Vaccines
Finding the data was a chore. The older stats are from PDFs provided by the CDC. I started with this one. It covers the top 10 or 15 or 20 causes of death from 1900 to 1998. My first thought was to write a python script to scrape what I was looking for, but it got annoying because the table format changes every few years. And sometimes even when it looks the same, it's different. So I gave up and transcribed it by hand. It's a one-off project, so not a big deal.
That PDF didn't have everything I wanted though, so I kept looking and found this rabbit hole. From there I manually parsed several badly photocopied scans of 100+ year old death stats and combined that with data from relatively modern PDFs. Then I filled in more recent years using CDC Wonder, and finally got everything I needed for this csv which provides the data for the plot.
Many of the older tables showed stats as a rate of total deaths, but fortunately with a combination of documents I always had the rates and totals for either tuberculosis or flu/pneumonia, which I used to derive the totals for everything else. I ended up splitting flu and pneumonia and then removing pneumonia altogether, since I was more interested in the flu (there's no vaccine for pneumonia). They're often grouped together because one leads to the other and it's not always clear which thing killed the person. So I'm not sure how statistically relevant it is. Which brings me to...
Disclaimer
As I've just described, this was pieced together by a rando in Oregon (me). Even if I transcribed everything perfectly (which I did, of course) the data source is not very accurate. There were a couple decades where whooping cough and diphtheria had slightly different rates depending on which PDF I referenced. All of the early 1900s are super inaccurate because not all states participated until 1933. And they didn't count black people at all until 1960 — nineteen-fucking-sixty. So think of this as a rough approximation at best.