Background: Hypoxaemia in the acute phase of stroke might damage the ischaemic penumbra and worsen clinical outcome. We determined the frequency of hypoxaemia on admission with stroke and assessed whether it was related to outcome. Methods: We measured arterial oxygen saturation (SaO(2)) and breathing effort with portable monitoring equipment in a large cohort of acute stroke patients, continually from arrival at hospital, during interdepartment transfer, in imaging and on the ward. Patients received best medical care according to current guidelines. Baseline neurological examination and 3-month outcome (Modified Rankin Scale) were assessed blind to other data. Hypoxaemia was defined as SaO(2) < 90% for >= 10% of each assessment stage. Results: Mean SaO(2) was lowest during transfers (p < 0.01), but hypoxaemia was common in all assessment stages. Patients with hypoxaemia (30/153, 20%) were more likely to have a pre-existing respiratory disease on admission than those without hypoxaemia (p < 0.04). More patients with hypoxaemia (40%) died than those without hypoxaemia (20%) (hazard ratio, 2.0; 95% CI, 1.0 - 4.1), though after adjusting for National Institute of Health Stroke Scale and age this association was not statistically significant (hazard ratio, 1.5; 95% CI, 0.7 - 3.1). There were similar numbers of dependent survivors in both groups. Conclusion: Hypoxaemia during acute stroke assessment was associated with increased risk of death. Although SaO(2) is lower during transfers, hypoxaemia occurs in all stages of the admission process. Further research is necessary to determine whether strategies to avoid hypoxaemia during acute assessment improve stroke survival. Copyright (c) 2006 S. Karger AG, Basel.