BACKGROUND: Patients are increasingly being admitted with chronic atrial fibrillation, and disease-specific management might reduce recurrent admissions and prolong survival. However, evidence is scant to support the application of this therapeutic approach. We aimed to assess SAFETY--a management strategy that is specific to atrial fibrillation. METHODS: We did a pragmatic, multicentre, randomised controlled trial in patients admitted with chronic, non-valvular atrial fibrillation (but not heart failure). Patients were recruited from three tertiary referral hospitals in Australia. 335 participants were randomly assigned by computer-generated schedule (stratified for rhythm or rate control) to either standard management (n=167) or the SAFETY intervention (n=168). Standard management consisted of routine primary care and hospital outpatient follow-up. The SAFETY intervention comprised a home visit and Holter monitoring 7-14 days after discharge by a cardiac nurse with prolonged follow-up and multidisciplinary support as needed. Clinical reviews were undertaken at 12 and 24 months (minimum follow-up). Coprimary outcomes were death or unplanned readmission (both all-cause), measured as event-free survival and the proportion of actual versus maximum days alive and out of hospital. Analyses were done on an intention-to-treat basis. The trial is registered with the Australian New Zealand Clinical Trials Registry (ANZCTRN 12610000221055). FINDINGS: During median follow-up of 905 days (IQR 773-1050), 49 people died and 987 unplanned admissions were recorded (totalling 5530 days in hospital). 127 (76%) patients assigned to the SAFETY intervention died or had an unplanned readmission (median event-free survival 183 days [IQR 116-409]) and 137 (82%) people allocated standard management achieved a coprimary outcome (199 days [116-249]; hazard ratio 0·97, 95% CI 0·76-1·23; p=0·851). Patients assigned to the SAFETY intervention had 99·5% maximum event-free days (95% CI 99·3-99·7), equating to a median of 900 (IQR 767-1025) of 937 maximum days alive and out of hospital. By comparison, those allocated to standard management had 99·2% (95% CI 98·8-99·4) maximum event-free days, equating to a median of 860 (IQR 752-1047) of 937 maximum days alive and out of hospital (effect size 0·22, 95% CI 0·21-0·23; p=0·039). INTERPRETATION: A post-discharge management programme specific to atrial fibrillation was associated with proportionately more days alive and out of hospital (but not prolonged event-free survival) relative to standard management. Disease-specific management is a possible strategy to improve poor health outcomes in patients admitted with chronic atrial fibrillation. FUNDING: National Health and Medical Research Council of Australia.