A number of OECD countries have introduced waiting time prioritisation policies which give explicit priority to severely ill patients with high marginal disutility of waiting. There is however little empirical evidence on how patients are actually prioritised. We exploit a unique opportunity to investigate this issue using a large national dataset with accurate measures of severity on over 200,000 patients. We link data from a national patient-reported outcome measures survey to administrative data on all patients waiting for a publicly funded hip and knee replacement in England during the years 2009-12. We find that patients suffering the most severe pain and immobility have shorter waits than those suffering the least, by about 29% for hip replacement and 9% for knee replacement, and that the association is approximately linear. These differentials are more closely associated with pain than immobility, and are larger in hospitals with longer average waiting times.