EDITOR—The article by Roberts et al on stroke mortality in hospital and at one year after stroke raises several issues.1 The case-mix ratio, presence or absence of on-site neurosurgical services, and stroke unit care should be emphasised as important and easily measured explanatory covariates.

Unfortunately, the most important predictor of 30 day mortality is stroke severity at baseline, and this variable …