Abstract 25: Predicting Outcomes among Patients with Heart Failure Receiving Anticoagulation with Warfarin

Abstract

Background: Many patients receiving oral anticoagulation for atrial fibrillation (AF) have heart failure (CHF). CHF predicts a greater risk of stroke and relatively poor anticoagulation control. However, it is not known which patients with CHF are at greatest risk of poor anticoagulation control or bleeding complications. Therefore, we risk-stratified the patients with CHF who are receiving warfarin for AF according to their risk of experiencing poor anticoagulation control and major hemorrhage using data available in the Veterans Health Administration (VA) Clinical Data Warehouse. Methods and results: We examined 66,099 subjects who had been anticoagulated for stroke prevention in AF for at least 6 months between 10/1/06-9/30/08. Based on ICD-9 codes, patients were divided between those who carried a diagnosis of CHF and those who did not. Among patients with CHF, we examined various predictors of percent time in therapeutic range (TTR), a measure of anticoagulation control. The predictors were chosen by the principle that more severe CHF would cause hepatic congestion, leading to a greater variability of anticoagulation control and greater risk of adverse events. We identified ten severity markers which predicted worse anticoagulation control (i.e. lower TTR) and chose the variables with the largest adverse impact on TTR such as aspartate transaminase (AST) >80 IU/L, alkaline phosphatase >150 IU/L, sodium <130 mEq/L, any receipt of metolazone, and any inpatient admissions for CHF exacerbation. We then developed a composite score of heart failure severity with CHF with no criteria, CHF with 1 criterion, CHF with 2 criteria, and CHF with 3 or more criteria. This composite score performed well to predict TTR; patients without CHF (referent) had a mean TTR of 65%, while groups of CHF patients with increasing severity had a mean TTR of 62%, 57%, 53%, and 50%, respectively (p < 0.001 for difference among groups). In addition, the composite CHF severity score also predicted the age-adjusted hazard of major hemorrhage. Compared to patients without CHF, patients with CHF of increasing severity had hazard ratios of 1.84, 3.04, 3.65, and 5.13 respectively (all p < 0.001 compared to no CHF). Conclusion: We developed a composite score for CHF severity which predicted both anticoagulation control and the rate of major hemorrhage among patients anticoagulated with warfarin. This study suggests that relatively easily observable characteristics can be used to risk-stratify patients with CHF who are receiving anticoagulation for AF.