Polypharmacy in Older Adults Hospitalized for Heart Failure

Abstract

Background:

Despite potential harm that can result from polypharmacy, real-world data on polypharmacy in the setting of heart failure (HF) are limited. We sought to address this knowledge gap by studying older adults hospitalized for HF derived from the Reasons for Geographic and Racial Differences in Stroke (REGARDS) study.

Methods:

We examined 558 older adults aged ≥65 years with adjudicated HF hospitalizations from 380 hospitals across the United States. We collected and examined data from the REGARDS baseline assessment, medical charts from HF-adjudicated hospitalizations, the American Hospital Association annual survey database, and Medicare’s Hospital Compare website. We counted the number of medications taken at hospital admission and discharge; and classified each medication as HF-related, non-HF cardiovascular-related, or non-cardiovascular-related.

Results:

The vast majority of participants (84% at admission and 95% at discharge) took ≥5 medications; and 42% at admission and 55% at discharge took ≥10 medications. The prevalence of taking ≥10 medications (polypharmacy) increased over the study period. As the number of total medications increased, the number of non-cardiovascular medications increased more rapidly than the number of HF-related or non-HF cardiovascular medications.

Conclusion:

Defining polypharmacy as taking ≥10 medications might be more ideal in the HF population as most patients already take ≥5 medications. Polypharmacy is common both at admission and hospital discharge, and its prevalence is rising over time. The majority of medications taken by older adults with HF are non-cardiovascular medications. There is a need to develop strategies that can mitigate the negative effects of polypharmacy among older adults with HF.

Introduction

With an expanding armamentarium of guideline-directed medical therapy (GDMT) to treat heart failure (HF), polypharmacy, broadly defined as the use of a high number of medications, has become increasingly relevant for older adults with HF. Indeed, with recent approval of sodium-glucose cotransporter 2 (SGLT-2) inhibitors for HF with reduced ejection fraction (HFrEF),¹ there are now 10 different classes of GDMT for HFrEF, many of which are recommended for concurrent use.^{2, 3} Older adults with HF also contend with multiple chronic conditions— over 60% of Medicare beneficiaries with HF have at least 5 other chronic medical conditions—which further contributes to their medication burden.⁴ Polypharmacy is important because it is associated with a myriad of adverse outcomes^{5, 6} including falls,^7–9 disability,^10–12 and hospitalizations.^{13, 14} This is especially relevant for older adults with HF, which is a population that is particularly vulnerable to the adverse effects of polypharmacy due to age-related alterations in pharmacokinetics and pharmacodynamics,¹⁵ changes in cardiovascular structure and function,¹⁶ and the coexistence of geriatric conditions like frailty¹⁷ and cognitive impairment.¹⁸

Despite the potential harm that can result from polypharmacy, real-world detailed information on polypharmacy in the setting of HF are limited. Characterizing medication patterns specifically around the time of a hospitalization is especially important because this is a period when medication errors and subsequent adverse drug events are common.¹⁹ We therefore sought to better understand polypharmacy in the context of a HF hospitalization by studying a cohort of older adults hospitalized for HF derived from the geographically-diverse Reasons for Geographic and Racial Differences in Stroke (REGARDS) study.

Methods

Study Population:

We examined adjudicated HF hospitalizations of Medicare beneficiaries aged at least 65 years who were hospitalized between 2003 and 2014. The study sample was a subset of participants from the REGARDS study, which is a national population-based prospective observational cohort of 30,239 community-dwelling black and white men and women aged ≥45 years from all 48 contiguous states in the United States and the District of Columbia originally recruited from 2003–2007.²⁰ The REGARDS study team used commercially-available lists of residents for recruitment. Baseline data collection included a 45-minute telephone interview followed approximately one month later by an in-home visit, when blood and urine samples as well as physiologic measures were obtained. For follow-up, participants received routine telephone calls every 6 months and reported if they had been hospitalized for a cardiovascular condition including HF. Two expert clinicians reviewed the hospitalization records of participants who reported a cardiovascular-related hospitalization to determine if the reason for hospitalization included a HF exacerbation, using a structured adjudication process. Adjudication was based on consideration of clinical presentation which included symptoms, physical exam findings, laboratory values, imaging, and medical treatments. In cases of disagreement, a committee made the final decision.

The medical records of participants hospitalized for HF were manually abstracted using a process previously described.²¹ We excluded participants who did not have medication data at both hospital admission and hospital discharge (N=132), and those who were referred to hospice at the time of discharge (N=25) (Supplemental Figure 1). Of note, prior work using this cohort has shown that there were few differences between individuals with medication data and those without medication data at both admission and discharge.²¹

All participants received information about the study and provided consent at the time of their enrollment. The REGARDS study previously had approvals from the Institutional Review Boards governing research in human subjects at the participating centers. Additionally, Weill Cornell Medicine Institutional Review Board approved this ancillary study.

Data Sources:

For this study, we used data from the REGARDS baseline assessment, medical charts from adjudicated HF hospitalizations, the American Hospital Association annual survey database, and Medicare’s Hospital Compare website.

REGARDS baseline data included sex, race, education level, household income, and geriatric conditions. Geriatric conditions included functional impairment (defined as physical component summary score from the Short Form 12 as <30), cognitive impairment (defined as a six-item screener score of <5)²², history of falls (defined as having at least one self-reported fall), and hypoalbuminemia (which has been described as a marker of frailty and is defined as albumin level ≤3.3 grams/deciliter).²³

Documents from each adjudicated HF hospitalization available for chart review included admission and progress notes, consultation notes, discharge summaries, medication reconciliation reports, laboratory and diagnostic testing reports, and nursing notes. For this study, we examined number of standing medications at admission and discharge (defining polypharmacy as taking at least 10 medications), echocardiogram parameters including left ventricular ejection fraction (LVEF), intensive care unit (ICU) stay, inpatient cardiology consultation, length of stay, discharge to home, and discharge year.

We also stratified patients by LVEF and examined them in two groups—we defined heart failure with reduced ejection fraction (HFrEF) as those with an LVEF<50% (inclusive of those with a mid-range or borderline ejection fraction) and/or a qualitative description of abnormal systolic function;^{2, 3, 24} and heart failure with preserved ejection fraction (HFpEF) as those with LVEF ≥50% and/or a qualitative description of normal systolic function.^{2, 3}

We classified medications according to the Multum Lexicon Drug Database,²⁵ which then facilitated a broader classification scheme of three major medication groups—HF medications, non-HF cardiovascular medications, and non-cardiovascular medications. These classifications were based on direct indications. For example, supplemental potassium was considered a non-cardiovascular medication because it is indicated for hypokalemia, even though hypokalemia often results from the use of diuretics (a HF medication).

We also used the survey database of the American Hospital Association, which collects data annually on approximately 6500 hospitals and over 400 health systems across the United States, to examine teaching status and hospital size (small hospital size defined as <200 beds); and Medicare’s Hospital Compare website that includes data about the quality of care of over 4000 Medicare-certified hospitals to examine hospital rating (range of 1–5, with 3 being average and a higher score of 5 reflecting higher quality care) corresponding to the year of hospital discharge.

Statistical analysis

We calculated medians and interquartile range (IQR) for continuous variables and percentages with 95% confidence intervals (CI) for categorical variables to summarize participant and hospital characteristics, comparing participants taking <10 medications with those taking ≥10 medications at both admission and discharge. We compared participant and hospital characteristics using the Mann Whitney U test for continuous variables and the Pearson chi-square for categorical variables. To assess for the unadjusted temporal trends of polypharmacy over the study period, we used Poisson regression with robust error variance.

We conducted a multivariable logistic regression analysis to identify the determinants of polypharmacy. We chose variables for the model based on published literature, prior knowledge, and clinical judgment; these included socio-demographics (age ≥75, sex, race, education, household income), heart failure subtype (HFpEF vs. HFrEF), geriatric conditions (functional impairment, cognitive impairment, history of falls, and hypoalbuminemia), a count of comorbid conditions, hospitalization factors (ICU stay, inpatient cardiology consultation, length of stay, discharge to home, and discharge year after 2010), and hospital characteristics (teaching status, hospital size, and Medicare hospital rating).

Results

We examined 558 older adults hospitalized for HF at 380 different hospitals across the United States (Supplemental Figure 1). The median age of the sample was 76 (interquartile range 72–83), 44% were female, and 34% were black (Table 1).

Table 1.

Population and Hospital Characteristics, According to Polypharmacy (at least 10 medications) at Admission and Discharge

		At Admission			At Discharge
	All (n=558)	< 10 Medications (n=322)	≥ 10 Medications (n=236)	P-value	< 10 Medications (n=252)	≥ 10 Medications (n=306)	P-value
SOCIODEMOGRAPHICS (%)
Age, median (±IQR)	76.0 (72.0–82.0)	76.5 (72.0–83.0)	76.0 (71.0–82.0)	0.27	76.0 (72.0–83.0)	76.0 (71.0–82.0)	0.48
Age ≥ 75 years	326 (58.4%)	195 (60.6%)	131 (55.5%)	0.23	151 (59.9%)	175 (57.2%)	0.51
Female	243 (43.5%)	138 (42.9%)	105 (44.5%)	0.70	104 (41.3%)	139 (45.4%)	0.32
Black Race	189 (33.9%)	116 (36.0%)	73 (30.9%)	0.21	87 (34.5%)	102 (33.3%)	0.77
Education high school or below	292 (52.3%)	167 (51.9%)	125 (53.0%)	0.91	135 (53.6%)	157 (51.3%)	0.92
Household income <$20,000	156 (28.0%)	96 (29.8%)	60 (25.4%)	0.48	147 (58.3%)	185 (60.5%)	0.36
GERIATRIC CONDITIONS (%)
Functional Impairment	117 (21.0%)	56 (17.4%)	61 (25.8%)	0.02	47 (18.7%)	70 (22.9%)	0.32
Cognitive Impairment	90 (16.1%)	54 (16.8%)	36 (15.3%)	0.38	46 (18.3%)	44 (14.4%)	0.12
History of falls	120 (21.5%)	59 (18.3%)	61 (25.8%)	0.03	47 (18.7%)	73 (23.9%)	0.13
Hypoalbuminemia	218 (39.1%)	128 (39.8%)	90 (38.1%)	0.56	101 (40.1%)	117 (38.2%)	0.89
CO-MORBIDITIES (%)
Comorbidity Count, median (IQR)	5 (4–7)	5 (3–6)	6 (5–8)	<.0001	5 (3–6)	6 (5–8)	<.0001
HFpEF	235 (42.1%)	136 (42.2%)	99 (41.9%)	0.84	103 (40.9%)	132 (43.1%)	0.77
Coronary Artery Disease	395 (70.8%)	216 (67.1%)	179 (75.8%)	0.02	166 (65.9%)	229 (74.8%)	0.02
History of Revascularization	214 (38.4%)	104 (32.3%)	110 (46.6%)	<0.001	77 (30.6%)	137 (44.8%)	<0.001
Atrial fibrillation/flutter	235 (42.1%)	126 (39.1%)	109 (46.2%)	0.10	106 (42.1%)	129 (42.2%)	0.98
Ventricular Arrhythmia	46 (8.2%)	27 (8.4%)	19 (8.1%)	0.89	16 (6.3%)	30 (9.8%)	0.14
Hypertension	431 (77.2%)	243 (75.5%)	188 (79.7%)	0.27	185 (73.4%)	246 (80.4%)	0.06
Diabetes	258 (46.2%)	126 (39.1%)	132 (55.9%)	<0.001	97 (38.5%)	161 (52.6%)	<0.001
CVA/TIA	102 (18.3%)	53 (16.5%)	49 (20.8%)	0.19	37 (14.7%)	65 (21.2%)	0.05
Peripheral Vascular Disease	108 (19.4%)	50 (15.5%)	58 (24.6%)	0.008	41 (16.3%)	67 (21.9%)	0.09
COPD	186 (33.3%)	91 (28.3%)	95 (40.3%)	0.003	67 (26.6%)	119 (38.9%)	0.002
Asthma	55 (9.9%)	25 (7.8%)	30 (12.7%)	0.05	20 (7.9%)	35 (11.4%)	0.17
Liver Disease	2 (0.4%)	1 (0.3%)	1 (0.4%)	0.83	1 (0.4%)	1 (0.3%)	0.89
Chronic Kidney Disease	201 (36.0%)	99 (30.7%)	102 (43.2%)	0.002	66 (26.2%)	135 (44.1%)	<0.001
Cancer	86 (15.4%)	52 (16.1%)	34 (14.4%)	0.57	42 (16.7%)	44 (14.4%)	0.46
Anemia	153 (27.4%)	71 (22.0%)	82 (34.7%)	<0.001	58 (23.0%)	95 (31.0%)	0.03
Osteoarthritis	148 (26.5%)	82 (25.5%)	66 (28.0%)	0.51	71 (28.2%)	77 (25.2%)	0.42
Benign prostatic hyperplasia	61 (10.9%)	29 (9.0%)	32 (13.6%)	0.09	22 (8.7%)	39 (12.8%)	0.13
HOSPITALIZATION DATA (%)
Intensive Care Unit Stay	116 (20.8%)	73 (22.7%)	43 (18.2%)	0.20	47 (18.7%)	69 (22.6%)	0.26
Cardiology Consultation	391 (70.1%)	231 (71.7%)	160 (67.8%)	0.32	175 (69.4%)	216 (70.6%)	0.77
Length of Stay, median days (IQR)	5.0 (3.0–8.0)	5.0 (3.0–8.0)	5.0 (3.0–7.0)	0.15	5.0 (3.0–8.0)	5.0 (3.0–8.0)	0.88
Discharge to Home	456 (81.7%)	264 (81.9%)	192 (81.4%)	0.39	217 (86.1%)	239 (78.1%)	0.01
Discharge year (after 2010)	137 (24.6%)	62 (19.3%)	75 (31.8%)	0.001	44 (17.5%)	93 (30.4%)	0.0004
HOSPITAL CHARACTERISTICS (%)
Teaching Hospital	270 (48.4%)	164 (50.9%)	106 (44.9%)	0.16	118 (46.8%)	152 (49.7%)	0.51
Small Hospital size (<200 beds)	135 (24.2%)	77 (23.9%)	58 (24.6%)	0.85	71 (28.2%)	64 (20.9%)	0.05
Hospital Rating below average (<3)	179 (32.1%)	112 (34.8%)	67 (28.4%)	0.08	82 (32.5%)	97 (31.7%)	0.63

COPD, Chronic Obstructive Pulmonary Disease; CVA, Cerebrovascular Accident; HFpEF, Heart failure with preserved ejection fraction; IQR, Interquartile range; TIA, Transient Ischemic Attack

Histograms showing the frequency of medication counts at hospital admission and discharge are shown in Figure 1A and 1B. The vast majority of older adults with HF took at least 5 medications (84% at admission and 95% at discharge), and approximately half took at least 10 medications, which we define here as polypharmacy (42% at admission and 55% at discharge). These patterns were similar for those with HFrEF and HFpEF (Supplemental Figures 2 and 3). The median number of medications taken by those with polypharmacy at admission was 12 (IQR 11–14) and at discharge was 12 (IQR 11–14); and the median number of medications taken by those without polypharmacy at admission was 6 (IQR 4–8) and at discharge was 7 (IQR 6–9). Those with polypharmacy had a higher median comorbidity count on both admission and discharge and higher prevalence of functional impairment upon admission. There were no significant differences in age, race, sex, education level, or household income between those with and without polypharmacy at either admission and discharge (Table 1). We found similar patterns for both HFrEF and HFpEF (Supplemental Table 1).

Figure 1. Distribution of medication count.

Polypharmacy, defined as the condition of taking at least 10 medications, occurred in 42% of adults with heart failure at hospital admission (A) and 55% at hospital discharge (B)

A. At Hospital Admission

B. At Hospital Discharge

The prevalence of polypharmacy at hospital admission increased from 25% in 2003–2006 to 55% in 2011–2014 (p for trend<0.0001), and at hospital discharge increased from 41% in 2003–2006 to 68% in 2011–2014 (p for trend<0.0001) (Figure 2). Similar trends were observed when HFrEF and HFpEF participants were analyzed separately, with a higher numerical increase among those with HFpEF (Supplemental Figure 4). The increase in medication count over time was greatest among non-cardiovascular medications at both admission (Figure 3A) and discharge (Figure 3B).

Figure 2. Temporal Trends of Polypharmacy.

Polypharmacy, defined as the condition of taking at least 10 medications, at both admission and discharge increased in prevalence between the 2003–2006 period and 2011–2014 period.

Figure 3. Median Number of Medications by Medication Classes Over Time.

The median number of medications increased over time. This increase was most notable among non-cardiovascular (non-CV) medications. Abbreviations: CV, cardiovascular; HF, heart failure; Non-CV, Non-cardiovascular

A. At Hospital Admission

B. At Hospital Discharge

Between hospital admission and hospital discharge, the prevalence of polypharmacy increased from 42% to 55% (p<0.001). Between admission and discharge, the median number of all medications increased from 9.0 to 10.0 (p<.0001), the median number of HF medications increased from 2 to 2.5 (p<.0001), the median number of non-HF cardiovascular medications increased from 2 to 3 (p<.0001), and the median number of non-cardiovascular medications increased from 4 to 4.5 (p<.0001). The most commonly initiated medications are shown in Table 2. As shown, loop diuretics, beta-blockers, aspirin, and electrolyte supplements were the most commonly initiated medications between admission and discharge. These patterns were similar for both HFrEF and HFpEF (Supplemental Table 2 and Supplemental Figure 5).

Table 2.

Most Common Medications Initiated Between Admission and Discharge

	Medication	Number (Percentage)*
1	Loop Diuretics	149 (27%)
2	Beta-blockers	116 (21%)
3	Aspirin	105 (19%)
4	Electrolyte Supplements	97 (17%)
5	ACE inhibitors	95 (17%)
6	Quinolones	67 (12%)
7	Anticoagulation	67 (12%)
8	Proton Pump Inhibitors	64 (11%)
9	Statins	61 (11%)
10	P2Y12 inhibitors	57 (10%)
11	Adrenergic Bronchodilators	54 (10%)
12	Multivitamins	46 (8%)
13	Nitrates	44 (8%)
14	Anti-arrhythmic	44 (8%)
15	MRA	36 (6%)

ACE, angiotensin converting enzyme

MRA, mineralocorticoid receptor antagonist

^*Percentages calculated by using total number of patients (N=558)

The most common class of medications at both admission and discharge were non-cardiovascular medications. Upon admission, non-cardiovascular medications comprised a median of 50.0% (IQR 33.3–66.7%) of medications, while HF medications comprised a median of 20.0% (IQR 11.1–30.8%) and other non-HF cardiovascular medications comprised a median of 25.0% (16.7–36.4%). At discharge, non-cardiovascular medications comprised a median of 46.7% (IQR 33.3–60%) of medications, while HF medications comprised a median of 25.0% (IQR 16.7–33.3%) and other non-HF cardiovascular medications comprised a median of 26.7% (IQR 17.6–36.4%). These patterns were similar for both HFrEF and HFpEF (Supplemental Figure 5). As the total medications at both admission and discharge increased, the number of non-cardiovascular medications increased at a faster rate than the number of either HF medications or non-HF cardiovascular medications at both admission and discharge (Figure 4). This was consistent for both HFrEF (Supplemental Figure 6) and HFpEF (Supplemental Figure 7). Table 3 shows the most common HF medications, non-HF cardiovascular medications, and non-cardiovascular medications taken at admission and discharge. As shown, beta-blockers were the most common HF medications; aspirin and statins were the most common non-HF cardiovascular medications; and proton pump inhibitors, electrolyte supplements, and multivitamins were the most common non-cardiovascular medications taken. These patterns were consistent for both HFrEF and HFpEF (Supplemental Table 3). Although not among the top 5 most common HF medications, it was notable that mineralocorticoid receptor antagonists were prescribed to 38 (6.8%) patients on admission and 64 (11.5%) patients on discharge.

Figure 4. Medication Class Counts According to Total Medication Count.

As the number of total medications increased, the number of non-cardiovascular (non-CV) medications increased more rapidly compared to either heart failure (HF) medications or non-HF cardiovascular (CV) medications at both admission (A) and discharge (B). Abbreviations: CV, cardiovascular; HF, heart failure; Non-CV, Non-cardiovascular

A. At Hospital Admission

B. At Hospital Discharge

Table 3.

Most Common Medications Prescribed at Admission and at Discharge, Stratified by Medication Class

HF*						Non-HF Cardiovascular*					Non-Cardiovascular*
	Admission		Discharge			Admission		Discharge			Admission		Discharge
1	Beta-blocker	333 (60%)	Beta-blocker	411 (74%)	1	Aspirin	277 (50%)	Aspirin	349 (63%)	1	PPI	177 (32%)	PPI	218 (39%)
2	Loop Diuretic	264 (47%)	Loop Diuretic	372 (67%)	2	Statin	275 (49%)	Statin	306 (55%)	2	Electrolyte Supplements	152 (27%)	Electrolyte Supplements	213 (38%)
3	ACE Inhibitor	186 (33%)	ACE Inhibitor	233 (42%)	3	CCB	171 (31%)	CCB	155 (28%)	3	Multi-vitamin	151 (27%)	Multi-vitamin	152 (27%)
4	ARB	108 (19%)	Nitrate	119 (21%)	4	A/C	134 (24%)	A/C	175 (27%)	4	Thyroid Hormone	105 (19%)	Adrenergic Bronchodilator	108 (19%)
5	Nitrate	85 (15%)	ARB	98 (18%)	5	P2Y12 Inhibitors	99 (18%)	P2Y12 Inhibitors	140 (25%)	5	SSRI	82 (15%)	Thyroid Hormone	102 (18%)

^*Percentages calculated by using total number of patients (N=558)

A/C, Anticoagulation; ACE, Angiotensin Converting Enzyme; ARB, Angiotensin Receptor Blocker; CCB, Calcium Channel Blocker; HF, Heart Failure; PPI, Proton Pump Inhibitors; SSRI, Selective Serotonin Reuptake Inhibitors

In the multivariable regression analysis identifying determinants of polypharmacy at discharge, each additional comorbid condition increased the relative risk of taking at least 10 medications at discharge by 13% (RR: 1.13 per comorbid condition, 95% CI [1.08, 1.19]) (Table 4). Interestingly, geriatric conditions including functional impairment (RR: 1.00, 95% CI [0.75, 1.32]), cognitive impairment (RR: 0.85, 95% CI [0.60, 1.18]), history of falls (RR: 1.15, 95% CI [0.88–1.51]) and hypoalbuminemia (RR: 0.92, 95% CI [0.71–1.18]) were not independently associated with polypharmacy.

Table 4.

Determinants of Polypharmacy at Discharge

Variable	Univariable Model		Multivariable Model
	Relative Risk (95% CI)	P-value	Relative Risk (95% CI)	P-value
Population Characteristics
Age ≥75	0.95 (0.82, 1.11)	0.51	0.98 (0.78, 1.24)	0.88
Female	1.08 (0.93, 1.25)	0.32	1.17 (0.91, 1.50)	0.23
Black Race	0.98 (0.83, 1.15)	0.77	0.95 (0.73, 1.24)	0.71
Education (High School or below)	0.96 (0.83, 1.12)	0.60	1.03 (0.80, 1.32)	0.82
Household Income (<$20,000)	0.92 (0.77, 1.10)	0.37	0.90 (0.67, 1.20)	0.47
Heart Failure with Preserved Ejection Fraction	1.03 (0.89, 1.21)	0.67	1.00 (0.79, 1.28)	0.97
Functional Impairment	1.10 (0.92, 1.30)	0.31	1.00 (0.75, 1.32)	0.98
Cognitive Impairment	0.84 (0.67, 1.06)	0.14	0.85 (0.60, 1.18)	0.33
History of Falls	1.15 (0.97, 1.36)	0.11	1.15 (0.88, 1.51)	0.31
Hypoalbuminemia	0.99 (0.83, 1.18)	0.88	0.92 (0.71, 1.18)	0.50
Comorbidity Count	1.14 (1.10, 1.17)	<0.0001	1.13 (1.08, 1.19)	<.0001
Intensive Care Unit Stay	1.11 (0.93, 1.32)	0.24	1.04 (0.78, 1.38)	0.79
Cardiology Consultation	1.03 (0.87 1.21)	0.77	0.91 (0.70, 1.18)	0.48
Length of Stay (per day)	1.01 (1.01, 1.01)	<0.0001	1.00 (0.99, 1.00)	0.66
Discharge to home	0.77 (0.66, 0.91)	0.002	0.86 (0.63, 1.18)	0.35
Discharge year (after 2010)	1.34 (1.16, 1.56)	0.0001	1.26 (0.98, 1.63)	0.07
Hospital Characteristics
Teaching Hospital	1.05 (0.91, 1.22)	0.51	0.95 (0.74, 1.23)	0.70
Small Hospital Size (<200 beds)	1.21 (0.99, 1.47)	0.06	1.25 (0.92, 1.70)	0.15
Hospital Rating Below Average (<3)	0.92 (0.78, 1.09)	0.33	0.96 (0.73, 1.25)	0.74

Discussion

During an era when the armamentarium for managing HF continues to expand, the issue of polypharmacy is becoming increasingly relevant for older adults with HF. In this real-world study of medication patterns in older adults hospitalized with HF, there were several important findings. First, we found that almost everyone took at least 5 medications and over half took at least 10 medications, underscoring the high prevalence of polypharmacy in this population. Second, we found that the prevalence of polypharmacy has increased over time, reflecting an urgent need to develop new strategies for managing high medication burden. Finally, we found that the majority of medications prescribed to older adults with HF were non-cardiovascular in nature, supporting the need for processes that can optimize prescribing practices for this subset of medications.

Polypharmacy is broadly defined as the condition of taking a high number of medications. A cutoff of 5 has been arbitrarily chosen to differentiate those with polypharmacy and those without polypharmacy.^{26, 27} This cutoff is logical for the broader population because there is evidence that the number of adverse events increases beyond 5 medications.¹⁴ In our study, we found that the vast majority of older adults with HF took at least 5 medications. This is important because this suggests that older adults with HF are inherently at risk for adverse drug events, and reinforces the idea that polypharmacy should routinely be considered when caring for older adults with HF.^{28, 29} It also suggests that using a cutoff of 5 medications to risk-stratify older adults with HF for the potential harms of a high medication burden may not be optimal in this population. Given our observation that about 50% of older adults with HF take at least 10 medications coupled with the known link between this extreme version of polypharmacy^30–32 and an incrementally increased risk of adverse events,^{12, 14, 33} it may be reasonable to use a cutoff of 10 medications to define polypharmacy in older adults with HF, which can identify individuals at highest risk for adverse events related to high medication burden.

Using a cutoff of 10 medications, we found that the prevalence of polypharmacy in this population at both hospital admission and hospital discharge has risen significantly between the early period of the REGARDS study (2003–2006) and later period (2011–2014). While this could represent better guideline-concordant therapy for the many medical conditions that older adults contend with (comorbidity burden was the only factor in our analysis that was independently associated with polypharmacy), this is also potentially concerning because a high medication burden contributes to an increased risk for adverse drug reactions,¹⁴ and is associated with a number of adverse outcomes including falls,^7–9 disability,^10–12 and hospitalizations.^{13, 14} High medication burden is also associated with treatment burden,^{34, 35} reduced medication adherence,³⁶ and low quality of life.³⁷ Issues like therapeutic competition, defined as the clinical situation where a medication that treats one condition is harmful for another condition,³⁸ and competing health priorities³⁹ further complicate optimizing medication regimens in older adults with HF, and support the need to develop evidence-based patient-centered strategies for managing patients with polypharmacy and, relatedly, multiple chronic conditions. Tinetti et al recently outlined a framework and formal process that can be used when caring for older adults with multiple chronic conditions (and polypharmacy) called Patient Priorities Care⁴⁰ which involves identifying, communicating, and providing care consistent with patients’ health priorities. This approach has shown some promise for addressing challenges in decision-making faced by older adults with multiple chronic conditions (and polypharmacy) when competing priorities are present. Whether this model can be applied to older adults with HF is unknown and warrants further investigation.

Irrespective of the total medication count, we found that non-cardiovascular medications comprised the majority of medications taken by older adults with HF. This directly reflects the significant comorbidity burden that most older adults with HF contend with; approximately 90% of older adults with HF have at least 3 other medical conditions, and over 60% have at least 5 other medical conditions.⁴ It may be hard to argue that the risks of polypharmacy outweigh the benefits of guideline-based care for conditions like chronic obstructive pulmonary disease and/or diabetes which frequently require multiple pharmacologic agents. Accordingly, continuing or starting medications that directly treat these conditions may be reasonable, and moreover support the development of medications that can concurrently treat multiple conditions such as SGLT-2 inhibitors which can treat both diabetes and HF.^{1, 41, 42} On the other hand, the value of starting and/or continuing proton pump inhibitors (PPIs) and multivitamins, which we found represented two of the top three most common non-cardiovascular medications taken at admission and discharge, is debatable. Indeed, PPIs have been described as one of the most over-prescribed medications for older adults;⁴³ and the data supporting the use of multivitamins are limited.⁴⁴ Thus, to address polypharmacy in this population, it may be reasonable to consider eliminating these agents in the absence of compelling indications. In addition to reducing medication burden, eliminating agents with limited benefit may have important implications on the use of GDMT for HF, which has persistently been underutilized in the United States and requires novel strategies to improve its uptake.⁴⁵ Indeed, polypharmacy has been described as an important physician-reported barrier to initiating GDMT,⁴⁶ is associated with underprescribing,⁴⁷ and is an important predictor of reduced adherence.³⁶ Whether eliminating non-cardiovascular medications with limited benefit can improve prescribing rates of GDMT and/or improve patient adherence by reducing medication burden is unknown and warrants further investigation.

We also found that electrolyte supplements represented one of the top 3 most common non-cardiovascular medications at both admission and discharge. Electrolyte supplements like potassium are often used to treat hypokalemia, a well-known effect of diuretic therapy that is associated with adverse outcomes.⁴⁸ This observation highlights yet another potential opportunity to improve GDMT. For individuals requiring potassium supplements, initiation of a mineralocorticoid antagonist, which is considered a part of GDMT for HF, could provide an alternative strategy to combat hypokalemia. Consistent with prior studies, the use of mineralocorticoid antagonists in our study was low,⁴⁹ despite their potential to improve mortality in HFrEF and reduce hospitalizations (and possibly mortality) in HFpEF.^{2, 3} Future study is thus warranted to explore whether increased focus on managing non-cardiovascular medications (discontinuing potassium supplementation) can improve rates of GDMT and thereby improve outcomes in HF.

There are several strengths to this study. REGARDS is a national sample that includes participants and hospitals from all regions of the contiguous 48 United States and thus has a high degree of generalizability to the United States population.⁵⁰ Another strength was the availability of chart-level medical record data for each hospitalization which allowed detailed data collection of medications and comorbid conditions. There are also some important limitations to our study. This study was observational in nature and thus precluded establishing a causal relationship between variables. This study examined older adults with Medicare, and may not be generalizable to younger individuals and those without Medicare. While medication data were obtained at the time of the hospitalization, some participant characteristics including geriatric conditions were obtained at the time of study enrollment which are subject to change by the time of their hospitalization. The number of medications did not account for multiple pharmacologically-active ingredients in a single pill (e.g. combination pills), and did not account for pill burden (multiple pills of the same medication). We also did not account for dosages or the schedule for medications. All of these aspects can contribute further to medication burden and regimen complexity, and thus warrant future investigation. Finally, this study was conducted prior to the approval of angiotensin receptor-neprilysin inhibitors, ivabradine, and sodium-glucose cotransporter 2 inhibitors. Consequently, the prevalence of polypharmacy and the contribution of HF medications to polypharmacy are likely even more substantial today than shown in this study.

In conclusion, our study showed that polypharmacy (defined as at least 10 medications) is common among hospitalized HF patients and its prevalence is rising over time, supporting the use of ≥10 medications as a marker for high risk in this population. Our study also showed that the majority of medications taken by older adults with HF are non-cardiovascular medications, some of which may have limited benefit. Taken together, our findings support the need to develop strategies to mitigate the negative effects of polypharmacy among older adults with HF, potentially starting with formalized processes that can improve prescribing practices for non-cardiovascular medications.

Clinical Implications.

What is new?

• Polypharmacy, defined as taking ≥10 medications here, is common in older patients hospitalized for heart failure.

• The prevalence of polypharmacy increases after a hospitalization for heart failure; and has increased between 2003 and 2014.

• The majority of medications prescribed to older adults with heart failure are non-cardiovascular medications.

What are the Clinical Implications?

• A cutoff of 10 medications to identify polypharmacy may be optimal for identifying older adults with heart failure at greatest risk for harms related to high medication burden

• Strategies are needed to mitigate the potential negative effects of polypharmacy in older adults with heart failure.

• Reconciliation and optimization of non-cardiovascular medications could help reduce medication burden, and indirectly impact the use of guideline-directed medical therapy for heart failure