Abstract
Importance:
Efforts to improve heart failure (HF) outcomes have focused on prescribing guideline-directed medical therapy at hospital discharge. Whether new prescriptions translate to sustained medication use after HF hospitalization is largely unknown.
Objective:
To characterize medication use patterns following HF hospitalizations.
Design:
Observational cohort study
Setting:
Regional US health system
Participants:
Patients discharged home after a HF hospitalization between July 1, 2017 and March 30, 2023.
Main Outcomes and Measures:
Medications of interest included beta-blockers, renin-angiotensin system inhibitors (RASis), mineralocorticoid receptor antagonists (MRAs), and sodium-glucose co-transporter 2 inhibitors (SGLT2is). New prescriptions were identified by in-hospital administration or discharge medication orders. Using electronic health records linked to pharmacy dispensation records, medication initiation (prescription dispensation) and primary non-adherence (discharge prescriptions which were not filled) were assessed at 7- and 90-days post-discharge. Adherence (proportion of days covered ≥80%), and persistence (continuous days’ supply) were assessed over 6 months post-discharge.
Results:
The cohort included 6,111 HF hospitalizations (mean (SD) age 70 (14) years, 37% female, 79% White, 19% Black) of whom 51% of were prescribed at least one new HF medication at discharge. At admission, 58% of patients were prescribed beta-blockers, 41% RASis, 16% MRAs, and 5% SGLT2is, and upon discharge, use increased to 73%, 53%, 28%, and 9%, respectively. Primary non-adherence was observed in 51% of new prescriptions for beta-blockers, 48% for RASis, 39% for MRAs, and 35% for SGLT2is. Thus, of 4,873 new discharge prescriptions, only 54% were initiated within 7 days of discharge and an additional 20% had delayed initiation, between 7- and 90-days post-discharge. At 6 months, persistence to beta-blockers was 70%, RASis was 60%, MRAs was 55%, and SGLT2is was 56%. As a result, at 6 months, only 42% of patients were adherent and 51% were persistent to all HF medications used at discharge.
Conclusions and Relevance:
Following HF hospitalization most new guideline-directed medical therapy prescriptions went unfilled. Moreover, most newly initiated prescriptions did not persist at 6 months, indicating a need to develop interventions to support patients’ medication use beyond the initial prescription at discharge.
Keywords: pharmacoepidemiology, heart failure, medication adherence, post-discharge care
Heart failure (HF) accounts for 1.2 million hospitalizations annually (1). In 2017, the American Heart Association, American College of Cardiology, and Heart Failure Society of America (AHA/ACC/HFSA) recommendations defined guideline directed medical therapy (GDMT) for HF with reduced ejection fraction (HFrEF) as the combination of three medication classes: beta-blockers (βBs), renin-angiotensin system inhibitors (RASis), and mineralocorticoid receptor antagonists (MRAs) (2). In 2022, guideline recommendations were updated to include sodium-glucose co-transporter-2 inhibitors (SGLT2is) and to preference angiotensin receptor-neprilysin inhibitors (ARNIs) over other RASi agents (3).
Following a HF hospitalization, patients have a heightened risk of readmission and mortality (4). Multiple trials have demonstrated that hospital discharge offers an opportunity for titration and optimization of GDMT medications to improve clinical outcomes (5). While starting GDMT medications at hospital discharge is an important first step, ensuring adherence and persistence to newly initiated treatments after hospitalization is critical to improve patient outcomes (6).
Developing contemporary evidence on GDMT medication utilization patterns following hospitalization may inform strategies to improve HF outcomes. Therefore, we used a unique linkage of electronic health records from a large regional health system with prescription and pharmacy fill data to describe GDMT medication initiation, adherence, and persistence throughout the 6 months following HF hospitalizations.
Methods
We conducted an observational cohort study of adults hospitalized for HF across 32 hospitals from a regional health system spanning 3 states. Electronic health record data was linked to Surescripts pharmacy dispensation data for cohort construction and outcome assessment. The University of Pittsburgh Institutional Review Board determined this study as exempt.
Study Cohort
The study cohort included patients hospitalized with a primary discharge diagnosis of HF who were discharged home between July 2017 to March 2023 (eMethods, eFigure 1). We then identified the most recent ejection fraction (EF) to restrict to patients with HFrEF (EF <40%) or HFmrEF (EF 40–49%). Patients with multiple hospitalizations were treated as separate observations as each hospitalization serves as an opportunity for new prescribing decisions. Additional exclusion criteria are included in the eMethods.
GDMT Medication Use on Admission
We defined the four GDMT medication classes according to the Class 1 recommendations from the 2017 and 2022 AHA/ACC/HFSA guidelines for HFrEF and Class 2A (SGLT2i) and 2B (βBs, RASis, and MRAs) recommendations for HFmrEF (2, 3). We assessed use of all four GDMT medication classes throughout the study period but measured ‘triple/quadruple’ treatment according to the guidelines applicable at the time of discharge. We defined each medication class as present on hospital admission if 125% of the days’ supply of the most recent preceding pharmacy fill was sufficient to last through the hospital admission date (7, 8).
Study Outcomes
GDMT Prescription
We identified patients who were newly prescribed a GDMT medication class as those not using the class on admission and with either: in-hospital administration within the two days prior to discharge, a new prescription recorded in discharge medication orders, or an outpatient pharmacy fill within 7 days following discharge. We validated this algorithm in a chart review of 100 randomly sampled patients (eMethods).
GDMT Initiation
Among patients newly prescribed GDMT, we defined those with an outpatient pharmacy fill within 7 days post-discharge as ‘initiators’ and the remainder as exhibiting ‘primary non-adherence’ (9) (eFigure 2). We defined ‘late initiators’ as those who filled a prescribed GDMT class between 8 and 90 days of discharge. We assumed patients with a GDMT medication class present on admission were ‘continuers’ of that class at discharge.
GDMT Adherence and Persistence
Among GDMT initiators and continuers, we calculated medication adherence and persistence in the 6 months following discharge. We defined medication ‘adherence’ using proportion of days covered (PDC) and categorized patients as ‘adherent’ to each class, if they had on PDC ≥80% (10). We defined medication ‘persistence’ as consecutive days with medication on hand without any gaps in days’ supply greater than 45 days (11). To assess persistence at 6 months and time to non-persistence, we extended our follow-up period by an additional 45 days to account for possible dose titrations, stockpiling, and drug holidays (10) (eMethods).
Baseline Characteristics and Contraindications
We measured sociodemographic characteristics (e.g., age, sex, race and ethnicity, rurality, neighborhood-level resources), clinical factors (e.g., comorbidities, laboratory results, vitals), baseline medication use, and hospital-level factors (e.g. rurality, size, cardiology fellowship site). Operational definitions are described in the eMethods.
Potential contraindications to use of each GDMT medication class identified prior to discharge included: hypotension (systolic blood pressure <100) and bradycardia (hear rate < 60) for βBs, hypotension, hyperkalemia (potassium > 5 mEq/L), and estimated glomerular filtration rate (eGFR) <30 for RASis and MRAs, and eGFR <30 for SGLT2is (12).
Statistical Analysis
We first present descriptive statistics of the cohort characteristics and use of GDMT medications prior to admission. Second, we examined new prescriptions and initiations of GDMT, comparing the proportion of patients using of GDMT on admission and following discharge using McNemar’s test. Third, we assessed adherence and persistence using Fine and Gray models to compare time to non-persistence between continuers and initiators for each GDMT medication class while accounting for competing risk of death. We also report overall patient-level adherence and persistence measures to summarize use across GDMT medication classes (eMethods). Fourth, we classified non-users of each GDMT medication class by the presence of contraindications. Fifth, we describe late initiation of GDMT medication classes at 90-days post-discharge. We compared late initiation by presence or absence of any contraindication using chi-square tests of proportions. Sixth, we constructed multivariable Poisson regression models to evaluate the associations between cohort characteristics (aforementioned sociodemographic, clinical, medication use, and hospital-level factors) and the outcomes of a) primary non-adherence and b) non-persistence at 6 months to at least one discharge medication. Models included fixed effects for hospital site and patient-clustered robust standard errors, and we report risk ratios with 95% confidence intervals.
We conducted three sensitivity analyses varying definitions for baseline medication use, persistence, and adherence (eMethods). We determined statistical significance with an ⍺ < 0.05. Data were analyzed using R version 4.4.0.
Results
The cohort included 6,111 HF hospitalizations (mean age 70 years (SD 14), 37% female, 79% White, 19% Black) of whom 74% had HFrEF and 26% HFmrEF (Table 1).
Table 1.
Baseline characteristics of hospitalized patients with heart failure with reduced and mildly reduced ejection fraction
| Overall | HFrEF | HFmrEF | |
|---|---|---|---|
| N | 6111 | 4521 | 1590 |
| Sociodemographic Characteristics | |||
| Age at admission, mean (SD) | 69.9 (13.6) | 68.8 (13.8) | 72.9 (12.6) |
| Male, % | 3834 (62.7) | 2963 (65.5) | 871 (54.8) |
| Race, % | |||
| African American | 1153 (18.9) | 941 (20.8) | 212 (13.3) |
| Other Race | 39 (0.6) | 25 (0.6) | 14 (0.9) |
| White | 4833 (79.1) | 3485 (77.1) | 1348 (84.8) |
| Hispanic or Latino/a ethnicity, % | 42 (0.7) | 35 (0.8) | 7 (0.4) |
| Non-Hispanic or Latino/a ethnicity, % | 5784 (94.6) | 4266 (94.4) | 1518 (95.5) |
| Insurance, % | |||
| Medicare | 4658 (76.2) | 3355 (74.2) | 1303 (81.9) |
| Medicaid | 739 (12.1) | 599 (13.2) | 140 (8.8) |
| Other/Self-Pay | 106 (1.7) | 80 (1.8) | 26 (1.6) |
| Private | 523 (8.6) | 417 (9.2) | 106 (6.7) |
| RUCA Category, % | |||
| Metropolitan | 4374 (71.6) | 3277 (72.5) | 1097 (69.0) |
| Micropolitan | 962 (15.7) | 690 (15.3) | 272 (17.1) |
| Small Town/Rural | 634 (10.4) | 443 (9.8) | 191 (12.0) |
| Area Deprivation Index Quintile, % | |||
| 1–2 (Least disadvantaged) | 410 (6.7) | 307 (6.8) | 103 (6.5) |
| 3–4 | 716 (11.7) | 487 (10.8) | 229 (14.4) |
| 5–6 | 984 (16.1) | 726 (16.1) | 258 (16.2) |
| 7–8 | 1587 (26.0) | 1136 (25.1) | 451 (28.4) |
| 9–10 (Most disadvantaged) | 2133 (34.9) | 1635 (36.2) | 498 (31.3) |
| Hospitalization Characteristics | |||
| Length of Stay, days, mean (SD) | 6.3 (5.8) | 6.5 (6.0) | 5.9 (5.2) |
| Discharge date, % | |||
| After 2017 Guideline Update | 3024 (49.5) | 2250 (49.8) | 774 (48.7) |
| After SGLT2i FDA approval for HFrEF | 2144 (35.1) | 1552 (34.3) | 592 (37.2) |
| After 2022 Guideline Update | 943 (15.4) | 719 (15.9) | 224 (14.1) |
| Clinical Factors | |||
| High Frailty Risk Score, % | 3158 (51.7) | 2279 (50.4) | 879 (55.3) |
| Valvular Heart Disease, % | 3074 (50.3) | 2265 (50.1) | 809 (50.9) |
| Atrial Fibrillation, % | 3738 (61.2) | 2710 (59.9) | 1028 (64.7) |
| Ischemic Heart Disease, % | 5108 (83.6) | 3810 (84.3) | 1298 (81.6) |
| Diabetes, % | 3306 (54.1) | 2397 (53.0) | 909 (57.2) |
| Renal Failure, % | 3503 (57.3) | 2557 (56.6) | 946 (59.5) |
| Dementia, % | 369 (6.0) | 263 (5.8) | 106 (6.7) |
| Delirium, % | 185 (3.0) | 128 (2.8) | 57 (3.6) |
| Baseline Medication Use | |||
| Number of Unique Medications on Admission, mean (SD) | 8.3 (5.5) | 8.2 (5.5) | 8.6 (5.5) |
| βBs, % | 3549 (58.1) | 2632 (58.2) | 917 (57.7) |
| RASi, % | 2530 (41.4) | 1905 (42.1) | 625 (39.3) |
| ARNI, % | 1527 (8.6) | 474 (10.5) | 53 (3.3) |
| MRA, % | 997 (16.3) | 874 (19.3) | 123 (7.7) |
| SGLT2i, % | 299 (4.9) | 267 (5.9) | 32 (2.0) |
| Loop diuretics, % | 3451 (56.5) | 2604 (57.6) | 847 (53.3) |
| Number of GDMT classes, % | |||
| Untreated | 1707 (27.9%) | 1235 (27.3%) | 472 (29.7%) |
| Monotherapy | 2075 (34.0%) | 1468 (32.5%) | 607 (38.2%) |
| Dual Therapy | 1750 (28.6%) | 1301 (28.8%) | 449 (28.2%) |
| Triple Therapy | 516 (8.4%) | 460 (10.2%) | 56 (3.5%) |
| Quadruple Therapy | 63 (1.0%) | 57 (1.3%) | 6 (0.4%) |
| Clinical Measures at Discharge | |||
| Heart Rate (beats per minute), % | |||
| Normal (60–99) | 5145 (84.2) | 3809 (84.3) | 1336 (84.0) |
| Bradycardic (<60) | 520 (8.5) | 356 (7.9) | 164 (10.3) |
| Tachycardic (≥100) | 442 (7.2) | 353 (7.8) | 89 (5.6) |
| Blood Pressure, % | |||
| Normal/Elevated | 3054 (50.0) | 2301 (50.9) | 753 (47.4) |
| Hypotensive | 893 (14.6) | 780 (17.3) | 113 (7.1) |
| Stage 1 Hypertension | 1367 (22.4) | 949 (21.0) | 418 (26.3) |
| Stage 2 Hypertension | 793 (13.0) | 488 (10.8) | 305 (19.2) |
| eGFR (mL/min/1.73 m2), % | |||
| <30 | 1434 (23.5) | 1018 (22.5) | 416 (26.2) |
| 30–59 | 2681 (43.9) | 1985 (43.9) | 696 (43.8) |
| 60–89 | 1327 (21.7) | 1010 (22.3) | 317 (19.9) |
| ≥90 | 463 (7.6) | 372 (8.2) | 91 (5.7) |
| Hypokalemia (Potassium <3.5 mEq/L), % | 490 (8.0) | 361 (8.0) | 129 (8.1) |
| Hyperkalemia (Potassium >5.0 mEq/L), % | 162 (2.7) | 133 (2.9) | 29 (1.8) |
| Hyponatremia (Sodium <135 mEq/L), % | 1791 (29.3) | 1396 (30.9) | 395 (24.8) |
DBP: Diastolic blood pressure
eGFR: estimated glomerular filtration rate
SBP: Systolic blood pressure
Other Race include those reported as American Indian/Alaska native, Asian Indian, Filipino, Japanese, Other, Other Asian, Other Pacific Islander, or Vietnamese.
Discharge date was categorized into three time periods: after 2017 Guideline Update (7/2017–4/2020); after SGLT2i FDA approval for HFrEF (5/2020–3/2022), and after 2022 Guideline Update (4/2022–4/2023).
Blood pressure categories were defined as: Normal/Elevated (SBP 100–129 & DBP <80 mmHg); Hypotensive (SBP<100); Stage 1 Hypertension (SBP 130–139 or DBP 80–89 mmHg); and Stage 2 Hypertension (SBP ≥140 or DBP 90mmHg).
Rates of missingness (N, %) were as follows: race for 86 (1.4%), ethnicity 285 (4.7%), insurance 85 (1.4%), RUCA category 141 (2.3%), Area Deprivation Index 281 (4.6%), heart rate 4 (0.1%), blood pressure 4 (0.1%), eGFR 206 (3.4%), potassium 29 (0.5%), sodium 31 (0.5%).
Use of GDMT at Hospital Admission and Discharge
On admission, 58% of the cohort were using βBs, 41% RASis, 16% MRAs, and 5% SGLT2is (Figure 1). At discharge, 51% of patients were prescribed one or more new class and absolute estimated use increased from admission by 31 percentage points (pp) (58 pp vs 89 pp) for βBs, 22 pp (41 pp vs 64 pp) for RASis, 20 pp (16 pp vs 36 pp) for MRAs, and 6 pp (5 pp vs 11 pp) for SGLT2is. However, initiation of prescribed medications resulted in an absolute increase of only: 15 pp (58 pp vs 73 pp) for βBs, 11 pp (41 pp vs 53 pp) for RASis, 12 pp (16 pp vs 28 pp) for MRAs, and 4 pp (5 pp vs 9 pp) for SGLT2is.
Figure 1. Guideline-directed medical therapy use before and after heart failure hospitalization.
Stacked bar graph demonstrates use of guideline-directed medical therapy (GDMT) medication classes before and after HF hospitalization. Use of beta-blockers (βBs), renin angiotensin system inhibitors (RASis), mineralocorticoid receptor antagonists (MRAs), and sodium glucose cotransporter 2 inhibitors (SGLT2is) was determined by pharmacy fills. Prescriptions unfilled within 7 days of discharge represent primary non-adherence. Estimated use at discharge from fills prior to admission and new prescriptions is 89%, 64%, 36%, and 11% for βBs, RASis, MRAs, and SGLT2is, respectively. The estimated use of GDMT classes declined to 73%, 53%, 28%, and 9%, respectively, after accounting for primary non-adherence.
Among patients not treated prior to admission, βBs were newly prescribed in 74% of hospitalizations, RASis in 38%, MRAs in 24%, and SGLT2is in 7% (Figure 2). Among those with new prescriptions at discharge, primary non-adherence occurred in 64% of all patients and 46% of all new prescriptions: 51% for βBs, 48% for RASis, 39% of MRAs, and 35% of SGLT2is. Late initiation, between 7 and 90 days post-discharge, resulted in the initiation of an additional 20% of new discharge prescriptions, including 22% of βB, 19% of RASi, 13% of MRA, and 14% of SGLT2i discharge prescriptions filled.
Figure 2.
Initiation of guideline-directed medical therapy classes by varying post-discharge assessment windows
Panel A. βB treatment after heart failure hospitalization
Panel B. RASi treatment after heart failure hospitalization
Panel C. MRA treatment after heart failure hospitalization
Panel D. SGLT2i treatment after heart failure hospitalization
Sankey plots demonstrate the rate of new prescriptions at discharge and the resulting primary non-adherence and initiation rates post-discharge for each GDMT class. Around half of prescriptions are not filled within 7 days of discharge. Initiation rates from those without a prescription at discharge remain low throughout follow-up.
Patients with HFrEF had substantially higher proportions of MRA use (e.g. 12–21%) and modestly higher proportions of βB, RAAS, and SGLT2i use on admission, prescription at discharge, and initiation within 7 days post-discharge compared to those with HFmrEF (eFigures 3–5).
The proportion of patients continuing or initiating SGLT2is increased from 0% in 2017 to 37% in 2023, whereas other classes had more modest increases: βBs 72% to 75%, RASis 55% to 56%, and MRAs 29% to 32% (eFigure 6). Among those treated with RASis, the proportion using ARNIs increased from 5% to 48%.
The use of all recommended GDMT medication classes increased from 7% on admission to 15% within 7 days of discharge (p<0.01). Of the 5702 not on triple/quadruple GDMT regimens at admission, 3111 (55%) were prescribed at least one new GDMT class. This varied by site, ranging from 38% to 68% (eFigure 7).
Adherence and Persistence to GDMT Following Discharge
Among those filling GDMT medications, adherence at 6 months was 63% for βBs, 54% for RASis, 49% for MRAs and 49% for SGLT2is (Table 2). Six-month persistence was slightly higher: 70% for βBs, 60% for RASis, 55% for MRAs and 56% for SGLT2is. Among initiators, a steep drop in persistence was observed within 30-days in all classes: 17% of βB initiators, 23% of RASi initiators, 26% of MRA initiators, and 25% of SGL2i initiators. Thus, non-persistence as early as the first 30 days or first fill was observed in 22% of newly initiated GDMT overall. Compared to estimated use according to prescriptions at discharge, estimated use of each GDMT medication class decreased at 30 days: βB: 89% vs 71%, RASi: 64% vs. 51%, MRA: 36% vs 26%, SGLT2i: 11% vs 8%) (eTable 1).
Table 2.
Adherence and persistence to GDMT medication classes over 6 months post-hospitalization among patients continuing or initiating each GDMT class at discharge
| βBs | RASi | MRA | SGLT2i | Any GDMT use* | |
|---|---|---|---|---|---|
| N = 4481 | N = 3232 | N = 1740 | N = 548 | N=5183 | |
| Continuers†, n (%) | 3549 (79) | 2530 (78) | 997 (57) | 299 (55) | 4404 (85) |
| Initiators‡, n (%) | 932 (21) | 702 (22) | 743 (43) | 249 (45) | 1855 (36) |
| Persistence, n (%) | 3127 (70) | 1952 (60) | 961 (55) | 306 (56) | 2620 (51) |
| Adherence (PDC ≥ 80%), n (%) | 2830 (63) | 1749 (54) | 852 (49) | 268 (49) | 2188 (42) |
| PDC, median [IQR] | 93 [62, 100] | 86 [44, 100] | 79 [35, 100] | 78 [37, 99] | 80 [54, 98] |
Any GDMT use corresponds to continuers or initiators of any of the four GDMT classes. In these overall measurements across GDMT classes, overall adherence was measured as adherence to all medications present on discharge according to PDC ≥80% and similarly overall persistence was measured as persistence at 6-months to all medications present on discharge. For those with any GDMT use, initiators are labeled as those with at least one class initiated at discharge. Thus, 1076 (20.8%) patients are continuers of one or more GDMT classes and initiators of at least one other class.
Continuers are those with use of the respective GDMT class prior to admission.
Initiators are those without use of the respective GDMT class prior to admission that had a new pharmacy fill within 7 days following discharge.
At 6 months post-discharge, 16% of patients had died and mortality was consistently higher among continuers compared to initiators (eTable 2), however time to non-persistence accounting for competing risk of death was not meaningfully different between these groups (Figure 3). Overall 6-month adherence and persistence across GDMT classes was 42% and 51% respectively. Sensitivity analyses varying definitions of baseline medication use, allowable gaps in persistence, and days’ supply for adherence showed similar findings (eTables 3–5).
Figure 3. Time to GDMT class non-persistence by GDMT class continuer vs. initiator status.
Black and red shaded regions denote 95% confidence intervals for Fine and Gray estimates of time-to-nonpersistent for continuers and initiators, respectively. Numbers at risk presented are not estimated from Fine and Gray models and represent the observed data. Risk of death was higher among continuers compared to initiators (βB: 16.4 vs 10.8; RASi: 13.7 vs. 10.8; MRA: 18.8 vs 10.9; SGLT2i: 20/7 vs 11.6). No meaningful differences in time-to-non-persistence were observed among initiators and continuers (Gray’s test p-values all > 0.05, except for beta-blockers). After 30 days, persistence drops to about 75% in initiators of each of the GDMT class and follows a similar decline in persistence as continuers thereafter.
Contraindications to GDMT
At discharge, severe kidney disease was present in 24% of patients, hypotension in 15%, bradycardia in 9%, and hyperkalemia in 3%. Thus, 22% of patients had at least one contraindication to βBs, 37% to RASis, 37% to MRAs, and 24% to SGLT2is. After accounting for clinical contraindicators the estimated proportions of eligible patients using GDMT classes at discharge increased moderately (i.e. βB: 89% vs 93%, RASi: 64% vs. 78%, MRA: 36% vs 49%, SGLT2i: 11% vs 14%) (eTable 1).
Late Initiation of GDMT
Late initiation between 7 days the 90 days following discharge occurred for βBs in 33% (532/1630) of patients, RASi in 20% (562/2879), MRAs in 13% (556/4371), and SGLT2is in 4% (234/5563). The majority of late initiators of βBs (81%) and RASis (51%) were individuals who had been prescribed these classes at discharge, whereas only 33% of MRA and 25% of SGLT2i late initiators had been prescribed these classes at discharge. Late initiation of GDMT was significantly more common among patients without contraindications at discharge than those with contraindications: 35% versus 26% initiated βBs, 24% versus 13% initiated RASi, 15% versus 9% initiated MRAs, and 5% versus 1% initiated SGLT2is (all p<0.001).
Characteristics Associated with Primary Non-Adherence and Non-Persistence
In adjusted analyses, there were no significant differences in primary non-adherence or non-persistence by age, sex, insurance, or area deprivation, Non-persistence was more common among Black compared to White patients (risk ratio (RR) 1.12; 95% CI, 1.02 to 1.23) (eTables 6–7).
Primary non-adherence was modestly more common among patients with ischemic heart disease and diabetes while non-persistence was more common only among those with ischemic heart disease. A greater number of GDMT medications on admission was associated with higher risk of primary non-adherence (e.g. dual/triple therapy vs no GDMT: RR 3.74; 95% CI, 3.18 to 4.39)) and non-persistence (e.g. triple/quadruple therapy vs no GDMT: RR 2.11; 95% CI, 1.80 to 2.48)). Similarly, patients prescribed or initiated on multiple new GDMT medications at discharge had a higher risk of primary non-adherence (RR 1.37; 95% CI 1.27 to 1.48) and non-persistence (RR 1.39; 95% CI, 1.29 to 1.51). Discharge vital signs and laboratories signaling possible safety concerns were also associated with a higher risk of primary non-adherence (e.g. hyperkalemia [RR: 1.23; 95% CI, 1.00 to 1.50]) and non-persistence (e.g. eGFR <30 [RR: 1.19; 95% CI, 1.01 to 1.39] and hypotension [RR: 1.12; 95% CI, 1.03 to 1.22]).
There were no significant differences in primary non-adherence by hospital-level factors, non-persistence was more common in smaller hospitals (RR: 1.63; 95% CI: 1.06 to 2.51) and those with a cardiology fellowship (RR: 1.26; 95% CI: 1.09 to 1.47).
Discussion
In this cohort study of patients hospitalized for HF in a large regional US healthcare system, half of patients were prescribed at least one new GDMT medication at discharge, yet only 15% of patients received triple/quadruple GDMT therapy. Nearly half of GDMT prescriptions went unfilled within 7 days of discharge, but within 90 days of discharge this improved to one-quarter. However, nearly a quarter of patients who filled a new discharge GDMT prescription did not fill a subsequent prescription for the same class. In the six months following hospitalization, over one-third of patients were non-adherent and half were non-persistent to GDMT medications. These findings identify critical windows for targeted interventions to close the gaps in GDMT medication use not only at discharge, but also in the weeks to months that follow HF hospitalizations.
Previous studies of post-hospitalization GDMT utilization have used registry and administrative claims data which, when unlinked, lack a bridge between prescription orders and patient pharmacy fills (13, 14). A recent registry study of patients hospitalized with incident HFrEF between 2021 and 2023 estimated prescription rates at discharge of: 92% for βBs, 76% for RASis, 41% for MRAs, and 24% for SGLT2is (15). We observed similar patterns of GDMT prescription orders, however nearly half of prescription orders were unfilled within 7 days of discharge. Including both prescriptions and subsequent fills provides insight into the early lapses in uptake of GDMT and highlights opportunities for interventions, such as discharge medication delivery programs, which may reduce primary non-adherence (16).
The differences between initiation and adherence to therapy are further underscored in prior studies that predate the current guidelines and have shown similar or lower adherence rates for βBs and RASis, particularly ARNIs (17–19). Our study suggests despite stronger GDMT guideline recommendations, gaps in GDMT medication use endure. Therefore, it is important to recognize that advancements in hospital initiation of GDMT are necessary, but not sufficient to overcome GDMT medication underutilization. During the three months following HF hospitalizations, we observed a decrease in primary non-adherence due to late initiation, which may reflect successful outpatient counseling or assistance overcoming access barriers to optimize GDMT post-discharge. In contrast, new prescriptions of GDMT were limited during this time, a finding consistent with prior literature (18, 20, 21). Our results suggest in-hospital prescription of GDMT promotes outpatient optimization and titration, yet GDMT initiation in outpatient follow-up care remains underutilized and warrants additional attention and resources.
Among patients with potentially transient medical contraindications to GDMT, initiation rates remained low at three months. Those with contraindications represent a medically complex population for which quadruple GDMT is not a one-size-fits all solution and necessitates more patient-centered approaches and benchmarks. Nonetheless, this finding highlights how clinical inertia may allow a transient concern to become a persistent barrier to therapy, signaling missed opportunities for re-evaluation and treatment initiation. These patterns may also be a reflection of unaddressed structural barriers to therapy and care (e.g. appointment availability, transportation) in the outpatient setting or, in some cases, inadequate transition to outpatient follow-up care altogether (4, 22, 23). As we observed few differences in adherence outcomes by sociodemographic factors and none by insurance, the wide variation in GDMT use by hospital site (i.e. 38% to 68%) may signal the need to prioritize clinician and system-level interventions.
The high burden of contraindications, occurring in one-fifth of discharged patients, also indicates the importance of systematic reassessment of medication tolerance and eligibility during post-discharge care (4, 5). Follow-up encounters offer an opportunity to evaluate tolerability and adherence to therapies initiated during hospitalization and to identify barriers to continued medication use (4, 5, 24). Current guidelines and quality improvement efforts in HF care emphasize timely post-discharge follow-up as a core metric to reduce the risk of readmissions (3, 13, 25). Yet less than half of Medicare beneficiaries receive an outpatient cardiology follow-up visit within 30-days post-discharge (26). To improve this, better coordination and integration between inpatient and outpatient care is needed, along with infrastructure to ensure timely follow-up. Effective interventions to increase GDMT medication utilization and improve clinical outcomes have included combinations of home visits, telephone follow-up calls, patient self-care education, interdisciplinary medication reconciliation, and handoff communication to outpatient providers (6, 27, 28). Future research should evaluate the frequency and effectiveness of treatment re-evaluations, identify optimal follow-up strategies, and develop algorithms to safely rechallenge patients with GDMT medications in outpatient care.
Over the study period, the use of newer medications, like SGLT2i and ARNI, increased substantially, while use of βB, RASi, and MRA classes was flat. This finding may be attributed to emphasis on these medications in updated guidelines, pharmaceutical marketing campaigns, differential side effect profiles, or secondary indications (3, 13, 29). Additional efforts are needed to increase the use of existing generic medications, especially MRAs, to reduce the broader gap in quadruple therapy. Our findings that greater baseline GDMT use and prescription of multiple GDMT at discharge were associated with higher risk of primary non-adherence and non-persistence are consistent with prospective cohort studies indicating cardiometabolic medication complexity is associated with medication errors (30). Taken together these data indicate a need to focus on improved communication around newly prescribed medications, particularly for those facing already complex regimens. Furthermore, as heart failure is largely a disease of aging, these results may also signal that some patients facing multimorbidity and frailty may prefer not to initiate additional medications or may not tolerate high medication burden. Thus, GDMT prescribing and HF quality metrics must prioritize patient-directed goals rather than population-level targets for medication use.
This study has limitations. First, pharmacy claims indicate receipt of medications but still may overestimate utilization as they do not measure consumption. Second, we are unable to identify whether medication discontinuations were clinician directed and may underestimate adherence and persistence. We conservatively assumed that medications present on admission were continued, which could misclassify admission medications that were discontinued during hospitalization due to side effects. Third, our study is limited to a single healthcare system and may not generalize to other systems. Nonetheless, our data include multiple payers and both academic and non-academic hospitals, supporting a broader relevance. Lastly, there is the potential for underestimation of GDMT medication use if patients fill medications at pharmacies outside the Surescripts network. However, Surescripts has been previously validated as including 91% of medication insurance claims and we excluded patients without any medication fills at baseline to reduce potential misclassification (31).
Following HF hospitalization, many new HF prescriptions went unfilled until late in the post-discharge period, and most newly initiated prescriptions did not persist at 6 months, indicating a need to develop interventions to support patients’ medication use beyond the initial prescription at discharge.
Supplementary Material
Key Points.
Questions:
How frequently is guideline-directed medical therapy prescribed by inpatient clinicians and continued through the 6 months following a heart failure (HF) hospitalization?
Findings:
In an observational cohort study of 6,111 patients hospitalized for HF, 54% of new prescriptions for guideline-directed medical therapy went unfilled within 7 days of discharge. At one-month post-discharge, 22% of the newly initiated prescriptions had been discontinued. At 6-months, adherence and persistence to all discharge medications were 42% and 51%, respectively.
Meaning:
HF quality improvement initiatives have focused on initiation of guideline-directed medical therapy at hospital discharge, these findings highlight a need to develop interventions to support patients’ medication use beyond the initial discharge prescription.
Funding:
Funded by the National Institute on Aging (PI Anderson, K76AG074878) and American Heart Association (PI Anderson, CDA940950). Ms. Bessette was additionally supported by the National Institute of General Medical Sciences (T32GM144300) awarded to the University of Pittsburgh-Carnegie Mellon University MD-PhD Program and training grants from the National Institute on Aging (T32AG021885) and National Heart, Lung, and Blood Institute (T32HL083825). Dr. Magnani was additionally supported by the National Heart, Lung, and Blood Institute (K24HL160527).
Footnotes
Disclosures: No conflicts of interest due to relationships with industry or other relevant entities to report.
References
- 1.Fonarow GC. HF STATS 2025: Heart Failure Epidemiology and Outcomes Statistics An Updated 2025 Report from the Heart Failure Society of America. J Card Fail. 2025. Epub 20250829. doi: 10.1016/j.cardfail.2025.07.007. [DOI] [PubMed] [Google Scholar]
- 2.Yancy CW, Jessup M, Bozkurt B, Butler J, Casey DE Jr, Colvin MM, et al. 2017 ACC/AHA/HFSA Focused Update of the 2013 ACCF/AHA Guideline for the Management of Heart Failure: A Report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines and the Heart Failure Society of America. Circulation. 2017;136(6):e137–e61. Epub 20170428. doi: 10.1161/CIR.0000000000000509. [DOI] [PubMed] [Google Scholar]
- 3.Heidenreich PA, Bozkurt B, Aguilar D, Allen LA, Byun JJ, Colvin MM, et al. 2022 AHA/ACC/HFSA Guideline for the Management of Heart Failure: A Report of the American College of Cardiology/American Heart Association Joint Committee on Clinical Practice Guidelines. J Am Coll Cardiol. 2022;79(17):e263–e421. Epub 20220401. doi: 10.1016/j.jacc.2021.12.012. [DOI] [PubMed] [Google Scholar]
- 4.Oskouie S, Pandey A, Sauer AJ, Greene SJ, Mullens W, Khan MS, et al. From Hospital to Home: Evidence-Based Care for Worsening Heart Failure. JACC Adv. 2024;3(9):101131. Epub 20240731. doi: 10.1016/j.jacadv.2024.101131. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Hollenberg SM, Stevenson LW, Ahmad T, Bozkurt B, Butler J, Davis LL, et al. 2024 ACC Expert Consensus Decision Pathway on Clinical Assessment, Management, and Trajectory of Patients Hospitalized With Heart Failure Focused Update: A Report of the American College of Cardiology Solution Set Oversight Committee. J Am Coll Cardiol. 2024;84(13):1241–67. Epub 20240808. doi: 10.1016/j.jacc.2024.06.002. [DOI] [PubMed] [Google Scholar]
- 6.Ruppar TM, Cooper PS, Mehr DR, Delgado JM, Dunbar-Jacob JM. Medication Adherence Interventions Improve Heart Failure Mortality and Readmission Rates: Systematic Review and Meta-Analysis of Controlled Trials. J Am Heart Assoc. 2016;5(6). Epub 20160617. doi: 10.1161/JAHA.115.002606. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Anderson TS, Jing B, Wray CM, Ngo S, Xu E, Fung K, et al. Comparison of Pharmacy Database Methods for Determining Prevalent Chronic Medication Use. Med Care. 2019;57(10):836–42. doi: 10.1097/MLR.0000000000001188. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Anderson TS, Xu E, Whitaker E, Steinman MA. A systematic review of methods for determining cross-sectional active medications using pharmacy databases. Pharmacoepidemiol Drug Saf. 2019;28(4):403–21. Epub 20190213. doi: 10.1002/pds.4706. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Hutchins DS, Zeber JE, Roberts CS, Williams AF, Manias E, Peterson AM, et al. Initial Medication Adherence-Review and Recommendations for Good Practices in Outcomes Research: An ISPOR Medication Adherence and Persistence Special Interest Group Report. Value Health. 2015;18(5):690–9. Epub 20150516. doi: 10.1016/j.jval.2015.02.015. [DOI] [PubMed] [Google Scholar]
- 10.Thai TN, Winterstein AG. Core concepts in pharmacoepidemiology: Measurement of medication exposure in routinely collected healthcare data for causal inference studies in pharmacoepidemiology. Pharmacoepidemiol Drug Saf. 2024;33(3):e5683. Epub 20230926. doi: 10.1002/pds.5683. [DOI] [PubMed] [Google Scholar]
- 11.Anderson TS, Jing B, Fung K, Steinman MA. Older Adults' Persistence to Antihypertensives Prescribed at Hospital Discharge: a Retrospective Cohort Study. J Gen Intern Med. 2021;36(12):3900–2. Epub 20210119. doi: 10.1007/s11606-020-06401-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Zheng J, Sandhu AT, Bhatt AS, Collins SP, Flint KM, Fonarow GC, et al. Inpatient Use of Guideline-Directed Medical Therapy During Heart Failure Hospitalizations Among Community-Based Health Systems. JACC Heart Fail. 2025;13(1):43–54. Epub 20240911. doi: 10.1016/j.jchf.2024.08.004. [DOI] [PubMed] [Google Scholar]
- 13.Tang AB, Lewsey SC, Yancy CW, Heidenreich PA, Greene SJ, Allen LA, et al. Get With the Guidelines-Heart Failure: Twenty Years in Review, Lessons Learned, and the Road Ahead. Circ Heart Fail. 2025:e012936. Epub 20250512. doi: 10.1161/CIRCHEARTFAILURE.125.012936. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Baez-Gutierrez N, Galindo-Garcia C, Rodriguez-Ramallo H, Sanchez-Fidalgo S. Adherence and persistence to heart failure guideline-directed medical therapy: A systematic review of studies based on electronic healthcare data. Res Social Adm Pharm. 2025;21(12):1013–23. Epub 20250717. doi: 10.1016/j.sapharm.2025.07.005. [DOI] [PubMed] [Google Scholar]
- 15.Greene SJ, Ayodele I, Pierce JB, Khan MS, Lewsey SC, Yancy CW, et al. Eligibility and Projected Benefits of Rapid Initiation of Quadruple Therapy for Newly Diagnosed Heart Failure. JACC Heart Fail. 2024;12(8):1365–77. Epub 20240325. doi: 10.1016/j.jchf.2024.03.001. [DOI] [PubMed] [Google Scholar]
- 16.Witcraft EJ, Norris AM, Fudzie SS, Vest MH, Johnson N, Rush J, et al. Impact of medication bedside delivery program on hospital readmission rates. J Am Pharm Assoc (2003). 2021;61(1):95–100 e1. Epub 20201113. doi: 10.1016/j.japh.2020.09.023. [DOI] [PubMed] [Google Scholar]
- 17.Carnicelli AP, Lippmann SJ, Greene SJ, Mentz RJ, Greiner MA, Hardy NC, et al. Sacubitril/Valsartan Initiation and Postdischarge Adherence Among Patients Hospitalized for Heart Failure. J Card Fail. 2021;27(8):826–36. doi: 10.1016/j.cardfail.2021.03.012. [DOI] [PubMed] [Google Scholar]
- 18.Carnicelli AP, Li Z, Greiner MA, Lippmann SJ, Greene SJ, Mentz RJ, et al. Sacubitril/Valsartan Adherence and Postdischarge Outcomes Among Patients Hospitalized for Heart Failure With Reduced Ejection Fraction. JACC Heart Fail. 2021;9(12):876–86. Epub 20210908. doi: 10.1016/j.jchf.2021.06.018. [DOI] [PubMed] [Google Scholar]
- 19.Chang LL, Xu H, DeVore AD, Matsouaka RA, Yancy CW, Fonarow GC, et al. Timing of Postdischarge Follow-Up and Medication Adherence Among Patients With Heart Failure. J Am Heart Assoc. 2018;7(7). Epub 20180401. doi: 10.1161/JAHA.117.007998. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Curtis LH, Mi X, Qualls LG, Check DK, Hammill BG, Hammill SC, et al. Transitional adherence and persistence in the use of aldosterone antagonist therapy in patients with heart failure. Am Heart J. 2013;165(6):979–86 e1. Epub 20130418. doi: 10.1016/j.ahj.2013.03.007. [DOI] [PubMed] [Google Scholar]
- 21.Bhatt AS, Fonarow GC, Greene SJ, Holmes DN, Alhanti B, Devore AD, et al. Medical Therapy Before, During and After Hospitalization in Medicare Beneficiaries With Heart Failure and Diabetes: Get With The Guidelines - Heart Failure Registry. J Card Fail. 2024;30(2):319–28. Epub 20230925. doi: 10.1016/j.cardfail.2023.09.005. [DOI] [PubMed] [Google Scholar]
- 22.Swat SA, Helmkamp LJ, Tietbohl C, Thompson JS, Fitzgerald M, McIlvennan CK, et al. Clinical Inertia Among Outpatients With Heart Failure: Application of Treatment Nonintensification Taxonomy to EPIC-HF Trial. JACC Heart Fail. 2023;11(11):1579–91. Epub 20230816. doi: 10.1016/j.jchf.2023.06.022. [DOI] [PubMed] [Google Scholar]
- 23.Bilicki DJ, Reeves MJ. Outpatient Follow-Up Visits to Reduce 30-Day All-Cause Readmissions for Heart Failure, COPD, Myocardial Infarction, and Stroke: A Systematic Review and Meta-Analysis. Prev Chronic Dis. 2024;21:E74. Epub 20240926. doi: 10.5888/pcd21.240138. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Maddox TM, Januzzi JL Jr., Allen LA, Breathett K, Brouse S, Butler J, et al. 2024 ACC Expert Consensus Decision Pathway for Treatment of Heart Failure With Reduced Ejection Fraction: A Report of the American College of Cardiology Solution Set Oversight Committee. J Am Coll Cardiol. 2024;83(15):1444–88. Epub 20240308. doi: 10.1016/j.jacc.2023.12.024. [DOI] [PubMed] [Google Scholar]
- 25.Balasubramanian I, Andres EB, Malhotra C. Outpatient Follow-Up and 30-Day Readmissions: A Systematic Review and Meta-Analysis. JAMA Netw Open. 2025;8(11):e2541272. Epub 20251103. doi: 10.1001/jamanetworkopen.2025.41272. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Anderson TS, Yeh RW, Herzig SJ, Marcantonio ER, Hatfield LA, Souza J, et al. Trends and Disparities in Ambulatory Follow-Up After Cardiovascular Hospitalizations : A Retrospective Cohort Study. Ann Intern Med. 2024;177(9):1190–8. Epub 20240806. doi: 10.7326/M23-3475. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Albert NM, Barnason S, Deswal A, Hernandez A, Kociol R, Lee E, et al. Transitions of care in heart failure: a scientific statement from the American Heart Association. Circ Heart Fail. 2015;8(2):384–409. Epub 20150120. doi: 10.1161/HHF.0000000000000006. [DOI] [PubMed] [Google Scholar]
- 28.Zheng J, Mednick T, Heidenreich PA, Sandhu AT. Pharmacist- and Nurse-Led Medical Optimization in Heart Failure: A Systematic Review and Meta-Analysis. J Card Fail. 2023;29(7):1000–13. Epub 20230331. doi: 10.1016/j.cardfail.2023.03.012. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.DeJong C, Inoue K, Durstenfeld MS, Agarwal A, Chen JC, Tseng CW, et al. Direct-to-Physician Marketing and Uptake of Optimal Medical Therapy for Heart Failure With Reduced Ejection Fraction. JACC Heart Fail. 2025;13(7):102380. Epub 20250319. doi: 10.1016/j.jchf.2024.11.020. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Anderson TS, Wilson LM, Wang BX, Steinman MA, Schonberg MA, Marcantonio ER, et al. Medication Errors and Gaps in Medication Discharge Planning for Hospitalized Older Adults: A Prospective Cohort Study. J Gen Intern Med. 2026;41(3):697–706. Epub 20251119. doi: 10.1007/s11606-025-09973-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Blecker S, Adhikari S, Zhang H, Dodson JA, Desai SM, Anzisi L, et al. Validation of EHR medication fill data obtained through electronic linkage with pharmacies. J Manag Care Spec Pharm. 2021;27(10):1482–7. doi: 10.18553/jmcp.2021.27.10.1482. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.