One of the issues in medical data, brought into sharp relief by the pandemic, is what is written on death certificates. For a while, the presumption of a COVID-19 infection without a positive culture was enough to get it listed as a cause of death –- possibly creating a bit of an over-count. But it is not just COVID-19 that is problematic.
The table seen below this paragraph was taken from a poster [1] at the European and International Congress on Obesity, which shows the number of individuals with documented obesity where this diagnosis was and, more importantly, was not included among the causes of death.
The “undercounting” makes obesity “a deadly illness, no one dies from.” I mention this to create a more significant point as we rush to Big Data and artificial intelligence systems in medicine. These systems are predicated on the quality of the data, garbage in, garbage out. (GIGO) And for most electronic medical records, there are few standards permitting machine learning systems to roam through the data without first making sure it is consistent and accurate.
For those who want to read the abstract, here is the link. And here is the link to the larger poster.
[1] Posters are frequently used to highlight new information at these large meetings. News that doesn’t necessarily warrant a 20-minute presentation. While the selection committee of the conference has reviewed it, it is not “peer-reviewed” in a sense the general public would understand.
