Predicting common cardiovascular and non-cardiovascular outcomes after new-onset atrial fibrillation: an analysis from a UK nationwide primary-secondary care linkage dataset.

Aims: Atrial fibrillation (AF) increases the risk of major cardiovascular and neurological complications. Utilization of risk schemes for identifying patients at risk for stroke has improved outcomes for patients with AF. Objectives: Using linked UK primary and secondary care data, we identified risk predictors and develop risk models for heart failure (HF) hospitalization, myocardial infarction (MI) hospitalization, vascular dementia, sepsis, sudden cardiac death (SCD), and all-cause death in the AF population. Methods and results: We included 198 995 patients diagnosed with AF between January 1998 and May 2016. We developed multivariable Cox regression models to predict outcome incidence after AF diagnosis, using 70% of the data for derivation and 30% for validation. Model performance was assessed using Harrell's c-index. Predictor variables were selected based on their availability in routine clinical practice. Among the assessed predictors, excessive alcohol intake, HF, MI, sepsis, and stroke showed the strongest associations with adverse outcomes. The prediction models demonstrated fair to moderate discrimination, with c-index values of 0.70 (95% CI 0.70-0.71) for HF hospitalization, 0.64 (95% CI 0.63-0.65) for MI hospitalization, 0.76 (95% CI 0.75-0.77) for vascular dementia, 0.66 (95% CI 0.65-0.67) for sepsis, 0.65 (95% CI 0.62-0.68) for SCD, and 0.71 (95% CI 0.71-0.71) for all-cause death. Conclusion: In patients with AF, we developed prediction models that demonstrate fair to good performance. Our models, based on variables easily obtainable in routine clinical care, may facilitate patient risk stratification and guide further diagnostic or preventive interventions.