Background: The GARFIELD-AF risk tool was originally developed to predict future risk of adverse events in patients with atrial fibrillation (AF) using a range of baseline clinical variables. In the present work a new, improved risk tool was developed using data from all five GARFIELD cohorts gathered over 2 years' follow-up.

Purpose: To derive a new integrated risk tool for predicting mortality, stroke/systemic embolism (SE), and major bleeding in AF patients up to 2 years after enrolment and compare the risk tool versus CHA2DS2-VASc and HAS-BLED.

Methods: GARFIELD-AF is an international prospective registry of nonvalvular AF patients diagnosed within 6 weeks prior to enrolment and having at least one risk factor for stroke. In this study only the first occurrence of events was considered. Event rates were estimated using a Poisson model. Potential predictors of events including a large set of demographics, clinical characteristics, choice of treatment, and lifestyle factors were identified, and a Cox proportional hazards model chosen for each outcome by least absolute shrinkage and selection operator (LASSO). Indices were compared versus models of CHA2DS2-VASc and HAS-BLED.

Results: Among a total 52,080 patients enrolled 52,032 (male, 55.8%; median age, 71 years) had available follow-up data. At 2 years, 3702 patients had died (event rate, 3.82 [95% CI, 3.70–3.95] per 100 patient-years) whereas non-haemorrhagic stroke/SE was noted in 957 patients (rate, 1.00 [95% CI, 0.94–1.06] per 100 patient-years) and major bleed/haemorrhagic stroke in 673 patients (rate, 0.70 [95% CI, 0.65–0.75] per 100 patient-years). The GARFIELD risk tool outperformed CHA2DS2-VASc and HAS-BLED at predicting all adverse events in the overall population and pre-selected subpopulations over 2 years. Notably, the new model identified use of OAC therapy, which is not included in CHA2DS2-VASc, as one of the strongest predictors of risk of mortality and stroke, and unlike HAS-BLED, could discriminate a lower risk of bleeding in patients treated with NOACs versus VKAs.

Conclusions: The GARFIELD-AF risk tool demonstrated good calibration and discrimination, outperforming CHA2DS2-VASc at predicting risk of death and non-haemorrhagic stroke and HAS-BLED for major bleed in AF patients over 2 years.