Association Between Beta-Blockers and Mortality and Readmission in Older Patients with Heart Failure: an Instrumental Variable Analysis

Abstract

Background

The demographics of heart failure are changing. The rate of growth of the “older” heart failure population, specifically those ≥ 75, has outpaced that of any other age group. These older patients were underrepresented in the early beta-blocker trials. There are several reasons, including a decreased potential for mortality benefit and increased risk of side effects, why the risk/benefit tradeoff may be different in this population.

Objective

We aimed to determine the association between receipt of a beta-blocker after heart failure discharge and early mortality and readmission rates among patients with heart failure and reduced ejection fraction (HFrEF), specifically patients aged 75+.

Design and Participants

We used 100% Medicare Parts A and B and a random 40% sample of Part D to create a cohort of beneficiaries with ≥ 1 hospitalization for HFrEF between 2008 and 2016 to run an instrumental variable analysis.

Main Measure

The primary measure was 90-day, all-cause mortality; the secondary measure was 90-day, all-cause readmission.

Key Results

Using the two-stage least squared methodology, among all HFrEF patients, receipt of a beta-blocker within 30-day of discharge was associated with a − 4.35% (95% CI − 6.27 to − 2.42%, p < 0.001) decrease in 90-day mortality and a − 4.66% (95% CI − 7.40 to − 1.91%, p = 0.001) decrease in 90-day readmission rates. Even among patients ≥ 75 years old, receipt of a beta-blocker at discharge was also associated with a significant decrease in 90-day mortality, − 4.78% (95% CI − 7.19 to − 2.40%, p < 0.001) and 90-day readmissions, − 4.67% (95% CI − 7.89 to − 1.45%, p < 0.001).

Conclusion

Patients aged ≥ 75 years who receive a beta-blocker after HFrEF hospitalization have significantly lower 90-day mortality and readmission rates. The magnitude of benefit does not appear to wane with age. Absent a strong contraindication, all patients with HFrEF should attempt beta-blocker therapy at/after hospital discharge, regardless of age.

Supplementary Information

The online version contains supplementary material available at 10.1007/s11606-021-06901-7

INTRODUCTION

Heart failure is one of the most common chronic conditions among older adults.¹ As modern medicine has transformed many previously fatal diseases, like heart failure, into chronic conditions, today’s heart failure with reduced ejection fraction (HFrEF) population has become larger and older, often with more chronic comorbidities.^2,3 These evolving demographics have created a number of challenges for clinical care, including well-founded clinical equipoise about how best to use standard neurohormonal antagonists, such as beta-blockers, in these older, often more complex HFrEF patients.

On basis of robust clinical trial data, we know beta-blocker decreases mortality in HFrEF.^4–7 However, many if not most, of today’s Medicare HFrEF population would have been excluded from these early, landmark trials. Moreover, by virtue of their older age, modern HFrEF patients have less potential to reap the mortality benefit of beta-blockers and, due to their higher comorbidity burden, simultaneously are at a higher risk for adverse events.

Unfortunately, the clinical trial data in this space is limited. In the SENIORS trial of nebivolol in patients aged ≥ 70, the authors found a 14% reduction in composite outcome of mortality or cardiovascular hospitalization.⁸ While informative, the use of a non-guideline-recommended HFrEF-specific beta-blocker and the use of > 70 to define “older” (today the average HFrEF beneficiary is 80 years old)^9,10 somewhat limit the modern application of these results. More recent work, using Medicare claims data, found that, on net, the mortality benefits of beta-blockers outweighed potential harms in HFrEF patients across the age spectrum.¹¹ The primary limitation of this study however was its retrospective design and the concern for residual confounding by indication; i.e., patients that were “healthier” on non-measurable covariates were both more likely to receive a beta-blocker and more likely to have better outcomes.

An instrumental variable (IV) analysis is one way to address this problem of residual confounding by using “natural” randomization to control for unmeasured confounding. In recent years, IV analyses have become increasingly common, particularly in situations such as this where the aim is to evaluate the average treatment effect of a standard therapy in an uncommon or understudied patient population.^12,13 Since a randomized controlled trial is unlikely to ever be done in this space, we aim to use an IV analysis to determine the association between beta-blocker use at/after HFrEF discharge and mortality and readmission rates and determine whether that association varies across the age spectrum.

METHODS

Study Cohort and Data Sources

We used the 100% national sample of patients enrolled in both Medicare Parts A and B and a random 40% sample of Part D enrollment to create a cohort of fee-for-service (FFS) beneficiaries with at least one hospitalization (index admission) for HFrEF between 2008 and 2016. Only the patient’s first hospitalization for HFrEF during the study period was included to avoid double counting patients. HFrEF was defined using International Classification of Diseases (ICD) 9 and 10 codes and distinguished from heart failure with preserved ejection fraction using previously validated methodology and are detailed in Appendix 1.^14,15 We required 1 year of FFS coverage prior to the index HFrEF admission to determine heart failure type, preexisting comorbidities, and exclusions. We required 1 year of FFS coverage after discharge (or until death) to determine outcomes. We also required 3 months of Part D coverage prior to the index admission to determine the medications that the patient was taking at the time of the index admission. The distribution of the number of index admissions per hospital is provided in Appendix 2.

Patients who died during admission were excluded because they were not eligible for beta-blocker therapy after discharge. Because they represent advanced HFrEF in which beta-blocker therapy may or may not be clinically feasible, those who underwent cardiac transplant or placement of a durable mechanical circulatory support (MCS) device during their index admission and those admitted from or discharged to hospice or with home inotropes were excluded. Patients with previously placed, durable MCS devices or a prior cardiac transplant were also excluded for the same reason.

Treatment Variable (Exposure)

We used Part D data and national drug codes (NDC) to determine beta-blocker exposure after hospitalization for HFrEF. We defined drug exposure as binary based on whether or not there was ≥ 1 fill for a HFrEF-specific beta-blocker (bisoprolol, carvedilol, or metoprolol succinate, as recommended by the HFrEF clinical guidelines¹⁶) within 30 days of hospital discharge.

Main Outcome Measure

The primary outcome of this study was all-cause mortality within 90 days of hospital discharge. The secondary outcome was all-cause readmissions within 90 days of discharge.

Instrumental Variable (IV) Selection

In order to support causal inference, an IV must satisfy three criteria (Fig. 1):

1. First, the instrument must have a strong association with the exposure (relevance).

2. Second, the effect of the instrument on the outcome must operate solely through the exposure variable (exclusion).

3. Third, the instrument must not share any common causes with the outcome (exchangeability).

Figure 1.

Instrumental variable analysis assumptions. This figure shows the key components and assumptions of the instrumental variable analysis model. We tested the relevance assumption with linear regression, by calculating F-statistics and examining the variation of the exposure across levels of the instrument. Neither the exclusion nor the exchangeability assumption can be empirically proven. However, we chose to further explore the exclusion assumption by measuring whether the instrument attenuated the imbalance of observed confounders between exposed and unexposed groups. For the exchangeability assumption, we contend that there is unlikely to be any common cause that affects both hospital-level rate of beta-blocker and patient-level outcomes.

For this analysis, we chose the hospital-level rate of beta-blocker use among eligible HFrEF patients.^17–21 We used Medicare data and the same HFrEF cohort to generate the instrument, but did so at the hospital (or facility) level. The use of a facility-level rate as an instrument has been successfully demonstrated in other studies from cardiology^{12,13,22–24} and pharmacoepidemiology^25–27 literature. As we were interested in whether the association between beta-blocker use and outcomes varied across the age spectrum, we performed stratified analyses by age: < 75 years old and ≥ 75 years old, as well as a pooled IV analysis with all eligible HFrEF beneficiaries.

Instrument Assessment

The first assumption of an IV analysis, the relevance assumption, can be empirically examined. To do this, we computed F-statistics for both age strata and overall.²⁸ Among those ≥ 75 years old, the F-statistic was 237; among those < 75 years old, it was 92; and overall, it was 343. In previous studies, an F-statistic > 10 has been shown to be sufficiently strong for use in IV analysis.²⁹

The second assumption of the IV analysis, the exclusion restriction, requires that the instrument not be independently correlated with the outcomes (90-day morality/readmission). This criterion is technically unverifiable, but as in other studies, we chose to examine exploratory evidence.^18,25,30,31 First, consistent with prior studies, we found that the IV analysis attenuated the imbalance of measured confounders between exposed and unexposed groups in our study (Appendix 3a–c). Next, we examined how the probability of receiving a beta-blocker after hospital discharge varied among the different quartiles of the instrument.³⁰ Patients admitted to a hospital in the highest IV quartile received a beta-blocker after discharge 70.1% more often than patients admitted to the lowest IV quartile (p < 0.001). For comparison, the rate of ACE inhibitor/angiotensin receptor blocker use varied only 4% (40 to 44%) across the instrument quartiles. Additional information about the gradient of beta-blocker use across IV quartiles for both age groups is available in Appendix 4.

The final requirement of the IV analysis is the exchangeability criterion. This requires that the instrument and the outcome not share any common causes. In thinking about this clinically, the only plausible “common cause” we could identify would be hospital characteristics and/or quality; i.e., “better” hospitals were both more likely to prescribe beta-blockers and do other things that improved patient outcomes. However, we compared the 90-day mortality rates of patients who received beta-blockers at high and low prescribe hospitals and found that the odds ratios for “beta-blocker fill” were very similar (0.4–0.5) across high and low prescribing hospitals. This also suggests that it is not differences in hospital characteristics or quality, i.e., other “common causes,” that is driving our results, but rather receipt of a beta-blocker after hospital discharge (Appendix 3c).

Statistical Methods

In this analysis, we used the two-stage least squares methodology. All models are adjusted for sex, race, dual eligibility, socioeconomic status, geographic region, number of inpatient/post-acute days in the year prior, number of hospitalizations in the year prior, beta-blocker use prior to hospitalization, the 29 Elixhauser comorbidities, and hospital-level fixed effects. Standard errors were estimated and clustered at the hospital level. A full description of the calculation of the IV estimator is included in Appendix 5.

In stage one, we used linear regression to predict beta-blocker fill within 30 days of discharge using the instrument and adjusted for all the patient-level covariates. We denote the respective variables by Death90_ij, BB_ij, IV_ij, and Cov_ij, where I denotes the ith patient to receive care at the jth hospital. The first-stage equation is given by:

$$\mathrm{B}{\mathrm{B}}_{\mathit{ij}}={\theta }_0+{\theta }_1\mathrm{I}{\mathrm{V}}_j+{\theta }_2\mathrm{Co}{\mathrm{v}}_{\mathrm{ij}}+{\mu }_j+{\delta }_{\mathit{ij}}$$

Consistent with prior studies, in the second stage, we used continuous estimates from the first-stage regression (continuous form of the IV)¹⁶ to predict 90-day mortality and/or readmissions. To do this, we estimated the equation using ordinary least squares (OLS) regression to obtain fitted values of BB_ij, denoted ${\widehat{\mathrm{BB}}}_{\mathit{ij}}={\widehat{\theta }}_0+{\widehat{\theta }}_1\mathrm{I}{\mathrm{V}}_j+{\widehat{\theta }}_2^T\mathrm{Co}{\mathrm{v}}_{\mathit{ij}}+{\widehat{\mu }}_j$. The second-stage equation is then given by:

$$\text{Death}{90}_{\mathit{ij}}={\beta }_0+{\beta }_1{\widehat{\mathrm{BB}}}_{\mathit{ij}}+{\beta }_2^T\mathrm{Co}{\mathrm{v}}_{\mathit{ij}}+{\lambda }_j+{\epsilon }_{\mathit{ij}}$$

Separate models were run for each age strata to test for effect modification by age. Finding none, we performed a pooled analysis using all HFrEF patients. All models are linear regressions and use the patient as the unit of analysis.³² To ensure adequate intra-hospital variability, hospitals with < 11 patients (based on CMS suppression rules) were excluded from these analyses. A sensitivity analysis using a cutoff of ≥ 25 index admission per hospital resulted in similar estimates, but the sample sizes were too small for stratified analyses (Appendix 6). Since the instrument was calculated at the facility-level, hospital characteristics by IV quartile are displayed in Appendix 7. Overall hospital characteristics were balanced across the IV quartiles. The only variable that reached significance was “hospital type” and this is likely an artifact of the small number of governmental (n = 2) and private/proprietary (n = 13) hospitals, compared to not-for-profit hospital (n = 97) in our analyses.

Analyses were preformed between January 2020 and January 2021. The study was approved by the Dartmouth College Institutional Review Board and is compliant with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for observational studies. p values are for two-sided hypothesis tests with values < 0.05 considered significant. All analyses were performed using SAS 9.4 and STATA version 15.1.

RESULTS

In total, we analyzed 136,456 patients aged ≥ 66 years old from 526 hospitals discharged after admission for HFrEF. Seventy percent of patients were aged 75+ years (n = 96,034) and 30% were aged < 75 years (n = 40,422). The baseline characteristics of patients in each age strata are presented in Table 1. The number of days inpatient/in post-acute care in the year prior were higher among those aged 75+ (7.61 days vs. 3.91 days, p < 0.001) and the rates of previously implanted cardiac defibrillators were lower among those aged 75+ (12.56% vs. 19.25%, p < 0.001). Comorbidity rates were similar between the two age strata (24.93% vs. 24.57% with ≥ 6 comorbidities and 2.97 vs. 2.19 with ≥ 10 comorbidities).

Table 1.

Baseline Characteristics of Patients Discharged After Admission for HFrEF, by Age

	Age strata
	Age < 75		Age 75+		Standard difference
	N	%/Mean	N	%/Mean
Sex					0.27
Male	23248	57.51	42268	44.01
Female	17174	42.49	53766	55.99
Race					0.27
White	32409	80.18	82728	86.14
Black	5135	12.70	7671	7.99
Other	2878	7.12	5635	5.86
Socioeconomics
Dual eligibility	13634	33.73	28369	29.54	− 0.09
% with bachelor’s degree*	40422	25.14	96034	29.34	0.27
% below the poverty line*	40422	16.66	96034	15.05	− 0.17
Geography					0.19
Midwest	11504	28.46	27635	28.78
Northeast	8110	20.06	26227	27.31
South	18481	45.72	36281	37.78
West	2327	8.76	5891	6.13
HFrEF disease severity
Number of days inpatient/in post-acute care in year prior	40422	3.91	96034	7.61	0.17
Number of hospitalizations in year prior	40422	1.07	96034	1.02	− 0.03
On beta-blocker prior to admission	22484	55.62	49906	51.97	− 0.07
Use of ACEi/ARB after hospital discharge	40422	0.42	96034	0.42	− 0.05
Implanted cardiac defibrillator	7783	19.25	12063	12.56	− 0.18
Comorbidities (Elixhauser)
6 or more comorbidities	10079	24.93	23594	24.57	− 0.01
10 or more comorbidities	1202	2.97	2101	2.19	− 0.05
Instrument quartile distribution					0.12
Instrument quartile 1	8945	22.13	24727	25.75
Instrument quartile 2	10037	24.83	24584	25.60
Instrument quartile 3	10367	25.65	23641	24.62
Instrument quartile 4	11073	27.39	23082	24.04
Beta-blocker fill after hospital discharge	22810	56.43	45767	47.66	− 0.18

*Analysis completed at the ZCTA level

HFrEF, heart failure with reduced ejection fraction; ACEi, angiotensin-converting enzyme inhibitor; ARB, angiotensin receptor blocker

Across the 4 quartiles of the instrument, the distribution of age, sex, race, socioeconomics, and geography was similar (Table 2). While the number of days inpatient/in post-acute care in the year prior was higher (7.51 days vs. 5.76 days, p < 0.001) among patients cared for at hospitals in the lowest instrument quartile, the distribution of comorbidity burden was similar (≥ 6 comorbidities Q1 26.35%, Q2 25.11%, Q3 23.92%, Q4 23.34%; ≥ 10 comorbidities Q1 2.72%, Q2 2.54%, Q3 2.24%, Q4 2.17%). Tables with complete listings of the individual Elixhauser comorbidities and the baseline characteristics of patients who either received or did not receive a beta-blocker (fill status) is displayed in Appendix 3.

Table 2.

Baseline Characteristics of Patients Discharged After Admission for HFrEF, by Instrument Quartile

	Instrumental variable quartiles
	Quartile 1		Quartile 2		Quartile 4		Quartile 4		Standard difference
	N	%/Mean	N	%/Mean	N	%/Mean	N	%/Mean
Age									− 0.08
66–74	8945	26.57	10037	28.99	10367	30.48	11073	32.42
75+	24727	73.44	24584	71.01	23641	69.52	23082	67.58
Sex									− 0.04
Male	15534	46.13	16638	48.06	16485	48.47	16859	49.36
Female	18138	53.87	17983	51.94	17523	51.53	17296	50.64
Race									0.10
White	28844	85.66	29584	85.45	28430	83.60	28279	82.80
Black	2677	7.95	2926	8.45	3454	10.16	3749	10.98
Other	2151	6.38	2111	6.09	2124	6.25	2127	6.22
Socioeconomics
Dual eligibility	10220	30.35	10224	29.53	10547	31.01	11012	32.24	0.04
% with bachelor’s degree*	33672	28.69	34621	28.67	34008	28.20	34155	26.81	− 0.07
% below the poverty line*	33672	14.54	34621	14.99	34008	15.87	34155	16.71	0.16
Geography									0.14
Midwest	10167	30.19	9875	28.52	9049	26.61	10048	29.42
Northeast	9444	28.05	9.58	27.03	8516	25.04	7019	20.55
South	11532	34.25	13802	39.87	14592	42391	14836	43.44
West	2529	7.51	1586	42.58	1851	5.44	2252	6.59
HFrEF disease severity
Number of days inpatient/in post-acute care in year prior	33672	7.51	34621	6.79	34008	6.01	34155	5.76	− 0.05
Number of hospitalization in year prior	33672	1.08	34621	1.03	34008	1.03	34155	1.00	− 0.03
On beta-blocker prior to admission	16646	4944	18160	52.45	18287	53.77	19297	56.50	0.08
Use of ACEi/ARB after hospitalization	33672	0.40	34621	0.40	34008	0.43	34155	0.44	0.31
Implanted cardiac defibrillator	4703	13.97	5072	14.65	4997	14.69	5074	14.86	0.01
Comorbidities (Elixhauser)
6 or more comorbidities	8874	26.35	8591	25.11	8134	23.92	7973	23.34	− 0.05
10 or more comorbidities	917	2.72	881	2.54	763	2.24	742	2.17	− 0.03
Beta-blocker fill after hospital discharge	12527	37.20	16250	46.94	18184	53.47	21616	63.29	0.33

*Analysis completed at the ZCTA level

HFrEF, heart failure with reduced ejection fraction; ACEi, angiotensin-converting enzyme inhibitor; ARB, angiotensin receptor blocker

Exposure and outcome rates, by age, are displayed in Figure 2a. Average instrument rates were similar (53.0% vs. 52.0%) between the two age strata. Instrument rates (rates of beta-blocker use after discharge) were, not surprisingly, higher among patients < 75 years old compared to patients 75+ years old (56.4% vs. 47.7, p < 0.001). Ninety-day mortality rates were also higher among patients aged 75+ years old compared to those < 75 years old (15.6% vs. 8.7%, p < 0.001). Ninety-day readmission rates were similar between the two age groups (36.2% among those < 75 years vs. 37.7% among those 75+ years old).

Figure 2.

a Exposure and raw outcome rates, by age. This figure shows rates of the instrument (hospital-level rates of beta-blocker use), exposure (patient-level rates of beta-blocker fills after hospital discharge), and patient-level outcomes stratified by age. BB is beta-blocker. b Exposure and raw outcome rates, by instrument quartile. This figure shows rates of the instrument (hospital-level rates of beta-blocker use), exposure (patient-level rates of beta-blocker fills after hospital discharge), and patient-level outcomes across the four quartiles of the instrument. Quartile 1 includes hospitals with the lowest rates of beta-blocker use in the year of analysis and quartile 4 includes hospitals with the highest rates of beta-blocker use after HFrEF discharge. BB is beta-blocker.

Exposure and outcome rates, by instrument quartile, are displayed in Figure 2b. As expected, rates of both the instrument (Q1 38.0%, Q2 49.0%, Q3 55.0%, Q4 65.0%) and the exposure (Q1 37.2%, Q2 46.9%, Q3 53.5%, Q4 63.3%) increase linearly across the 4 quartiles. The rates of 90-day mortality and readmission were highest among the highest instrument quartile (14.7% mortality, 38.3% readmission) and lowest among the lowest instrument quartile (12.7% mortality, 36% readmission). Exposure and outcome rates, by fill status, are included in Appendix 3b.

Among patients aged 75+ years old, receipt of a beta-blocker within 30 days of discharge was associated with a − 4.78% decrease (95% CI − 7.19 to − 2.40%, p < 0.001) in 90-day mortality (Table 3) and a − 4.67% (95% CI − 7.89 to − 1.45%, p = 0.005) decrease in 90-day readmissions (Table 4). Among patients aged < 75, mortality results were similar, receipt of a beta-blocker was associated with a − 5.83% (95% CI − 8.84 to − 2.81%, p < 0.001) decrease in mortality and a trend toward a decrease in 90-day readmissions as well (− 4.67%, 95% CI − 9.80 to + 0.04%, p = 0.075). As no effect modification by age was found, we also performed pooled analysis of all HFrEF patients and found that receipt of a beta-blocker after discharge was associated with a − 4.34% (95% CI − 6.27 to − 2.42%, p < 0.001) decrease in 90-day mortality and a − 4.66% (95% CI − 7.40 to − 1.91%, p = 0.001) decrease in 90-day readmission rates.

Table 3.

90-day Death Stratified by Age

Death within 90 days of hospital discharge
Age category	Coefficient	Standard error	95% confidence interval		p-value
< 75 years old (n = 40,422)	− 5.83%	1.54%	− 8.84%	− 2.81%	< 0.001
75+ years old (n = 96,034)	− 4.78%	1.22%	− 7.19%	− 2.4%	< 0.001
All patients (n = 136,456)	− 4.35%	0.98%	− 6.27%	− 2.42%	< 0.001

Table 4.

90-day Readmission Stratified by Age

Readmission within 90 days of hospital discharge
Age category	Coefficient	Standard error	95% confidence interval		p-value
< 75 years old (n = 40,422)	− 4.67%	2.62%	− 9.80%	+ 0.04	0.075
75+ years old (n = 96,034)	− 4.67%	1.64%	− 7.89%	− 1.45%	0.005
All patients (n = 136,456)	− 4.66%	1.40%	− 7.40%	− 1.91%	0.001

DISCUSSION

Based on robust clinical trials such as CIBIS-II,⁴ COPERNICUS⁷, and MERIT-HF,⁵ beta-blockers are now a cornerstone of HFrEF therapy and their use after HFrEF hospitalization is a class I recommendation for all patients,³³ regardless of age. However, in several key ways, today’s heart failure population is different from those who enrolled in the landmark beta-blocker trials two decades ago.³⁴ These demographic changes may have altered the risk/benefit ratio of standard therapies like beta-blockers.

Today, more than half of Medicare’s heart failure population is over age 75³⁵ and over two-thirds have multiple chronic conditions.³⁶ In contrast, in the CIBIS-II trial, the mean age was 61 years;⁴ in COPERNICUS, the mean age was 63 years⁷; and in MERIT-HF, it was 63 years.⁵ For most patients with HFrEF, beta-blockers are a net benefit. However, for older and often more medically complex patients, there is concern that the potential for mortality benefits is outweighed by the risk for harm due to side effects. Given the significant heterogeneity that exists within today’s older and more medically complex heart failure population,³⁷ there is, now more than ever, a pressing need to better understand the association between the routine use of beta-blocker therapy after HFrEF hospitalization and patient outcomes, and whether these associations vary by age.

Unfortunately, randomized data in this space is limited and additional randomized data is unlikely to ever be forthcoming. In 2005, Flather et al. investigated the utility of nebivolol, notably not a HFrEF-specific beta-blocker, in 2,128 patients aged ≥ 70 in the SENIORS trial. They found a HR = 0.86 (95% CI 0.74–0.99, p = 0.039) for the composite outcome of all-cause mortality or cardiovascular hospitalization among those exposed to nebivolol.⁸ However, both the choice of nebivolol and the age cutoff of 70 years limit the applicability of these results to a modern HFrEF population.

In the CIBIS-ELD study, the authors compared bisoprolol and carvedilol (both guideline-recommended HFrEF beta-blockers) in 883 patients ≥ 65 years of age.³⁸ The authors found that only 25% of patients were able to reach and maintain guideline-recommend doses of either beta-blocker, though tolerability rates did not vary significantly between the two beta-blockers. Unfortunately, because it enrolled fairly “young” HFrEF patients (≥ 65 years of age) and did not examine either short- or long-term outcomes, this study is unable to address the question of whether older patients with HFrEF who receive beta-blockers derive a similar net benefit to their younger counterparts.

To address this gap in knowledge, our group recently used Medicare claims data to examine the association between beta-blocker use after HFrEF hospitalization and short- and long-term outcomes.¹¹ To control for confounding by indication, inverse probability weighting across a myriad of demographic and clinical variables (including comorbidities, drug use, and prior healthcare utilization) was used. We found a strong net mortality benefit for beta-blockers in both the short (30-day mortality: OR ranged = 0.32–0.37) and long (1-year mortality: HR ranged = 0.35–0.52) terms. Moreover, we found that this benefit did not diminish with age. However, given the obligate retrospective study design and the lack of granular data, the study was limited by concerns for residual confounding by indication.

To address the concern for residual confounding, in this study, we used nationally representative observational data and an IV analysis to investigate the association between beta-blocker use after HFrEF hospitalization and outcomes and determine whether that association varied by age. We found that beta-blocker use after HFrEF hospitalization was associated with a significant decrease in mortality and readmission rates, across the age spectrum—importantly also among older adults ≥ 75 years old. These results support prior work, finding that, on net, beta-blocker therapy benefits outweigh the risks in both young and older HFrEF patients. While there are certainly clinically valid reasons why a patient might not be prescribed a beta-blocker at hospital discharge, such as low heart rate, soft blood pressure, syncope, or advanced disease,¹⁶ these findings support current clinical guidelines and quality metrics that recommend the use of beta-blockers at the time of HFrEF discharge, absent a clinical contraindication, regardless of age.

Limitations

Our primary limitations stem from the instrumental variable analysis method. Although we have tested the strength of the association between the instrument and exposure, the other assumptions of an IV analysis can only be evaluated by indirect methods, including clinical reasoning, rather than formal statistical testing. It is possible that hospitals that are more aggressive about giving beta-blockers also have better outcomes, but we find no evidence of this (see Appendix 3c). However, we are also reassured by prior work that has examined the strength of association between process measure adherence (e.g., giving a drug) and patient outcomes (e.g., readmissions/death) and found it to be weak.³⁹ In addition, our ability to detect small differences in outcome rates may be somewhat limited by the lower precision associated with instrumental variable approaches. This may account for the inability to detect a significant difference in readmission rates among the subgroup of HFrEF patients < 75 years old. Next, these estimates are at a population level and, as such, are not able to identify whether specific, smaller subgroups of older patients would have benefitted more or less than the average patient. Finally, our population is derived from longitudinal claims data and we required continuous enrollment in both fee-for-service and pharmacy coverage to enter the cohort. While this is commonly done too ensure completeness of follow-up, care should be used in applying the results to populations that might not meet the inclusion/exclusion criteria used in this study.

CONCLUSIONS

At a population level, patients aged ≥ 75 who receive a beta-blocker after HFrEF hospitalization have significantly lower 90-day mortality and readmission rates than those who do not. On net, beta-blocker therapy benefits outweigh the risks, even among older HFrEF patients. Thus, absent a clinical contraindication, beta-blocker therapy should at least be attempted in all HFrEF patients after hospitalization, regardless of age.

Funding

Dr. Gilstrap is supported by K23HL142835 from the National Heart Lung and Blood Institute. Drs. Austin, Skinner, Barnato, and O’Malley are supported, in part, by P01AG019783 from the National Institute on Aging. Dr. Tosteson is supported by UL1TR001086 from the National Center for Advancing Translational Sciences. Dr. Barnato is supported by the Susan J. and Richard M. Levy 1960 Academic Cluster in Health Care Delivery, administered by the Office of the Provost at Dartmouth College.

Declarations

Conflict of Interest

Dr. Skinner is a consultant for Sutter Health, and an investor in Dorsata Inc. The remaining authors have no relationships with industry to disclose.