BACKGROUND AND PURPOSE: several studies have assessed delirium post-stroke but conflicting results have been obtained. Also, the natural history and outcome of delirium post-stroke need to be fully elucidated. METHODOLOGY: eligible stroke patients were assessed for delirium on admission and for four consecutive weeks using the Confusion Assessment Method (CAM). Risk factors for delirium were recorded. Our outcome measures were length of stay, inpatient mortality and discharge destination. RESULTS: of 110 eligible patients, 82 were recruited over 7 months. Delirium was detected in 23 patients (28%); 21 of these were delirious on their first assessment. Sixty-nine per cent of patients who had four weekly assessments were delirious at 4 weeks. Multivariate logistic regression analysis was performed, and two models were identified. With unsafe swallow in the analysis, delirium was associated with an unsafe swallow on admission (OR 28.4, P<0.001), Barthel score < 10 (OR 32.1, P = 0.004) and poor vision pre-stroke (OR 110.8, P = 0.01). With unsafe swallow removed from the analysis, delirium was associated with an admission C-reactive protein (CRP) > 5 mg/l (OR 10.2, P = 0.009), Barthel score < 10 (OR 46.5, P = 0.001) and poor vision pre-stroke (OR 85.2, P = 0.01). Delirious patients had a higher mortality (30.4% vs. 1.7%, P<0.001), longer length of stay (62.2 vs. 28.9 days, P<0.001) and increased risk of institutionalisation (43.7 vs. 5.2%, OR 14, P<0.001). CONCLUSIONS: delirium is common post-stroke. Most cases develop at stroke onset and remain delirious for an appreciable period. Delirium onset is associated with stroke severity (low admission Barthel), unsafe swallow on admission, poor vision pre-stroke and a raised admission CRP. Delirium is a marker of poor prognosis.