Author + information
- Received January 19, 2009
- Revision received March 24, 2009
- Accepted April 3, 2009
- Published online September 29, 2009.
- Jonathan P. Piccini, MD*,
- Adrian F. Hernandez, MD, MHS*,* (, )
- Xin Zhao, PhD*,
- Manesh R. Patel, MD*,
- William R. Lewis, MD†,
- Eric D. Peterson, MD, MPH*,
- Gregg C. Fonarow, MD‡,
- Get With The Guidelines Steering Committee and Hospitals
- ↵*Reprint requests and correspondence:
Dr. Adrian F. Hernandez, Duke Clinical Research Institute, P.O. Box 17969, Durham, North Carolina 27715
Objectives This study sought to examine quality of care and warfarin use at discharge in patients with atrial fibrillation (AF) and heart failure (HF).
Background Atrial fibrillation is common in HF, and national guidelines recommend discharge on warfarin for stroke prophylaxis. However, the frequency and factors associated with the guideline adherence are poorly described.
Methods We analyzed 72,534 HF admissions from January 2005 through March 2008 at 255 hospitals participating in the American Heart Association's Get With The Guidelines HF program. Multivariable logistic regression was used to identify independent factors associated with warfarin use at discharge.
Results In this HF population, 20.5% (n = 14,901) had AF on admission, whereas another 13.7% (n = 9,918) had a prior history of AF but were in a regular rhythm at admission. Contraindications to warfarin therapy were documented in 9.2%. Among eligible HF patients without contraindications, the median prevalence of warfarin therapy at discharge was 64.9% (interquartile range 55.5 to 73.4) and did not improve during the 3.5 years of study. After adjustment, major factors associated with no warfarin use at discharge included increasing age, nonwhite race, anemia, and treatment in the south. Warfarin use also varied inversely with CHADS2(congestive heart failure, hypertension, age >75, diabetes, and prior stroke or transient ischemic attack) risk (70.9% to 59.5% for CHADS2score 1 to 6, p < 0.0001).
Conclusions Guideline-recommended warfarin use in patients with AF and HF is less than optimal, has not improved over time, and varies significantly according to age, race, risk profile, region, and hospital site.
Atrial fibrillation (AF) and heart failure (HF) are 2 prominent and converging epidemics in the U.S. health care system (1,2). After a hospitalization for HF, 1-year mortality exceeds 30% and readmission rates approach 50% (3,4). In patients with HF complicated by AF, these outcomes are even worse (5), with potentially longer-term impact on quality of life and disability because of the enormous risk for stroke. Further complicating the high mortality risk in this growing patient population are the challenges of appropriate treatment, including stroke prophylaxis. Despite the absence of contraindications to warfarin, many eligible patients do not receive oral anticoagulation as recommended in the American College of Cardiology/American Heart Association (ACC/AHA) HF management guidelines (6). In 2005, the ACC/AHA Clinical Performance Measures for Adults with Heart Failure included a performance measure of the use of warfarin at hospital discharge in all patients with HF and AF without contraindications (7).
Despite the very high prevalence of AF in patients with HF (8,9), little is known about the quality of care for AF in contemporary patients hospitalized with HF. The goal of this study is to examine characteristics, treatments, and quality measure performance in patients hospitalized with HF and AF in the Get With The Guidelines (GWTG)-HF registry. Furthermore, this analysis attempts to define factors associated with appropriate warfarin use in this rapidly enlarging patient population.
Data were collected through the GWTG program, a national, ongoing, prospective, observational data collection and quality improvement initiative started in the year 2000 under the guidance and sponsorship of the AHA. The GWTG program and its component data elements have been described previously (10,11). Beginning in January 2005, patients hospitalized with new or worsening HF or those patients in whom significant HF symptoms developed during hospitalization for another primary diagnosis (i.e., HF was the primary discharge diagnosis) were enrolled into the registry, regardless of their left ventricular function. Participating centers were instructed to submit data on consecutive eligible HF patients to the GWTG-HF registry. Hospital teams used HF case-ascertainment methods similar to those of the Joint Commission.
All participating institutions were required to comply with local regulatory and privacy guidelines and, if required, to secure institutional review board approval. Because data were used primarily at the local site for quality improvement, sites were granted a waiver of informed consent under the common rule. Outcome Sciences served as the registry coordinating center. The Duke Clinical Research Institute served as the data coordinating center and analyzed the aggregate de-identified data for research purposes.
Trained personnel abstracted the data using standardized definitions for all data elements, including AF, HF, ischemic versus nonischemic etiology, and comorbidities. Patients were assigned to race/ethnicity categories using options defined by the electronic case report form. Other variables included demographic and clinical characteristics, medical history, admission laboratory data, previous treatments, inpatient procedures (including cardioversion), contraindications for evidence-based therapies, and in-hospital outcomes.
Data collection on rhythm status included prior history of chronic or recurrent AF, prior history of atrial flutter, and the presence of AF on admission. With respect to warfarin use at discharge, appropriate eligibility was captured. The following contraindications to warfarin use were recorded: allergy to warfarin, pregnancy, excess bleeding risk, excess fall risk, other medical contraindications, or patient reasons. Using a web-based system, data quality was monitored to ensure the completeness and accuracy of the submitted data. Data edit checks were performed to ensure the validity of the collected data. Only sites and variables with a high degree of completeness were used in the analysis.
From January 1, 2005, through March 25, 2008, 104,268 admissions with HF were discharged from 340 hospitals participating in GWTG-HF. We excluded 31,548 patients from 85 sites that participated in GWTG-HF in a limited way and reported medical histories with <75% completeness. There were 72,720 admissions with HF that were discharged from 255 hospitals participating in the complete GWTG-HF registry. We excluded 186 patients because of conflicting data fields. The final overall study population included 72,534 HF admissions.
The primary outcome (and performance measure) was warfarin use at discharge among eligible patients with a history of AF or AF on admission without documented contraindications, intolerance, or other documented reasons for not prescribing warfarin.
Secondary analyses were planned to ascertain whether warfarin use at discharge varied according to risk stratification by the CHADS2(congestive heart failure, hypertension, age >75, diabetes, and prior stroke or transient ischemic attack) score, across hospital sites, and finally, whether warfarin use improved with time.
Using chi-square tests for categorical variables and the Kruskal-Wallis test for continuous variables, baseline characteristics were compared between: 1) patients with no history of AF; 2) patients with a history of AF only; and 3) patients with AF at the time of their admission. Medians and interquartile ranges (IQRs) were determined for continuous variables and percentages for categorical variables. Univariate analyses and multivariable logistic regression were used to identify important factors associated with warfarin use at discharge among eligible patients. In univariate analyses, the Wilcoxon rank sum test was used for continuous variables, and the Pearson chi-square test was used for categorical variables. Generalized estimating equations were used to adjust for clustering within hospitals (12). Candidate variable selection was based on prior GWTG analyses, essential baseline demographics, and clinical experience. The initial model included variables for age, sex, race, geographic region, hospital type, systolic blood pressure, heart rate (HR), body mass index, left ventricular ejection fraction (LVEF), anemia, AF, cerebrovascular disease (stroke or transient ischemic attack), chronic obstructive pulmonary disease, coronary artery disease, depression, diabetes mellitus, hyperlipidemia, hypertension, nonischemic cardiomyopathy, peripheral vascular disease, renal insufficiency, smoking, chronic dialysis, prior history of revascularization (coronary bypass grafting or percutaneous intervention), and antiplatelet therapy at discharge (aspirin, clopidogrel, or both). Interaction terms were included for race*geographic region. Significant variables were selected using backward elimination at the p < 0.1 level. Factors that were insignificant were removed from the logistic regression model. All tests were 2-tailed, and statistical significance was declared when p ≤ 0.05. All analyses were performed using SAS software version 9.0 (SAS Institute, Inc., Cary, North Carolina).
Among 72,534 admissions for HF, 47,715 (65.8%) did not have a history of AF and 9,918 (13.7%) had a prior history of AF but were in a regular rhythm at admission; 1 in 5 HF admissions presented in AF (n = 14,901 of 72,534 or 20.5%). Of those patients who presented in AF, 11,551 (77.5%) also had a history of prior AF. As shown in Table 1,compared with patients with no history of AF, patients who presented in AF more often were white, were anemic, and had a lower prevalence of diabetes and renal insufficiency. Patients admitted with AF were more likely to have HF with preserved systolic function because they had a higher median LVEF (40% vs. 35%) and a lower incidence of significant left ventricular dysfunction (40% vs. 47.8% with an LVEF <40%). Patients admitted in AF were more likely to have a history of stroke or transient ischemic attack (15.5% vs. 12.2%). Patients admitted with AF also were more likely to be transferred in from another institution. Among the patients with AF on admission, 23.9% (n = 3,172 of 13,269) had an HR >100 beats/min and 8.8% (n = 1,163 of 13,269) had an HR >120 beats/min, consistent with a rapid ventricular response. The median (IQR) LVEF of patients with an admission HR >120 beats/min was 40% (interquartile range 25% to 55%).
Among eligible patients with a prior history of AF, 51.3% were treated with warfarin before admission, and this decreased over time (p value for trend <0.0001). Among the 24,819 HF admissions with AF (prior history or AF on admission), 2,290 (9.2%) had documented contraindications to warfarin therapy, including an allergy to warfarin (n = 37 of 2,290; 1.6%), pregnancy (n = 0 of 2,290), excess bleeding risk (n = 713 of 2,290; 31.1%), excess fall risk (n = 428 of 2,290; 18.7%), or other documented contraindications (n = 1,208 of 2,290; 52.8%).
Warfarin use at discharge
Among eligible patients, the median prevalence of warfarin therapy at discharge was 64.9% (IQR 55.5% to 73.4%). Because the primary aim of our analysis was to describe the use of warfarin at discharge, we examined factors associated with warfarin use in the 15,748 patients with AF (prior history or AF on admission) who had no contraindications to warfarin. In univariate analysis, we found that older patients, women, black patients, Medicaid recipients, and those with a history of coronary artery disease, anemia, and renal insufficiency were less likely to be discharged on warfarin (Table 2).Additionally, HF admissions at smaller hospitals, nonacademic hospitals, and hospitals without house staff or interventional capabilities were less likely to discharge patients on warfarin.
As shown in Table 3,multivariable logistic regression identified discharge on antiplatelet therapy, anemia, renal insufficiency, and nonwhite race as factors independently associated with higher odds of no warfarin use at discharge in patients with AF (p < 0.001). With respect to sex, there was less warfarin use in women (p = 0.001). Finally, patients treated in the southern U.S. were less likely to receive warfarin at discharge (p = 0.047). The race*region interaction was not significant and was not included in the final model (Table 3). Finally, we conducted a sensitivity analysis restricted to only those patients with AF on admission. The results of this multivariable model for factors associated with nonwarfarin use at discharge were similar (Online Appendix 1). Eligible patients who were admitted with AF were more likely to receive anticoagulation at discharge than those patients with a history AF who were in a regular rhythm at admission (68.5% vs. 60.7%, p < 0.0001).
Risk stratification and warfarin use
The ACC/AHA/Physician Consortium AF and Atrial Flutter Performance Measures (13) call for assessment and documentation of thromboembolic risk factors. To examine the relationship between risk factors for stroke and use of anticoagulation, we analyzed warfarin use at discharge in eligible patients according to the CHADS2risk stratification scheme (14). As illustrated in Figure 1,there was significant risk–treatment mismatch. Warfarin use declined in patients with increasing risk for stroke, as indexed by higher CHADS2scores (p < 0.0001). Among patients with HF as their only risk factor, 71% were prescribed warfarin, compared with 60% among patients with a CHADS2score of 6.
Variation in warfarin prescribing
To assess how prescription of warfarin at discharge varied at different hospitals, we investigated the variance among warfarin use at discharge across sites. The use of warfarin at discharge at each site ranged from 0 to 95.5% and is shown in Figure 2.There was less interhospital variation in documented contraindications/other reasons for not prescribing warfarin in AF patients with HF (median 6.5%, IQR 0% to 13.5%). We also examined the trend in warfarin use over the 3.5 years of the study. In both quarterly and yearly assessments, there was no evidence of increasing use of warfarin at discharge with time (Fig. 3).In 2005, 62.2% of eligible patients were discharged on warfarin, compared with 66.8% in 2006, 65.5% in 2007, and 63.5% in the first quarter of 2008 (p = 0.1457).
Combination antithrombotic pharmacotherapy
Because some patients may have received stroke prophylaxis other than warfarin, or combination antithrombotic therapy, we also examined the use of other antithrombotic medications at discharge (Fig. 4).Among the 15,748 patients with a history of AF or AF on admission, the most common discharge antithrombotic strategy was warfarin monotherapy (n = 5,913, 37.6%). Eighty-nine percent of the cohort was taking at least some form of antithrombotic therapy (either warfarin or an antiplatelet agent). The use of combination antiplatelet therapy with warfarin, so-called triple therapy, was only documented in 3.4% (n = 531). Among patients on triple therapy, only 10.9% had a percutaneous coronary intervention (n = 58). Finally, 1 in 10 patients with a diagnosis of AF was not taking any medication for stroke prophylaxis.
Using GWTG-HF, a nationwide clinical practice registry, we examined the quality of care for AF in over 70,000 HF admissions at more than 250 clinical sites. There were 4 main findings in our analysis. First, stroke prophylaxis is significantly underused in eligible patients with AF and HF. Second, there is a significant risk–treatment mismatch in stroke prevention, such that patients with higher CHADS2scores were less likely to receive anticoagulation. Moreover, despite national guideline recommendations and an ACC/AHA performance measure for anticoagulation at the time of hospital discharge, warfarin use has not improved with time. Finally, we found that there are several important factors associated with warfarin underuse, including age, race, clopidogrel use, and treatment at smaller hospitals.
Given the increasing prevalence of both HF and AF, improving the quality of care for AF is paramount in this patient population. The ACC/AHA guidelines for the management of chronic HF have advocated the use of warfarin for stroke prophylaxis in patients with paroxysmal or persistent AF since 2005 (Class I recommendation; Level of Evidence: A) (6,7). The 2005 ACC/AHA HF performance measures include a hospital performance measure for the use of warfarin at hospital discharge in all eligible patients without contraindications. Although warfarin underuse has been well documented in other AF populations (15–17), the significant underuse of warfarin (a proven lifesaving intervention) in this cohort of hospitalized HF patients is cause for concern. Despite unequivocal guideline recommendations, 1 in 3 eligible patients was not discharged on warfarin. Furthermore, there was no evidence of longitudinal incorporation of guideline recommendations because rates of warfarin use did not improve over time.
The warfarin underuse in this real-world population has significant implications. The decision to anticoagulate patients requires a careful balance of the risks and benefits. Considering the mean CHADS2score of 3 in our hospitalized cohort, the risk of stroke or transient ischemic attack could be estimated at 5.9% per year (18). Therefore, the failure to prescribe warfarin (which is associated with a two-thirds reduction in stroke risk, or one-third reduction when compared with aspirin) (19) in the 5,475 HF admissions could have led to between 110 and 216 preventable thromboembolic events. If these findings are generalized to the entire U.S. populations hospitalized with HF and AF, there may be as many as 7,000 preventable thromboembolic events occurring each year.
An equally important finding in our analysis was the discovery of wide interhospital variance in warfarin use at discharge. Site-specific warfarin use at discharge varied widely, ranging between 0% and 96%. Significant interhospital variance has been described previously for other evidence-based therapies (10,20). In this HF registry, smaller hospital size was associated with decreased warfarin use. This association may reflect differences in hospital resources. Smaller hospitals may not have sufficient volume to afford the use of formal or automated stroke prophylaxis screening tools. Hopefully, emphasis on performance measures will help ameliorate this variance.
In addition to interhospital variation, we also identified significant regional variation in the use of warfarin at discharge. Warfarin use at discharge was greatest in the Northeast, whereas treatment in the South was associated with lower rates of warfarin use. The relationship between decreased warfarin use and treatment in the South persisted after adjustment for multiple clinical factors, including race. This regional variation has important implications, given the higher prevalence of stroke and higher stroke-related mortality in the South (21,22).
The CHADS2index is a validated risk stratification tool for stroke risk in patients with nonvalvular AF, and its use is advocated in the ACC/AHA/European Society of Cardiology 2006 guidelines for the management of patients with AF (18). Increasing CHADS2scores reflect increasing risk for stroke. In this cohort, higher CHADS2scores were paradoxically associated with decreased warfarin use. This risk–treatment mismatch is surprising given the increased magnitude of benefit and absolute risk reduction in these high-risk patients (23,24). Previously, it has been observed that patients at the highest risk are less likely to receive invasive or interventional procedures (25,26). Similarly, elderly patients with AF are also less likely to receive anticoagulation, presumably because of perceived increased risks and imbalance in the risk–benefit ratio. Given the extremely low rates of intracranial bleeding, the mortality benefit afforded with warfarin therapy, and the increased risk of thromboembolism in these patients, warfarin should be the preferred therapy (27,28). Substitution of aspirin should only occur in patients with significant risk of bleeding. The use of some form of antithrombotic therapy in 89% of patients, paired with warfarin underuse and the marked risk–treatment mismatch observed in this population, suggests that stroke risk is not adequately influencing anticoagulation decisions in patients with AF and HF.
Concordant with the observed risk–treatment mismatch between CHADS2scores and rates of warfarin use, our analysis also showed that nonwhite patients were 30% less likely to receive warfarin for stroke prophylaxis. Despite an increased risk for stroke, black patients were less likely to receive warfarin. Finally, the decreased warfarin use in women is also a cause for concern given the increased risk of stroke in women relative to men.
Among the warfarin-eligible patients in this cohort, approximately 1 in 4 were receiving antiplatelet therapy only (16% aspirin only, 2% clopidogrel only, and 5% aspirin and clopidogrel). Although one could advocate that these patients were low risk and could be appropriately treated with antiplatelet therapy only, recent and prior trial evidence has shown that oral anticoagulation is superior to antiplatelet therapy (19,29). Although some evidence suggests that HF is not a potent risk factor for stroke (30), several studies have clearly established increased risk in patients with HF (14,31). Consistent with this increased risk, the ACC/AHA guidelines for the management of patients with HF and the American College of Chest Physicians practice guidelines for antithrombotic therapy in AF recommend warfarin for all eligible HF patients with AF (6,32).
Currently, there is a great deal of controversy regarding the appropriate antithrombotic regimen in patients with AF and an indication for extended dual antiplatelet therapy (e.g., placement of a drug-eluting stent). So-called triple therapy (warfarin with aspirin and a thienopyridine) has been associated with increased bleeding; however, withholding anticoagulation in this patient population is also associated with increased mortality (28,33). The 2006 ACC/AHA/Heart Rhythm Society guidelines for the management of patients with AF advocate treatment with warfarin and the addition of a thienopyridine for the maintenance of stent patency (18). This recommendation is based on expert opinion (Level of Evidence: C). In our analysis, clopidogrel use at discharge was a strong, independent predictor of discharge without warfarin. Although there is no therapeutic consensus, given the available data and the magnitude of stroke risk compared with the relatively rarity of stent thrombosis (34), warfarin should not be withheld in patients with AF and HF. Additionally, the combination of aspirin and clopidogrel has been shown to be inferior to warfarin for the prevention of stroke in patients with AF (29). Although further research and long-term outcome data are needed, this is an important opportunity for targeted quality improvement initiatives.
First, although we controlled for many clinical variables known to be associated with an increased risk for bleeding and we excluded those patients in whom an increased risk of bleeding was documented, we did not have data on prior bleeding events. Second, data were collected by medical chart review and are dependent on the accuracy and completeness of documentation and abstraction. Contraindications and intolerance were recorded as noted in the medical record, but a proportion of untreated patients we classified as eligible for treatment may have had contraindications or intolerance that were indeed present but not documented. Fourth, the GWTG database currently does not track post-discharge treatment and outcomes. Although the link between the anticoagulation for AF as a process quality measure and adverse post-discharge outcomes may be valid, we were not able to directly explore the consequences of lack of anticoagulation in this patient population. Finally, the voluntary GWTG program may include hospitals with a higher likelihood of following evidence-based recommendations, therefore biasing our observations toward conservative estimates. Prior studies of the OPTIMIZE-HF (Organized Program to Initiate Lifesaving Treatment in Hospitalized Patients with Heart Failure) trial and other HF registries have shown that patients admitted with HF have similar baseline characteristics to patients from national datasets, suggesting that data from registry hospitals are likely to be representative of national trends and practices (18,35). Nonetheless, our study represents the largest description of warfarin use in an eligible, real-world HF population.
More than one-third of patients hospitalized with HF have a history of AF or present with AF on admission. Guideline-recommended warfarin use in eligible patients with HF and AF is less than optimal and varies significantly according to age, race, comorbidities, region, and hospital site. There is a significant risk–treatment mismatch, with HF patients with higher CHADS2scores less likely to receive anticoagulation. Anticoagulation in patients with AF and HF represents a significant opportunity for quality improvement.
For a table on the factors associated with nonwarfarin use at discharge among patients with atrial fibrillation on admission, please see the online version of this article.
The Get With The Guidelines Heart Failure program is supported by an unrestricted educational grant from GlaxoSmithKline. Dr. Piccini is supported by an American College of Cardiology Foundation/Merck Award. Dr. Hernandez is supported by American Heart Association Pharmaceutical Roundtable grant 0675060N; has received research grants from Johnson & Johnson (Scios), Merck, and GlaxoSmithKline; and has received honoraria from Novartis and AstraZeneca. Drs. Hernandez and Peterson are supported by grant 1U18HS016964 from the Agency for Health care Research and Quality. Dr. Patel is on the advisory board of Genzyme. Dr. Lewis has served as a consultant/speaker for Medtronic (modest). Dr. Peterson has received grant/research support from Bristol-Myers Squibb/Sanofi and Schering-Plough. Dr. Fonarow is supported by the Ahmanson Foundation (Los Angeles, California); has received research grants from GlaxoSmithKline/Medtronic; has received honoraria from Bristol-Myers Squibb/Sanofi, GlaxoSmithKline, Medtronic, and Schering-Plough; served as a consultant for GlaxoSmithKline, Medtronic, Sanofi-Aventis, Schering-Plough, and Scios; and serves as chair of the American Heart Association's Get With The Guidelines Steering Committee.
- Abbreviations and Acronyms
- American College of Cardiology/American Heart Association
- atrial fibrillation
- Get With The Guidelines
- heart failure
- heart rate
- interquartile range
- left ventricular ejection fraction
- Received January 19, 2009.
- Revision received March 24, 2009.
- Accepted April 3, 2009.
- American College of Cardiology Foundation
- Fang J.,
- Mensah G.A.,
- Croft J.B.,
- Keenan N.L.
- Hunt S.A.,
- Abraham W.T.,
- Chin M.H.,
- et al.
- Bonow R.O.,
- Bennett S.,
- Casey D.E. Jr..,
- et al.
- Piccini J.P.,
- Hernandez A.F.,
- Dai D.,
- et al.
- Estes N.A. 3rd.,
- Halperin J.L.,
- Calkins H.,
- et al.
- Stafford R.S.,
- Radley D.C.
- Fuster V.,
- Ryden L.E.,
- Cannom D.S.,
- et al.
- Kaul P.,
- Peterson E.D.
- Howard G.,
- Prineas R.,
- Moy C.,
- et al.
- Bungard T.J.,
- Ghali W.A.,
- McAlister F.A.,
- et al.
- Rogacka R.,
- Chieffo A.,
- Michev I.,
- et al.
- Wolf P.A.,
- Abbott R.D.,
- Kannel W.B.
- Ruiz-Nodar J.M.,
- Marín F.,
- Hurtado J.A.,
- et al.
- McNamara R.L.,
- Brass L.M.,
- Drozda J.P. Jr..,
- et al.