Fortunately we have dose maps from Chernobyl and have a rough understanding of biological damage vs. dose. Using conservative models, plus the medical records and interviews (WHO and UN got excellent access a few years after Chernobyl happened), we can build pretty good understandings even if census data was questionable. This was accounted for by UNSCEAR, who says "Up to 4000 may die early from Chernobyl" [1]

[1] https://www.unscear.org/unscear/en/chernobyl.html