Predicting outcomes among patients with atrial fibrillation and heart failure receiving anticoagulation with warfarin

Among patients receiving oral anticoagulation for atrial fibrillation (AF), heart failure (HF) is associated with poor anticoagulation control. However, it is not known which patients with heart failure are at greatest risk of adverse outcomes. We evaluated 62,156 Veterans Health Administration (VA) patients receiving warfarin for AF between 10/1/06-9/30/08 using merged VA-Medicare dataset. We predicted time in therapeutic range (TTR) and rates of adverse events by categorising patients into those with 0, 1, 2, or 3+ of five putative markers of HF severity such as aspartate aminotransferase (AST)> 80 U/l, alkaline phosphatase> 150 U/l, serum sodium< 130 mEq/l, any receipt of metolazone, and any inpatient admission for HF exacerbation. These risk categories predicted TTR: patients without HF (referent) had a mean TTR of 65.0 %, while HF patients with 0, 1, 2, 3 or more markers had mean TTRs of 62.2 %, 57.2 %, 53.5 %, and 50.7 %, respectively (p< 0.001). These categories also discriminated for major haemorrhage well; compared to patients without HF, HF patients with increasing severity had hazard ratios of 1.84, 3.06, 3.52 and 5.14 respectively (p< 0.001). However, although patients with HF had an elevated hazard for bleeding compared to those without HF, these categories did not effectively discriminate risk of ischaemic stroke across HF. In conclusion, we developed a HF severity model using easily available clinical characteristics that performed well to risk-stratify patients with HF who are receiving anticoagulation for AF with regard to major haemorrhage.