> Future historians will wonder why various medical records show people with acute idiopathic hypothermia in the middle of July 2020.

You underestimate how many people in the database (from any time period) will have a recorded body temperature of 0, thanks to weak typing and buggy null coalescence.

if it's worth anything, as someone who's worked with lots of manually entered data, i expect to see 0.5% to 5% errors in most publicly sourced data collected in this fashion. i guess what I'm saying is, everyone actually professionally working with such on good faith knows its also stacked with a good proportion of bullshit/error.