecognized COVID-19 deaths. This indicates that 19% more COVID-19 deaths occurred in the US than officially reported. Predicted unrecognized COVID-19 deaths occurred disproportionately among decedents with less than a high school education; decedents identified as Hispanic, American Indian, Alaska Native, Asian, and/or Black; counties with lower household incomes and worse preexisting health; and counties in the South. These findings suggest that the US death investigation system undercounted COVID-19 deaths unevenly, hiding the true extent of inequities.
The unrecognized COVID-19 deaths that our machine learning approach predicted were attributed to a variety of underlying causes of death including Alzheimer’s disease and related dementia, cardiovascular disease, and diabetes (table S8)
I wasn’t surprised by Alabama having a big undercount, by the way. Notice New York, as well.
The thing is, for most death certificates, cause of death is not going to come from any deep investigation, autopsy, or anything like that. If you have lung cancer (as in the table S8 up there), if you die in your sleep (for example), the assumption will be the UCD is lung cancer. Even if (perhaps) you had COVID.
COVID is associated with blood clots, and I did see a lot of ischemic heart disease/stroke deaths increase over 2020/2021… which are associated with blood clots. I also thought of other major natural CoDs that I saw increase over that period.
Is Connecticut missing because it’s missing, or because someone didn’t account for the change from historic counties to planning regions as county-equivalents?
Maybe not too surprising there were undercounts. It seems like a combination of perhaps both political bias influencing death certificate coding (as you pointed out, there could be two reasonable codes in many cases) as well as a healthcare system that underserves the poor areas where those who were already unhealthy produced a large number of ambiguous deaths.
Weren’t we (actuaries and other smarty-pans) always attributing deaths-over-expected to COVID because COVID was not being listed as a cause of death? Maybe not auto accidents, but certainly a driver of deaths. (Yes, I chose that word on purpose, because I could not help myself.)
The difference seems to be the next step in differentiating excess deaths not related to those secondary pandemic causes of mortality like suicide or overdoses.
Interesting. Thanks for that. It could be argued that those who died after missing out on treatment for unrelated conditions due to isolation or being unable to get into the hospital could also be included in the count as indirect victims of the pandemic. I had a friend who was unable to go to the doctor for some time and by the time he was diagnosed with cancer it was too late. It’s possible (but hard to prove) that had he been able to get an earlier checkup he may have survived the cancer.
The shutdowns did cause some extra deaths but as with any triage situation, if the total number of deaths was less, than the shutdown was justified. Measuring that and convincing the public is difficult.
I’ll have to dig into why CT is missing, but one needs to remember that CT had data reporting problems in the early years of the pandemic… which led to the mysterious bump in cancer deaths in 2022…..
note: it was a model error, not realizing that CT had fixed its data-reporting problems
I spent some weeks annoying the CDC about their model until they fixed it (and then they dropped doing an IBNR altogether, and then stopped the tracker)
