This paper explores the relationship between health risks from COVID-19 and UK labour supply, using pre-existing conditions as a source of variation in COVID-19 health risk. We find that those with pre-existing conditions were less likely to work during the pandemic after controlling for a rich set of covariates, including labour supplied pre-pandemic, but only when remote work was unavailable. This relationship begins by April 2020, persists through to September 2021, and shows signs of fading after COVID-19 risks had fallen in 2022. Our results are strong enough to explain a 1–1.5 percentage point drop in employment during the pandemic. Placebo tests confirm that our estimates do not reflect labour demand shocks, and that a negative relationship between pre-existing conditions and labour supplied, conditional upon the covariates, did not exist pre-pandemic.