OBJECTIVE: We examined cognitive function in older hospitalised patients with chronic atrial fibrillation (AF). DESIGN: A prospective substudy of a multicentre randomised trial of an AF-specific disease management intervention (the Standard versus Atrial Fibrillation spEcific managemenT studY; SAFETY). SETTING: Three tertiary referral hospitals within Australia. PATIENTS: A total of 260 patients with chronic AF: mean age 72±11 years, 53% men, mean CHA2DS2-VASc score 4±2. INTERVENTIONS: Cognitive function was assessed at baseline (during inpatient stay) using the Montreal Cognitive Assessment (MoCA). MAIN OUTCOME MEASURES: The extent of mild cognitive impairment (MCI-defined as a MoCA score <26) in AF patients and identification of independent predictors of MCI. RESULTS: Overall, 169 patients (65%, 95% CI 59% to 71%) were found to have MCI at baseline (mean MoCA score 21±3). Multiple deficits in cognitive domains were identified, most notably in executive functioning, visuospatial abilities and short-term memory. Predictors of MCI (age and sex-adjusted) were lower education level (technical/trade school level OR 6.00, 95% CI 2.07 to 17.42; <8 years school education OR 5.29, 95% CI 1.95 to 14.36 vs 8-13 years), higher CHA2DS2-VASc score (OR 1.46, 95% CI 1.23 to 1.74) and prescribed digoxin (OR 2.19, 95% CI 1.17 to 4.10). CONCLUSIONS: MCI is highly prevalent amongst typically older high-risk patients hospitalised with AF. Routine assessment of cognitive function with adjustment of clinical management is indicated for this patient group.