Digitally enabled acute atrial fibrillation care: safety and cost-analysis in a propensity-matched real-world cohort

Introduction Atrial fibrillation management consumes significant health budget with a majority due to hospital utilisation costs. Digitally enabled care models could provide hospital-level care at home, but evidence in acute care remains limited, and evidence on costs and value remains inconsistent. This study evaluates the safety and NHS provider-side direct costs of an acute AF virtual ward (AFVW) and explores readmission predictors. Methods We implemented a novel AFVW at a UK tertiary centre in January 2022. Two analyses were performed. First, a propensity-matched non-randomised comparator study with standard of care (SOC), evaluating a primary composite outcome of non-planned readmissions, emergency department or urgent care attendance, or death within 90 days. A cost-analysis was undertaken from the NHS provider perspective using patient-level and spell-level costing data. Total cost and cost per day were analysed using Gamma generalized linear models with log link, reported in 2022 GBP. Second, a long-term real-world cohort analysis (Jan 2022 - Nov 2025) examining readmission and length-of-stay (LoS) predictors using regression modelling. Results Ninety-one AFVW patients were matched to 91 SOC patients. Matching included 49 baseline variables; absolute standardised differences were <0.1 for most variables. The SOC group demonstrated a higher risk of the composite outcome at 90 days, yet not statistically significant (hazard ratio (HR) 1.63, 95% CI 0.91 to 2.90). Thirty-day readmissions, emergency attendances, outpatient appointments, and mortality were similar across groups. Emergency bed-day use following readmission was higher in the AFVW cohort (0.42 vs 0.07 bed days per patient, ratio 6.04; 95% CI 2.70 to 15.90), although absolute utilisation remained low. Cost analysis included 329 SOC admissions and 75 VW patients (8 had more than one care episode). Mean total cost was higher in SOC than VW (£1,561.57 vs £947.85). The unadjusted mean cost difference favoured VW by £614 (95% CI £393 to £841), bootstrap estimates were consistent. In a real-world cohort of 742 AFVW patients, the HR on discharge from the virtual ward was independently associated with re-admission (adjusted odds ratio (OR) 1.1, 95% CI 1 – 1.1, P<0.001), whilst medication uptitrations and diabetes were associated with a longer LoS. Mean length of stay decreased from 20 to 15 days over time. Conclusion An acute AF virtual ward model delivered comparable safety outcomes to SOC while reducing direct provider costs. Longer, multicentre, prospective studies are needed to facilitate wider implementation.