Author + information
- Received April 25, 2008
- Revision received August 25, 2008
- Accepted September 22, 2008
- Published online January 13, 2009.
- Adrian F. Hernandez, MD, MHS⁎,†,⁎ (, )
- Bradley G. Hammill, MS⁎,
- Christopher M. O'Connor, MD, FACC⁎,†,
- Kevin A. Schulman, MD⁎,†,
- Lesley H. Curtis, PhD⁎,† and
- Gregg C. Fonarow, MD, FACC‡
- ↵⁎Reprint requests and correspondence:
Dr. Adrian F. Hernandez, Duke Clinical Research Institute, P.O. Box 17969, Durham, North Carolina 27715
Objectives We sought to examine associations between initiation of beta-blocker therapy and outcomes among elderly patients hospitalized for heart failure.
Background Beta-blockers are guideline-recommended therapy for heart failure, but their clinical effectiveness is not well understood, especially in elderly patients.
Methods We merged Medicare claims data with OPTIMIZE-HF (Organized Program to Initiate Lifesaving Treatment in Hospitalized Patients With Heart Failure) records to examine long-term outcomes of eligible patients newly initiated on beta-blocker therapy. We used inverse probability–weighted Cox proportional hazards models to determine the relationships among treatment and mortality, rehospitalization, and a combined mortality–rehospitalization end point.
Results Observed 1-year mortality was 33%, and all-cause rehospitalization was 64%. Among 7,154 patients hospitalized with heart failure and eligible for beta-blockers, 3,421 (49%) were newly initiated on beta-blocker therapy. Among patients with left ventricular systolic dysfunction (LVSD) (n = 3,001), beta-blockers were associated with adjusted hazard ratios of 0.77 (95% confidence interval [CI]: 0.68 to 0.87) for mortality, 0.89 (95% CI: 0.80 to 0.99) for rehospitalization, and 0.87 (95% CI: 0.79 to 0.96) for mortality–rehospitalization. Among patients with preserved systolic function (n = 4,153), beta-blockers were associated with adjusted hazard ratios of 0.94 (95% CI: 0.84 to 1.07) for mortality, 0.98 (95% CI: 0.90 to 1.06) for rehospitalization, and 0.98 (95% CI: 0.91 to 1.06) for mortality–rehospitalization.
Conclusions In elderly patients hospitalized with heart failure and LVSD, incident beta-blocker use was clinically effective and independently associated with lower risks of death and rehospitalization. Patients with preserved systolic function had poor outcomes, and beta-blockers did not significantly influence the mortality and rehospitalization risks for these patients.
Heart failure is the most common discharge diagnosis for Medicare beneficiaries, and more than one-half of patients admitted for heart failure will be readmitted within 6 months (1,2). Although outcomes for heart failure are poor, advances in understanding of heart failure over the past 2 decades have led to important advances in therapy. One major advance is the clinical trial evidence for beta-blockers, which is now a cornerstone of heart failure treatment (3).
Among patients with heart failure and reduced systolic function who are enrolled in randomized clinical trials, beta-blockers confer a 10% to 40% reduction in mortality and hospitalization within 1 year (4–6). Patients enrolled in randomized clinical trials tend to be relatively young, have relatively few comorbid conditions, and are not representative of the general heart failure population encountered in clinical practice (7). Moreover, patients' care in clinical trials may be significantly better compared with that of the general population of patients with heart failure. Thus, there is limited data on the effectiveness and safety of beta-blockers in non–clinical trial populations of patients with systolic dysfunction, and there are almost no data for patients with preserved systolic function.
A number of studies have found that evidence-based heart failure therapies are underused, especially in the elderly (8–11). Although current guidelines recommend that populations underrepresented in heart failure clinical trials—including elderly patients, women, racial/ethnic minorities, and patients with comorbid conditions—should be treated like the broader population in the absence of specific evidence to the contrary, some physicians may be hesitant to do so without additional evidence of effectiveness in these populations (3–6,12).
Given the limited evidence and the uncertainty, we sought to measure the clinical effectiveness of beta-blockers in a broad group of eligible elderly patients with heart failure who were newly initiated on beta-blocker therapy and discharged after a heart failure hospitalization.
We merged data from the OPTIMIZE-HF (Organized Program to Initiate Lifesaving Treatment in Hospitalized Patients With Heart Failure) registry and from enrollment files and in-patient claims from the Centers for Medicare and Medicaid Services (CMS). OPTIMIZE-HF was a national registry and performance-improvement program for patients hospitalized with heart failure (13–16). This nationwide patient registry was used to gather data on various patient characteristics with a web-based information system. In an approach similar to Joint Commission case ascertainment of heart failure hospitalizations, eligibility for the OPTIMIZE-HF registry required patients to be adults hospitalized for an episode of new or worsening heart failure as the primary cause of admission or with significant heart failure symptoms that developed during a hospital stay with a primary discharge diagnosis of heart failure. Similar to other national cardiovascular registries and quality-reporting initiatives, data regarding medical history, signs and symptoms, medications, and diagnostic tests were collected via a web-based registry. The registry also includes data on contraindications, intolerance, and other reasons for not prescribing therapy. All regions of the U.S. are represented in the registry, and a variety of institutions participated, from community hospitals to large tertiary medical centers. The OPTIMIZE-HF protocol was approved by each participating center's institutional review board or by a central institutional review board. Automated electronic data checks were used to prevent out-of-range entries and duplicate patients. A source data verification audit of a random 5% of the first 10,000 records collected showed better than 99% concordance on 53% of the fields (118 of 223) and 95% concordance on 91% of the fields (205 of 223). Fields with less than 95% concordance were not used in this analysis.
The CMS files include all fee-for-service Medicare beneficiaries age 65 years and older who were hospitalized with a diagnosis of heart failure (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] 428.x, 402.x1, 404.x1, 404.x3). We matched patients in the OPTIMIZE-HF registry to Medicare enrollment and in-patient data according to date of birth, sex, admission date, discharge date, and hospital site. Of the 36,165 hospitalizations involving patients age 65 years and older, we matched 29,301 (81%) to CMS claims, representing 25,901 unique patients. The OPTIMIZE-HF patients linked to Medicare claims had similar demographic characteristics and outcomes as other Medicare beneficiaries hospitalized for heart failure (17). For example, compared with 925,161 non–OPTIMIZE-HF fee-for-service Medicare beneficiaries, the OPTIMIZE-HF patients were not different than non–OPTIMIZE-HF patients with respect to age (79.5 years vs. 79.8 years), sex (44.0% vs. 42.5% male), or race (89.1% vs. 88.6% nonblack).
The institutional review board of the Duke University Health System approved this study.
After exclusion of 1,212 patients who died before discharge, the merged dataset included 24,689 patients with heart failure age 65 years and older who were discharged alive from the hospital and for whom Medicare data were available (Fig. 1). For patients with multiple hospitalizations recorded in the registry, we used information from the earliest hospitalization. We included only patients documented as eligible for beta-blocker therapy. We successively excluded patients who had documented contraindications or intolerance to beta-blockers (n = 2,975), who were discharged or transferred to other short-term hospitals (n = 309) or hospice (n = 399), or who left against medical advice (n = 49) (18). We also excluded patients who were missing information on admission beta-blocker status, discharge beta-blocker status, or discharge status (n = 114) and patients with incomplete or inconsistent Medicare follow-up history (n = 364).
From the remaining patients, we created 2 analysis cohorts. The first cohort—those with left ventricular systolic dysfunction (LVSD)—included patients with either documented left ventricular ejection fraction (LVEF) of <40% or qualitative documentation of LVSD (n = 7,529). The second cohort—those with preserved systolic function—included patients with either documented LVEF of at least 40% or qualitative documentation of preserved systolic function (n = 9,712). We excluded patients without documentation of left ventricular function (n = 3,238).
The outcomes of interest were time to death, time to first rehospitalization, and time to death or first rehospitalization within 1 year after hospital discharge. We obtained dates of death from the CMS enrollment files and data regarding hospital admissions from the CMS in-patient files. We did not count subsequent admissions for rehabilitation as rehospitalizations in this analysis.
The primary analysis estimated the effect of newly initiated beta-blocker therapy versus no beta-blocker therapy on mortality and all-cause rehospitalization at 1 year in eligible patients with heart failure who were naïve to beta-blockers at hospital admission. There were 3,001 patients in the LVSD cohort and 4,153 patients in the preserved systolic function cohort. We analyzed the 2 cohorts separately. In sensitivity analyses, we replicated the primary analysis in all eligible patients regardless of whether they were taking beta-blockers before admission. That is, we included both patients newly initiated on beta-blocker therapy and those who were continuing beta-blocker therapy. We also explored whether findings for clinical effectiveness of beta-blockers differed between patients with LVEF of 40% to 49% compared with the overall cohort with preserved systolic function.
In the LVSD and preserved systolic function cohorts, we also evaluated whether there was heterogeneity in the findings regarding beta-blocker effectiveness. Specifically, we examined whether the effectiveness of beta-blockers was the same for patients with and without chronic obstructive pulmonary disease, renal insufficiency, and diabetes mellitus.
We summarized baseline characteristics by treatment group using percentages for categoric variables and medians and interquartile ranges for continuous variables. For comparisons by treatment group, we used chi-square tests for categoric variables and Wilcoxon rank sum tests for continuous variables. We summarized observed mortality using a Kaplan-Meier estimator and rehospitalization using the cumulative incidence function (19). We estimated the unadjusted relationship between treatment and outcome using a Cox proportional hazards model in which treatment group was the only variable. We estimated the adjusted relationship between treatment and outcome using an inverse probability–weighted Cox proportional hazards model (20,21). The weights in this model were based on propensity to receive treatment (Online Appendix). Specifically, we used logistic regression models to estimate the propensity to receive any beta-blocker at discharge (vs. no beta-blocker) as a function of age at admission, sex, race, ischemic heart failure status, LVEF, systolic blood pressure, smoking within the prior year, rales, or lower-extremity edema at admission; serum sodium, serum creatinine, and hemoglobin levels at admission; other discharge medications (i.e., angiotensin-converting enzyme [ACE] inhibitor or angiotensin receptor blocker [ARB]); in-hospital procedures (i.e., angiography, mechanical ventilation, implantable cardioverter-defibrillator); and history of diabetes mellitus, hyperlipidemia, atrial arrhythmia, chronic obstructive pulmonary disease, reactive airway disease, anemia, peripheral vascular disease, thyroid abnormality, prior cerebrovascular accident or transient ischemic attack, depression, or renal insufficiency. To assess the heterogeneity of effect by subgroup, we included treatment-by-subgroup interactions in separate inverse probability–weighted Cox proportional hazards models. Because this approach resulted in additional statistical comparisons, we required p < 0.01 for the interactions to be considered statistically significant.
In the primary analysis of the LVSD cohort (n = 3,001), which included patients with heart failure and LVSD who were eligible for beta-blockers and naïve to beta-blockers at admission, 1,800 patients (60.0%) were newly initiated on beta-blocker therapy. Table 1 shows the baseline characteristics of the cohort by discharge beta-blocker status. Patients discharged on beta-blockers were slightly younger and had less atrial arrhythmia and renal insufficiency compared with eligible patients discharged without beta-blockers.
In the preserved systolic function cohort (n = 4,153), which included patients with heart failure and preserved systolic function who were eligible for beta-blockers and naïve to beta-blockers at admission, 1,621 patients (39.0%) were discharged on beta-blocker therapy. Table 1 shows the baseline characteristics of the cohort by discharge beta-blocker status. Patients discharged on beta-blockers were more often nonwhite and had a higher frequency of ischemic heart failure than patients discharged without beta-blockers. Eligible patients discharged without beta-blockers had higher median LVEF and higher prevalence of atrial arrhythmia and chronic obstructive pulmonary disease.
Table 2 shows the discharge therapies for the 2 cohorts. In general, patients discharged on beta-blocker therapy in both cohorts were more likely to be taking ACE inhibitors or ARBs, antiplatelet agents, lipid-lowering agents, and aldosterone antagonists. In-patient cardiac catheterization rates were higher in both cohorts for patients discharged on beta-blocker therapy.
Observed 1-year mortality for all patients who were eligible for and naïve to beta-blockers at admission was 32.8%, and the 1-year all-cause rehospitalization rate was 64.3%. In the LVSD cohort, 1-year mortality was 34.3% and the 1-year rehospitalization rate was 63.2%. In the preserved systolic function cohort, 1-year mortality was 31.7% and the rehospitalization rate was 65.0%. One-year survival for eligible patients initiated on beta-blocker therapy compared with eligible patients not initiated on beta-blocker therapy at discharge is shown in Figure 2 for patients with LVSD and in Figure 3 for patients with preserved systolic function.
In both cohorts, the discrimination of the propensity model was moderate, indicating that the treatment groups were similar enough to be adjusted using propensity weights (22). The c statistics for the propensity models were 0.67 in the LVSD cohort and 0.63 in the preserved systolic dysfunction cohort. In both models, the most important variables, defined as those with the highest Wald chi-square values, included other discharge therapies (i.e., antiplatelet therapy, ACE inhibitor, or ARB) and within-hospital angiography. After observations were weighted by their propensity for treatment received, none of the factors listed in Tables 1 or 2 differed significantly between treatment groups in either cohort.
Table 3 shows the unadjusted and adjusted results for incident beta-blocker therapy at discharge compared with no beta-blocker therapy for both cohorts. In the LVSD cohort, we observed a protective effect of beta-blocker therapy on mortality and on the combined end point of mortality and rehospitalization. After adjustment, the hazard rate for mortality was 23% lower for patients newly initiated on beta-blocker therapy, compared with those eligible but not discharged on beta-blocker therapy. For the combined end point, the adjusted hazard rate was 13% lower for patients newly initiated on beta-blocker therapy. After adjustment, incident beta-blocker therapy at discharge did not have a statistically significant effect on the hazard of rehospitalization among patients with LVSD. Among patients with preserved systolic function, incident beta-blocker therapy at discharge did not have a statistically significant effect on mortality, rehospitalization, or the combined end point after adjustment.
In sensitivity analyses, we explored outcomes in subgroups of patients with LVSD, including patients with and without chronic obstructive pulmonary disease, renal insufficiency, and diabetes mellitus. Table 4 shows the relatively homogeneous findings for mortality in these subgroups. Based on α = 0.01 as a significant interaction, adjusted hazard ratios for beta-blocker therapy showed no significant heterogeneity in patients with chronic obstructive pulmonary disease, diabetes mellitus, and renal insufficiency and in patients age 65 to 74 years and 75 years and older. The findings were similar for rehospitalization and the combined end point of mortality and rehospitalization in these subgroups. We performed subgroup analyses of patients with preserved systolic function (Online Appendix). We also examined 937 patients with measured ejection fractions of 40% to 49%. In this group, 454 (48%) patients initiated beta-blocker therapy at discharge. The unadjusted and inverse-weighted hazard ratios for initiation of beta-blocker therapy compared with no beta-blocker were similar. The adjusted hazard ratios were 0.95 (95% confidence interval [CI]: 0.75 to 1.20) for mortality, 0.91 (95% CI: 0.78 to 1.07) for rehospitalization, and 0.90 (95% CI: 0.78 to 1.04) for mortality/rehospitalization.
This study of clinical effectiveness addresses important questions about heart failure and beta-blockers by linking a large contemporary heart failure registry containing data on clinical characteristics, treatment eligibility, and contraindications and intolerance with long-term outcome data from Medicare claims. The registry includes a broad cohort of patients (including elderly patients) with heart failure from a variety of settings and from all regions of the U.S. The registry contains far more detailed information on patient characteristics, presenting symptoms, treatments, and outcomes than has been available in other administrative datasets or registries. We confined the analysis to patients who were eligible for beta-blocker therapy and who did not have documented contraindications, intolerance, or other reasons for not receiving beta-blockers. We found that beta-blockers were clinically effective in patients with LVSD, a population that is older and has more comorbid conditions than populations usually enrolled in randomized clinical trials. Moreover, although patients with preserved systolic function had substantial morbidity and mortality after hospitalization for heart failure, in marked contrast to patients with LVSD, we did not find a substantial benefit of beta-blocker therapy in this cohort.
In comparison with clinical trials, in which the median age was 65 years (4–6,12), the median age in our study was 78 years, and the patients had significant comorbidities. Beta-blockers were prescribed at discharge for nearly 80% of all eligible patients with LVSD, a striking finding given the prevalence of comorbid conditions such as diabetes mellitus, chronic obstructive pulmonary disease, renal insufficiency, and peripheral vascular disease, which are often perceived by practitioners as barriers to beta-blocker therapy. The overall risk in this study population is further highlighted by the 1-year mortality rate of 33%. In contrast, the COPERNICUS (Carvedilol Prospective Randomized Cumulative Survival) trial, which enrolled patients with an LVEF <25%, had an overall mortality rate of 14% during a mean follow-up of 10 months (6). The observed 1-year mortality benefit in this study is especially remarkable, given that eligible patients who were discharged without beta-blocker therapy may have subsequently started treatment and patients discharged on beta-blockers may have discontinued therapy or been nonadherent.
Our findings also highlight the risks of decompensation and cardiovascular and noncardiovascular hospital admissions among patients with heart failure. We found an all-cause rehospitalization rate of more than 60% in the first year. Previous clinical trials that included patients with severe heart failure had rehospitalization rates of <30% within 1 year (6). We found a significant reduction in the combination of mortality and all-cause rehospitalization with beta-blocker therapy for patients with heart failure and LVSD, although the major effect was on mortality alone.
We also found that elderly patients with heart failure and preserved systolic dysfunction had very high mortality and morbidity, with a 1-year mortality rate of 32% and a 1-year rehospitalization rate of 65%. Despite the alarming risk, clinical trial data regarding treatment for patients with heart failure and preserved systolic function are sparse. Beta-blockers may be beneficial in patients with preserved systolic function who also have other indications, such as coronary artery disease, diabetes mellitus, or hypertension. However, we found no association between beta-blocker therapy at discharge and clinical outcomes in the first year of follow-up in patients admitted with preserved systolic function. There are several possible explanations for the lack of benefit of beta-blockers in this cohort, including the heterogeneity of the population and poor understanding of pathophysiology. The single randomized trial of beta-blockers that included patients with preserved systolic function did not find a mortality benefit in the subgroup of patients with LVEF of at least 40% (23). Thus, our findings among patients enrolled in the OPTIMIZE-HF registry are consistent with the existing clinical trial evidence. More studies are needed to address this high-risk population and to prospectively test new therapies that could improve outcomes.
Our study strengthens the findings of previous studies by including important clinical characteristics, such as LVEF, and prospective data on eligibility for treatment, including contraindications and intolerance. Go et al. (24) analyzed medical records from 2 health care systems to assess the comparative effectiveness of beta-blockers on the risk of rehospitalization for heart failure. After adjustment for risks of admission and propensity to receive beta-blockers, they did not find significant differences in rehospitalization within 12 months for patients on atenolol, metoprolol tartrate, carvedilol, or other beta-blockers. Because the study had limited data on beta-blockers primarily studied in clinical trials and did not have data on LVEF, the importance of evidence-based beta-blockers versus non–evidence-based beta-blockers could not be determined. The striking difference by left ventricular function in the association of beta-blocker therapy with post-discharge outcomes in our study highlights the importance of accurate documentation of left ventricular function for studies of effectiveness in patients with heart failure. The difference also makes it unlikely that observations of improved outcomes in patients with LVSD were simply a result of lower-risk patients being treated with beta-blockers or residual confounding, because these would have applied equally to patients with preserved systolic function.
First, hospitals participating in the OPTIMIZE-HF registry were self-selected and may not be representative of national care patterns and clinical outcomes. Second, the follow-up data included were derived from the subset of patients in the registry who were enrolled in fee-for-service Medicare. Third, eligibility for beta-blocker therapy is based on documentation in the medical record and is thus dependent on the accuracy of this documentation. Some patients may have had contraindications or intolerances that were not documented. Fourth, we did not assess exposure to beta-blockers after hospital discharge, and patients discharged without beta-blocker therapy may have started treatment and those discharged with beta-blocker therapy may have discontinued or not fully adhered to therapy after discharge. A prior analysis of the OPTIMIZE-HF registry showed that in a pre-specified 10% sample of patients with 60 to 90 days of follow-up, 95% of eligible patients discharged on beta-blocker therapy remained on therapy (25). Any drop-off in beta-blocker therapy would diminish the likelihood of finding a significant difference in post-discharge clinical outcomes, making the findings of a significant difference in survival based on discharge use of beta-blocker therapy more remarkable. Also, we did not assess health-related quality of life, functional capacity, patient satisfaction, or other outcomes that may be of interest; beta-blockers may or may not be associated with these outcomes in clinical practice. Also, we did not have data ascertaining causes of death and rehospitalization. Some studies have suggested that there are differences in mode of death between heart failure patients with reduced and preserved systolic functions (26).
The association between beta-blocker therapy and outcomes may be confounded by socioeconomic factors that are not included in the multivariable and propensity models. There may also be other unmeasured confounders that, had they been adjusted for, would have strengthened or weakened the association between beta-blocker use and clinical outcomes. This study was observational, and randomized, controlled trials are the best methods for establishing or comparing efficacy of treatment. However, it is unlikely that additional randomized comparisons for beta-blockers either broadly or in special populations, such as patients with preserved systolic function, will be conducted. Thus, our data serve as one of the few resources for improving understanding of clinical effectiveness in patients hospitalized for heart failure.
In an elderly population hospitalized for heart failure, incident beta-blocker therapy at discharge was associated with significantly improved risk and propensity-adjusted survival for patients with LVSD in the first year of follow-up. These findings extend the results of randomized clinical trials of beta-blockers conducted in selected outpatients with chronic systolic heart failure to a diverse cohort of patients hospitalized with heart failure and LVSD. Patients with heart failure and preserved systolic function have poor outcomes, and beta-blockers do not substantially influence the risk of death or rehospitalization in these patients.
The authors thank Damon M. Seils, MA, of Duke University for editorial assistance and manuscript preparation. Mr. Seils did not receive compensation for his assistance apart from his employment at the institution where the study was conducted.
For the supplementary tables on the results of a propensity model with outcome of initiation of beta-blocker therapy at discharge in the left ventricular systolic dysfunction cohort and preserved systolic dysfunction cohort, and a sensitivity analysis of subgroups in patients with heart failure and preserved systolic dysfunction, please see the online version of this article.
This work was supported by GlaxoSmithKline. Dr. Hernandez is a recipient of an American Heart Association Pharmaceutical Roundtable grant (0675060N). Drs. Curtis and Schulman were supported in part by grant 1R01AG026038-01A1 from the National Institute on Aging and grant 5U01HL66461-05 from the National Heart, Lung, and Blood Institute. Dr. Fonarow holds the Eliot Corday Chair in Cardiovascular Medicine and Science at the David Geffen School of Medicine, University of California, Los Angeles.
- Abbreviations and Acronyms
- angiotensin-converting enzyme
- angiotensin receptor blocker
- confidence interval
- left ventricular ejection fraction
- left ventricular systolic dysfunction
- Received April 25, 2008.
- Revision received August 25, 2008.
- Accepted September 22, 2008.
- American College of Cardiology Foundation
- ↵(2008) Heart disease and stroke statistics—2008 update: a report from the American Heart Association Statistics Committee and Stroke Statistics Subcommittee. Circulation 117:e25–e146.
- Hunt S.A.,
- Abraham W.T.,
- Chin M.H.,
- et al.
- Ezekowitz J.,
- McAlister F.A.,
- Humphries K.H.,
- et al.
- Fonarow G.C.,
- Abraham W.T.,
- Albert N.M.,
- et al.
- Curtis L.H.,
- Hernandez A.F.,
- Greiner M.A.,
- Fonarow G.C.,
- Schulman K.A.
- Fonarow G.C.,
- Abraham W.T.,
- Albert N.M.,
- et al.
- Flather M.D.,
- Shibata M.C.,
- Coats A.J.,
- et al.
- Henkel D.M.,
- Redfield M.M.,
- Weston S.A.,
- Gerber Y.,
- Roger V.L.