The Global Health and Economic Burden of Hospitalizations for Heart Failure

Lessons Learned From Hospitalized Heart Failure Registries

Patient Demographics and Clinical Characteristics

The mean age of patients admitted with a primary diagnosis of HF ranges from 70 to 75 years, with an SD of 15 years (Table 1). Regional variation in age is likely explained by differences in the prevalence of underlying risk factors as well as the standard of living. For example, patients participating in the major North American registries tended to be older than patients enrolled in countries with developing economies. In fact, the ADHERE-AP (Acute Decompensated Heart Failure National Registry International–Asia Pacific) registry found substantial variation in age of presentation among the 8 participating countries, which showed a strong inverse correlation with the human development index, a composite measure including life expectancy, adult literacy, educational level, and standard of living (25).

Approximately 40% to 50% of HHF patients are female, a group of patients that have traditionally been under-represented in clinical trials (29), with 1 notable exception (30). The available data suggest that a larger proportion of HHF patients in the United States are female compared with other regions of the world. This is a noteworthy observation, as female patients are unique in that they tend to be older at the time of initial diagnosis and are more likely to have heart failure with preserved ejection fraction (HFpEF) (31). However, after adjusting for differences in baseline characteristics, women have comparable outcomes to men (32). There are virtually no data on race and ethnicity outside of the United States, and when collected, these data have traditionally been limited by the accuracy and completeness of provider documentation. Despite these shortcomings, African Americans and Hispanics make up approximately 20% and 7% of HHF patients, respectively, and tend to present at a younger age and have a lower ejection fraction (EF) and higher prevalence of medical comorbidities. Although the relative burden and severity of HF is greater in these vulnerable groups, HF care and risk-adjusted outcomes appear to be equitable in the context of multiple national hospital-based registries (33–35).

An ischemic etiology is universally the most common cause of HF, whereas HF secondary to uncontrolled hypertension, valvular pathology, and congenital heart disease are likely to be more common in the developing world. Depending on how the EF is categorized (i.e., 40% to 45%) and the population under consideration, 50% to 60% of the HHF population is classified as heart failure with reduced ejection fraction (HFrEF) (36). However, hospital-based registries conducted to date have not routinely measured EF during index admission. Thus, the true epidemiologic breakdown of HF patients by EF is unknown. More importantly, it is unclear what proportion of HFpEF patients previously had a reduced EF that improved in response to evidence-based therapies. In contrast to HFrEF, HFpEF is poorly characterized as a clinical entity, and there are currently no available evidence-based therapies, although medical comorbidities should be treated accordingly (36–38). Interestingly, there are some data based on the American Heart Association’s GWTG-HF (Get With The Guidelines–Heart Failure) registry to suggest that the proportion of HHF patients classified as HFpEF is growing and may exceed HFrEF in the future (Fig. 1) (17). This may be due to increasing recognition of HFpEF by providers as well as demographic trends including aging of the population.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Breakdown of Patients by EF Over Time

The percentage of patients with preserved ejection fraction (EF) has gradually increased during the enrollment period of the GWTG-HF (Get With The Guidelines–Heart Failure) registry.

Reprinted with permission from Steinberg et al. (17).

Importantly, cardiac and noncardiac comorbidities are extremely prevalent among HHF patients globally. More than one-half of all HHF patients have known CAD, which is complicated by myocardial infarction in 20% to 30% of cases, frequently resulting in systolic dysfunction. In addition, approximately 70% and 40% of HHF patients, respectively, have a history of hypertension and atrial fibrillation, and the prevalence of these comorbidities is even higher in HFpEF where they directly contribute to diastolic dysfunction and impaired ventricular filling. Similarly, noncardiac comorbidities including diabetes mellitus, chronic kidney disease, and chronic obstructive pulmonary disease may be found in over one-third of HHF patients. These conditions not only impact the pathophysiologic progression of HF but also limit the initiation and titration of evidence-based drug and diuretic therapy.

Although regional differences in comorbid conditions likely exist, the variation is less pronounced than would be expected based on clinical trial experience (14). However, because many of these comorbidities may represent either the inclusion or the exclusion criteria for enrollment in clinical trials, it is possible that they are most accurately characterized in a trial setting. Furthermore, medical comorbidities in hospital-based studies are generally self-reported, extracted from chart review and/or billing codes, and lack formal diagnostic criteria and objective evidence of the severity. In addition, other important comorbidities among HHF patients may be systematically underappreciated due to the lack of comprehensive screening and documentation efforts in clinical practice (e.g., sleep disorders and depression). Finally, therapies that may concurrently impact the management of HF (i.e., bronchodilators in chronic obstructive pulmonary disease and thiazolidinediones in diabetes mellitus) have not been uniformly reported.

Initial Clinical Presentation and Classification

Most HHF patients are admitted for worsening chronic HF and have 1 or more identifiable acute precipitants leading to admission (Table 2). It is important to note that about one-half of patients are hypertensive at presentation. In contrast, only ∼2% of patients present with an initial systolic blood pressure <90 mm Hg, suggestive of cardiogenic shock and systemic hypoperfusion. As a result, signs and symptoms of low cardiac output and inadequate end-organ perfusion (i.e., altered mentation, weak peripheral pulses, and cold clammy extremities) are relatively uncommon among HHF patients, and the vast majority present with signs and symptoms of pulmonic and systemic congestion.

Dyspnea at rest is reported in only about 34% of patients at admission; however, with provocation (i.e., exertion and orthopnea) closer to 90% of patients report some level of dyspnea (39). Rales may be appreciated on physical examination in close to 70% of patients, and two-thirds of patients have signs of systemic congestion including elevated jugular venous pressure and/or peripheral edema. Chest x-ray is obtained in more than 90% of HHF patients at initial presentation and three-quarters show radiographic evidence of pulmonary congestion.

Laboratory, Electrocardiography, and Echocardiography Findings

Approximately one-half of HHF patients will exhibit some degree of anemia, with over one-fourth presenting with at least moderate anemia (40). Approximately 20% of patients are hyponatremic (i.e., serum sodium <135 mEq/l) at admission, but the prevalence and degree of hyponatremia is higher and more severe in patients admitted to the intensive care unit (ICU) or coronary care unit (CCU) (41). Although about one-third of HHF patients carry a formal diagnosis of chronic kidney disease, at admission only 10% of patients will have an estimated glomerular filtration rate ≥90 ml/min/1.73 m², and about 20% will have severe renal impairment with an estimated glomerular filtration rate <30 ml/min/1.73 m² (42). Despite the fact that natriuretic peptides (NPs) have been investigated as a diagnostic marker, prognostic indicator, and potential guide to therapy, NP data in many hospital-based registries is either not included or, when it is reported, it is unavailable for a large proportion of patients.

Similarly, global HHF registry data on electrocardiographic and echocardiographic findings at admission are limited. Notably, the prevalence of atrial fibrillation on baseline electrocardiogram (ECG) may be as high as 50%, and close to 20% of HHF patients may experience new-onset atrial fibrillation during hospitalization (Table 3). In addition, after excluding ECGs with paced rhythms, more than one-third of patients have a prolonged QRS duration, a finding consistent with recent clinical trial data (43). Finally, a substantial proportion of HHF patients have normal or near-normal chamber size and systolic function and no underlying valvular heart disease, a testament to the increasing prevalence of HFpEF as well as the complex relationship among cardiac structure, function, and the clinical manifestations of HF (44) (Table 4).

Inpatient Management

Despite recent developments in evidence-based drug and device therapy for ambulatory HF patients with reduced EF, there have been few advances in the management of HHF patients, of treatment still being intravenous diuretics and/or vasodilators (45,46). Although the use of diuretics is ubiquitous, the route, dose, and duration of therapy have not been routinely recorded in HHF registries (Table 5). In contrast, the utilization of vasoactive drugs (i.e., vasodilators and inotropes) exhibits substantial geographic variation and appears to play a less prominent role in HHF management in North America compared with other regions. This finding has important prognostic implications, as even short-term inotropic support has been associated with increased mortality (47).

It is also noteworthy that the vast majority of HHF patients do not receive any procedural interventions during their hospital stay. Although ischemia is by far the most common etiology of HF and many HHF patients may present with concomitant acute coronary syndrome or subclinical ischemia (48–50), <10% of patients undergo coronary angiography during index hospitalization. In addition, pulmonary artery catheters are not commonly placed during hospitalization (51). The utilization of intravenous therapies and procedural interventions in HF patients receiving ICU/CCU-level care is likely higher, but few large observational experiences in this high-acuity population have been conducted to date.

Utilization of Evidence-Based Therapies

HHF is an opportune time to review current management and implement evidence-based therapies for chronic HF in a controlled and monitored setting; indeed, in-hospital initiation of therapy is 1 of the best predictors of long-term use (52–54). A number of important insights can be discerned by examining the geographic variation and temporal trends in evidence-based medication utilization patterns (Fig. 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Evidence-Based Medication Utilization at Admission and Discharge

The rate of prescription of guideline-recommended therapies has increased over time. In addition, these life-prolonging medications are commonly initiated during hospitalization (A: admission, B: discharge). ACEI/ARB = angiotensin-converting enzyme inhibitor/angiotensin receptor blocker; ADHERE = Acute Decompensated Heart Failure National Registry; AHEAD = Acute Heart Failure Database; ATTEND = Acute Decompensated Heart Failure Syndromes; EFICA = Epidémiologie Francaise de l'Insuffisance Cardiaque Aiguë; EHFS II = European Heart Failure Survey II; ESC-HF = European Society of Cardiology–Heart Failure; IN-HF = Italian Registry on Heart Failure; MRA = mineralocorticoid receptor antagonist; OPTIMIZE-HF = Organized Program to Initiate Lifesaving Treatment in Hospitalized Patients With Heart Failure; RO-AHFS = Romanian Acute Heart Failure Syndromes.

First, adherence to evidence-based chronic HF therapies is highest in North America, Western Europe, and Japan and lowest in the developing economies of Eastern Europe and Asia. A noteworthy exception to this generalization is that the rate of mineralocorticoid receptor antagonist (MRA) prescription at discharge is markedly lower in North America. Although the indication for MRAs has recently been expanded to include HFrEF patients spanning from mild to severe symptoms (55), providers may still be reluctant to prescribe these medications due to the risk of hyperkalemia and arrhythmic events (56). In the European Society of Cardiology Heart Failure Long-Term registry, after taking into consideration reasons for nonadherence, the real rate of undertreatment at discharge for angiotensin-converting enzyme inhibitors/angiotensin receptor blockers, beta-blockers, and MRA was estimated to be <5% (57). In addition, although only 30% of patients achieved target dosing, a specific reason was documented in close to two-thirds of patients. However, few hospital-based registries have uniformly documented medication dosing or specific intolerance(s) or relative/absolute contraindication(s) to initiating and/or up-titrating pharmacotherapy. In addition, the utilization of evidence-based medications at target dosing is likely overestimated and substantially lower at community hospitals compared with academic centers.

Many HHF patients may also be eligible for implantable cardioverter-defibrillators and/or cardiac resynchronization therapy, yet the vast majority of patients without previous implantation are discharged without being considered for device therapy. Further research is necessary to clarify the optimal timing of device implantation during hospitalization or soon after discharge among HHF patients (58–60).

In-Hospital and Post-Discharge Outcomes

Global HHF registries show that the median length of stay (LOS) ranges from 4 to 20 days and in-hospital mortality from 4% to 30% (Fig. 3). The most recent data based on the American Heart Association’s ongoing GWTG-HF registry revealed a median LOS of approximately 4 days and in-hospital mortality of <3%. A notable outlier is the median LOS of 21 days reported in the Japanese ATTEND (Acute Decompensated Heart Failure Syndromes) registry, which is driven by differences in reimbursement and participation in inpatient disease management programs. Similarly, a 4-week mortality of 28% was reported in the EFICA (Epidémiologie Francaise de l'Insuffisance Cardiaque Aiguë) registry; however, this registry was restricted to patients requiring an ICU/CCU level of care. In general, registries with a shorter LOS tend to have lower in-hospital mortality. Thus, differences in LOS due to clinical practice and reimbursement patterns may largely explain the observed geographic variation in mortality during hospitalization.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Length of Stay and In-Hospital Mortality

There is a direct relationship between (A) length of stay and (B) in-hospital mortality across the globe. *Mean. **4-week mortality. ADHERE-AP = Acute Decompensated Heart Failure National Registry International–Asia Pacific; ALARM-HF = Acute Heart Failure Global Registry of Standard Treatment; GWTG-HF = Get With The Guidelines–Heart Failure; other abbreviations as in Figure 2.

Very few HHF registries have collected data on post-discharge mortality and readmissions. In the OPTIMIZE-HF (Organized Program to Initiate Lifesaving Treatment in Hospitalized Patients With Heart Failure) study follow-up cohort representing about ∼10% of the overall registry, the post-discharge readmission rate was approximately 30% within 60 to 90 days post-discharge, and mortality ranged from 5.4% to 14.0% based on admission systolic blood pressure. Similarly, the ESC-HF (European Society of Cardiology–Heart Failure) Pilot survey reported 1-year mortality and readmission rates of 17.4% and 31.9%, respectively, at representative centers from 12 European countries (61). Although the short-term rehospitalization rate is high, the cause of nearly one-half of all readmissions may be noncardiovascular in etiology (62). Interestingly, outcomes appear to be comparable or slightly better for HFpEF (Fig. 4).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

In-Hospital and Post-Discharge Outcomes by EF

Although there is a trend toward lower (A) in-hospital mortality among hospitalized heart failure patients with preserved ejection fraction (EF) participating in the GWTG-HF (Get With The Guidelines–Heart Failure) program, follow-up data from the OPTIMIZE-HF (Organized Program to Initiate Lifesaving Treatment in Hospitalized Patients With Heart Failure) registry suggest that (B) post-discharge survival is poor irrespective of EF. LVSD = left ventricular systolic dysfunction; PSF = preserved systolic function.

Reprinted with permission from Steinberg et al. (17) and Fonarow et al. (38).

Predictors of Morbidity and Mortality

The most widely accepted independent predictors of morbidity and mortality in HHF patients include age (63), cardiac (64,65) and noncardiac comorbidities (66–68), systolic blood pressure (69), renal function (i.e., blood urea nitrogen and serum creatinine) (42,69), serum sodium (41), hemoglobin (40), NP concentration (70,71), troponin (71,72), QRS duration (43,71), and evidence-based medication utilization (73,74). Data from the ADHERE (Acute Decompensated Heart Failure National Registry) (75) and OPTIMIZE-HF (69) registries both found renal function and systolic blood pressure at admission to be among the best discriminators between hospital survivors and non-survivors (Fig. 5). The OPTIMIZE-HF registry found that 8 factors, including age, weight, systolic blood pressure, sodium, serum creatinine, and comorbid disease states, could predict the combined endpoint of death or readmission with a c-index of 0.72 (76). The combination of an elevated B-type NP and positive troponin (Tn) levels (i.e., defined as TnI ≥1.0 ng/ml or TnT ≥0.1 ng/ml) in the ADHERE registry designated a small subset of patients (i.e., ∼5%) who were at a particularly high risk for short-term mortality (71).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

Admission Renal Function, SBP, and In-Hospital Mortality

Baseline renal insufficiency and low systolic blood pressure (SBP) denoted a subgroup of hospitalized heart failure patients at particularly high risk for in-hospital mortality in the OPTIMIZE-HF (Organized Program to Initiate Lifesaving Treatment in Hospitalized Patients With Heart Failure) registry. SCr = serum creatinine.

Reprinted with permission from Abraham et al. (69).

Although it may be argued that the aforementioned independent predictors of morbidity and mortality and risk models may have limited external validity, much of these data have been corroborated by post-hoc analysis of global clinical trial databases. Thus, a number of important conclusions can be drawn regarding outcomes and risk prediction among HHF patients. First, HHF patients experience relatively low in-hospital mortality but are at much higher risk for early post-discharge readmission and mortality. The risk for in-hospital and post-discharge adverse events can be predicted with a high degree of fidelity by a small number of widely-available clinical variables, but a multitude of novel candidate variables may have added prognostic value (71). Risk prediction for mortality appears to be more accurate than for readmission (77). Finally, absence of high-risk features does not define an absolute low-risk population.

Impact of Quality Improvement Initiatives

Although most global HHF registries have been purely observational and have not included a prospective quality improvement component, there are some data to suggest that simply participating in a registry with intermittent benchmarked data reports can lead to improvements in quality measures and outcomes over time (78). For instance, in the ADHERE registry, over a 3-year period, beta-blocker use during hospitalization and at discharge increased by 30% and 29%, LOS decreased from 6.3 to 5.5 days, and in-hospital mortality decreased from 4.5% to 3.2%.

Following the success of the ADHERE registry, the next major North American registry, OPTIMIZE-HF, was designed to be a national quality improvement registry (79). The process-of-care improvement component included evidence-based practice algorithms, standardized order sets, discharge checklists, and a variety of other features. The OPTIMIZE-HF registry found that beta-blocker use increased from 76% to 86%, and there was a significant reduction in LOS, but there was only a trend toward lower in-hospital or post-discharge mortality, or the composite of post-discharge mortality and readmission rate (80). However, use of the process-of-care components was associated with lower risk-adjusted in-hospital mortality and post-discharge deaths and readmissions.

The OPTIMIZE-HF registry experience is further corroborated by the American Heart Association’s GWTG-HF registry quality improvement initiative, which has shown that participating U.S. hospitals have better outcomes compared with nonparticipant hospitals (81). Furthermore, among participating hospitals, those receiving awards related to achieving a certain degree of benchmark compliance had lower risk-adjusted HF mortality. These improved outcomes are thought to be due to better adherence to the process-of-care intervention, because award-winning and non–award-winning hospitals performed comparably on other core hospital performance measures (i.e., pneumonia and surgical infection prevention) not specifically addressed by the GWTG-HF program.

Collectively, the data suggest that it is feasible to design and conduct a quality improvement initiative in HHF on a large scale and that adopting standardized clinical decision-making tools may increase adherence to quality measures and may improve in-hospital and post-discharge outcomes. Given the ongoing high post-discharge event rate, future global HHF registries may focus on implementing interventions aimed at the transition of care from the inpatient to the outpatient setting (82–85).

Limitations of Existing Data and Future Directions

Despite the immense benefit of past HHF registries, there remain important knowledge gaps. It may be prudent to systemically discuss these shortcomings, barriers to implementation, and strategies for success. Areas to focus on in the design and conduct of future global HHF registries include, but are not limited to, geographic representation, patient enrollment, data capture, and quality improvement interventions.

The single greatest limitation of the HHF registries conducted to date is the relative paucity of data collected outside of North America and Western Europe. Thus, the current knowledge base on HHF is largely derived from a nonrepresentative sample including only slightly more than 15% of the world’s population. Expanding recruitment to other regions may require an investment in infrastructure and personnel where it is more rudimentary or nonexistent. This will require the international medical community to partner with local governments and existing organizations to achieve shared goals. A recent post-hoc analysis of the EVEREST (Efficacy of Vasopressin Antagonism in Heart Failure: Outcome Study with Tolvaptan) database found significant regional differences in clinical characteristics, management, and unadjusted outcomes, suggesting that despite restrictive inclusion and exclusion criteria, differential enrollment across geographic regions has the potential to dramatically alter the study population and potentially impact the overall response to investigational therapies (14). Thus, these efforts will prove pivotal as the current trend is toward conducting global clinical trials and in some cases, moving clinical trials entirely outside of the United States and Western Europe, where the costs and regulatory burden may be perceived to be prohibitive (86,87).

A related issue is that most HHF registry data have been based on the nonconsecutive enrollment of patients admitted with a definitive primary diagnosis of HF. Nonconsecutive enrollment may result in the differential recruitment of lower-acuity patients and may not be truly representative of “all-comers.” Future HHF registries should include on-site chart review specialists in order to prospectively enroll patients and enter data into a centrally-managed, web-based reporting system. However, even with dedicated staffing and institutional support, it may only be possible to enroll consecutive patients a few days per week, a compromise that may still be more reflective of the true epidemiology of HHF yet logistically feasible.

There is also a need to standardize the data collected, which has been subject to resource availability and physician preference in previous hospital-based registries. The study protocol of future HHF registries should include basic laboratory values, biomarkers (i.e., NPs and troponin), an ECG, and a transthoracic echocardiogram during index admission. Similarly, although the prevalence of inpatient therapies has been previously characterized, less is known about the route, dose, and duration of diuretics and other vasoactive medications in more contemporary and international HHF registries. In addition, a more detailed description including procedural interventions and the clinical indication is needed for higher-acuity patients requiring ICU or CCU care.

Similarly, global HHF registry data clearly show that the utilization of evidence-based therapies with proven mortality benefit has increased over time. However, patients may not achieve target dosing. The reason for not initiating or up-titrating these medications, including intolerance(s) and/or relative/absolute contraindication(s), should be queried in future hospital-based registries. Finally, in contrast to in-hospital outcomes, data on short- and long-term post-discharge mortality and readmissions have rarely been collected in HHF registries, although the majority of adverse events are known to occur post-hospitalization. Strategies should be developed to assess post-discharge outcomes while simultaneously allowing for consecutive enrollment. The status of readmissions and deaths may be supplemented by existing national reporting databases.

Quality improvement initiatives have rarely been included in large-scale HHF registries and universally have been limited to the timeframe between initial presentation and discharge. Process-of-care measures in future global HHF registries should focus on transitioning the success achieved during hospitalization to the ambulatory setting in order to reduce the unacceptably high post-discharge event rate. This may include early follow-up, surveillance laboratory testing (i.e., electrolytes and renal function), medication reconciliation and titration, and appropriate referrals for device (i.e., ICD and/or CRT) and other procedural interventions (i.e., coronary revascularization, ventricular assist device and/or transplant evaluation). In addition, permanent hospital-based registries are preferable to studies of a defined duration and may facilitate inclusion of an adaptive intervention in order to monitor the impact on outcomes over time.

Conclusions

There is currently an unmet critical need in HHF to design and conduct rational, global, hospital-based registries to better understand this heterogeneous patient population, inform public policy decisions, and guide basic, translational, and clinical research. It is highly desirable to develop a hospital-based registry that is global and geographically representative, employs consecutive or intermittently consecutive enrollment, and captures comprehensive and longitudinal data including hospital course and post-discharge outcomes. Future studies should also incorporate quality improvement initiatives focusing on continuity of care from initial presentation to the early post-discharge vulnerable period. In addition to traditional endpoints (i.e., hospitalization and mortality), patient-centered outcomes should be designed that comprehensively and longitudinally capture the burden of worsening HF (i.e., quality of life impairments and functional limitations). Thus, additional hospital-based registries are desperately needed to further our understanding of HHF and guide research endeavors (27,28,86,87).