Long-term Survival After Heart Failure: A Contemporary Population-Based Perspective

Background  Heart failure (HF) is a major public health problem that is associated with substantial morbidity, impaired quality of life, and diminished survival. Despite the considerable prevalence of HF in the United States, there are limited published data describing the contemporary long-term prognosis of patients hospitalized with decompensated HF.

Methods  A total of 2445 residents in the Worcester metropolitan area discharged from 11 greater Worcester hospitals after confirmed acute HF during 2000 comprised the study sample. Follow-up of discharged hospital survivors was carried out through 2005.

Results  The mean age of the study population was 76 years, 43.4% were men, and approximately three quarters had been previously diagnosed as having HF. Among discharged hospital patients, 37.3% died during the first year after hospital discharge, while 78.5% died during the 5-year follow-up period. Several subgroups of patients were at significantly increased risk for dying during the first year after hospital discharge. This included older persons (≥85 years) (adjusted odds ratio [OR], 2.11; 95% confidence interval [CI], 1.35-3.29), patients with a history of chronic obstructive pulmonary disease (OR, 1.39; 95% CI, 1.15-1.69) or HF (OR, 1.26; 95% CI, 1.00-1.59), and patients with elevated serum urea nitrogen levels during hospitalization (OR, 1.02; 95% CI, 1.01-1.03).

Conclusions  The results of our communitywide study demonstrate the poor long-term prognosis of patients surviving hospitalization for decompensated HF. Despite advances in the therapeutic management of these patients, their long-term survival remains guarded. Efforts are needed to improve the long-term survival of patients with this clinical syndrome.

Heart failure (HF) is a growing national health concern that is widely considered to be the next cardiovascular “epidemic.”^1,2 Despite advances in the therapeutic and nonpharmacologic management of patients with HF, some data suggest that the long-term prognosis of patients with HF remains guarded.^3-10 Contemporary data, particularly from the more generalizable perspective of a population-based investigation, that describe the long-term survival of patients with decompensated HF are limited, however.

The objectives of our communitywide study are to describe the long-term survival outlook for patients discharged from all 11 greater Worcester medical centers after being hospitalized for acute HF in 2000 and to identify demographic and clinical factors associated with a poor long-term prognosis. Data from the Worcester Heart Failure Study are used for purposes of this investigation.^11,12

Male and female adult residents of all ages from the Worcester metropolitan area (478 000 individuals per the 2000 census estimate) hospitalized for possible HF at all 11 greater Worcester medical centers during 2000 comprised the study population. The medical records of patients with primary or secondary International Classification of Diseases, Ninth Revision codes consistent with the possible presence of HF were reviewed in a standardized manner.^11,12 Patients with a discharge diagnosis of HF (code 428) comprised the primary diagnostic rubric reviewed for the identification of cases of possible HF. In addition, the medical records of patients with discharge diagnoses of rheumatic HF (code 398.9), hypertensive heart and renal disease (codes 402 and 404, respectively), acute cor pulmonale (code 415), other diseases of the endocardium (code 424), cardiomyopathy (code 425.4), pulmonary heart disease and congestion (codes 416.9 and 514, respectively), acute lung edema (code 518.4), edema (code 782.3), and dyspnea and respiratory abnormalities (code 786) were reviewed by trained study physicians and nurses to identify patients who also may have had new-onset HF. Confirmation of the diagnosis of HF, based on use of the Framingham Study criteria,^13,14 included the presence of 2 major criteria or the presence of 1 major and 2 minor criteria. An incident (first) event of HF was defined as the absence of prior hospitalization for HF, physician diagnosis of HF, or past treatment for HF based on the review of data contained in hospital medical records. Patients who developed HF secondary to admission for another acute illness (eg, acute myocardial infarction) or after a procedure or surgery (eg, percutaneous coronary intervention and coronary artery bypass surgery) were excluded. For purposes of the present investigation, patients who died during the index hospitalization (5.5%) were excluded.

Information was collected about demographics, medical history, clinical characteristics, and laboratory test results among hospital survivors of HF through the review of information contained in hospital medical records. This included information about patients' age, sex, race/ethnicity, initial symptoms, body mass index (BMI), physical examination findings, clinical characteristics (eg, initial heart rate and blood pressure), prior comorbidities (eg, stroke, angina, hypertension, and diabetes mellitus), and laboratory findings (eg, serum levels of creatinine, hematocrit, and serum urea nitrogen).

Ejection fraction findings during the index hospitalization were available for only 36.8% of the study sample. We reviewed physicians' progress notes and daily medication logs for the prescribing of selected medications. We examined the use of cardiac medications that have been shown to be of benefit in improving the prognosis of patients with HF (β-blockers, angiotensin receptor blockers, and angiotensin-converting enzyme inhibitors), as well as the use of medications shown to be effective in improving the symptoms of patients with acute HF (digoxin and diuretics). Information about patients' long-term survival status through 2005 was obtained by the review of hospital medical records at all participating medical centers for subsequent hospitalizations or medical care contacts, as well as the review of the Social Security Death Index and death certificates at the Massachusetts State Health Department.

A life-table approach was used to examine mortality after greater Worcester residents were discharged from all area medical centers after an incident episode of HF or a new exacerbation of this clinical syndrome among patients with previously diagnosed HF in 2000 through 2005, including patients with varying lengths of long-term follow-up. Differences in the distribution of selected demographic and clinical characteristics between postdischarge decedents and survivors were examined using χ² and t tests for discrete and continuous variables, respectively. A logistic regression approach was used to identify demographic, medical history, laboratory, and clinical factors associated with a poor prognosis after HF, controlling for various potentially confounding factors. Multivariate adjusted odds ratios and accompanying 95% confidence intervals of factors associated with a poor long-term prognosis at 3 months after hospital discharge, at 1 year after hospital discharge, and at the end of our 5-year follow-up period were calculated in a standard manner.

A total of 2445 greater Worcester residents were discharged from area medical centers after being hospitalized for decompensated acute HF in 2000. The mean age of the study sample was 76 years, 43.4% were men, and approximately three quarters had been previously diagnosed as having HF.

The total death rates were high throughout the 5-year follow-up period among the study cohort (Figure). The all-cause death rates were 37.3% at 1 year after hospital discharge, 52.9% at 2 years, and 78.5% at 5 years. Among patients with an incident episode of HF, the cumulative death rates at 1, 2, and 5 years after hospital discharge were 27.8%, 40.1%, and 62.5%, respectively.

We examined differences in the characteristics of postdischarge decedents overall and separately among those who died during the high-risk first year after hospital discharge. Compared with survivors, postdischarge decedents were older, were generally of white race/ethnicity, and were more likely to have higher levels of serum urea nitrogen and serum creatinine, have a BMI (calculated as weight in kilograms divided by height in meters squared) of less than 25, and have lower blood pressure and hematocrit levels at the time of hospital presentation (Table 1). Compared with survivors, postdischarge decedents were significantly more likely to have been previously diagnosed as having stroke, HF, anemia, renal disease, peripheral vascular disease, and chronic obstructive pulmonary disease. Patients who died during the follow-up period were more likely to have edema and generalized weakness at the time of hospital presentation and were less likely to have chest pain, orthopnea, and nausea or vomiting. Similar differences in these characteristics were observed when we restricted our analysis to patients hospitalized for an incident episode of HF.

In terms of the use of different hospital treatment regimens, postdischarge decedents were significantly less likely to have been treated with each of the cardiac regimens examined except digoxin and diuretics. These findings are summarized in Table 1.

We carried out a multivariate adjusted regression analysis for purposes of identifying factors associated with a poor long-term prognosis separately for those who died during the first year after hospital discharge and for those who died during the entire follow-up period. Given the particularly high death rates during the first 3 months after hospital discharge, we also examined predictors of death during this high-risk period. The results of this analysis showed that the following factors were associated with increased mortality during the first 3 months after hospital discharge: older age and having a history of stroke, a BMI of less than 25, a higher serum urea nitrogen level during hospitalization, and a lower blood pressure or a higher heart rate at the time of hospital presentation (Table 2). Similar factors were associated with the death rates at 1 year after hospital discharge and during the entire follow-up period. In addition, patients with a history of chronic obstructive pulmonary disease and patients with previous HF were more likely to die during the first year after hospital discharge and during the entire follow-up period. The symptoms of chest pain and orthopnea at the time of hospital presentation were associated with significantly lower postdischarge death rates at 1 year after hospital discharge. The presence of edema and a history of peripheral vascular disease were associated with an increased risk of dying after hospital discharge.

Among patients with an incident episode of HF, several factors were associated with increased risk for dying during the entire follow-up period. These included older age, lower BMI, white race/ethnicity, history of chronic obstructive pulmonary disease or stroke, and higher serum urea nitrogen levels and increased heart rates at the time of hospital presentation.

The results of our large observational study suggest that the long-term survival outlook for patients hospitalized with HF in the community setting remains poor. Several readily identifiable demographic, medical history, and clinical characteristics were associated with an adverse long-term prognosis among hospital survivors of acute HF.

Few studies^{3,5-7,9,15,16} have examined the long-term prognosis associated with hospitalization for HF from a community-based perspective. In a large population-based study¹⁶ of Olmsted County, Minnesota, residents, the incidence rates of new-onset HF were unchanged between 1979 to 1984 and 1996 to 2000; however, the 5-year age-adjusted survival rates associated with HF significantly improved during the periods under study (from 43% to 52%). Similarly, in a comprehensive study¹⁷ of 1075 Framingham, Mass, residents from 1950 to 1999 who had new-onset HF, improvements in survival after new-onset HF were estimated to be approximately 12% per decade for men and women enrolled in this longitudinal study. From 1990 to 1999, the 5-year age-adjusted death rates after the development of new-onset HF were 59% for men and 45% for women. The incidence of new-onset HF remained virtually unchanged among Framingham men and declined among Framingham women during the period under study. Improving trends in the 3-year survival rates after new-onset HF also have been demonstrated in a national study¹⁸ from 1984 to 1992 of older and younger Scottish patients with HF.

Albeit based on limited data, these results suggest that the incidence of new-onset HF has been stable over time but that the long-term prognosis for patients experiencing an incident episode of HF may be improving. These improvements in the long-term prognosis may be the result of advances in medical treatment, increased numbers and effectiveness of outpatient HF clinics, greater use of effective pharmacologic and nonpharmacologic regimens for hospital survivors of HF, and the publication of guidelines in the mid 1990s for more effective management of patients with HF.¹⁹

These comments notwithstanding, the long-term mortality observed among our study sample is appreciably higher than that reported among the Olmsted County cohort¹⁶ or in the Framingham Study.¹⁷ These data highlight a different perspective of the same HF epidemic. Given the stable incidence rates but the improved survival noted by the aforementioned studies^16,17 of patients with an incident episode of HF, the death rates in our study of central Massachusetts residents are perhaps not surprising because approximately three quarters of patients in our community-based sample had a history of HF. In addition, most patients in this study population were older than 75 years, and these high-risk patients had many additional comorbidities present. In prior studies, only patients with an incident episode of HF were included, and many of these patients were diagnosed and treated on an outpatient basis, possibly reflecting a less severely ill patient population. Our data suggest that HF in an older population remains a terminal condition for most hospitalized patients.

Similar to most investigations that have examined the long-term prognosis among patients with HF, we report only information concerning all-cause mortality. The identification and monitoring of trends in mortality attributed to HF would be of considerable interest. However, identifying deaths caused by HF is difficult given the advanced age and high prevalence of accompanying comorbidities among our study population, as well as concerns about the accuracy of the cause of death listed on death certificates. Indeed, it could be argued that the high mortality observed in our study was primarily due to these associated comorbidities. The onset of acute HF also may be a marker of the progressive nature of many comorbid states, reflecting a more rapid decline in survival.

To better understand the factors affecting the prognosis after HF, particularly for patients encountered in the community setting, we evaluated the independent prognostic significance of several demographic and clinical factors on long-term survival at different time points. At 3 months after hospital discharge, older age, weight loss, increased heart rate, lower blood pressure, and higher serum urea nitrogen levels were predictive of mortality. Similar prognostic variables were identified in studies evaluating hospital mortality among patients with HF. These data suggest that early mortality after HF is strongly associated with the severity of HF at the time of hospital presentation.

As time passes from the initial hospitalization, BMI and the presence of other comorbidities seem to play an increasing role in mortality associated with HF. At 1 year after hospital discharge, patients with histories of stroke, HF, or chronic obstructive pulmonary disease were at increased risk for an adverse outcome; at the end of our study, each of these important prognosticators and a history of peripheral vascular disease were associated with greater mortality. Our study and other studies^20-22 found inverse associations between BMI and survival among patients with HF. A growing body of evidence suggests that the cause of cardiac cachexia among patients with severe HF is multifactorial.^23,24 The observed increase in mortality among patients with decompensated HF and their weight loss also may be a result of multiple factors, including weight reduction, severity of the underlying HF, and high levels of circulating inflammatory cytokines. It remains unclear if patients who are obese to begin with are protected from this syndrome and the extent to which obesity affects subsequent HF-associated mortality. These issues merit further study, particularly with regard to better understanding the possible mechanisms involved so that appropriate advice and treatment regimens may be prescribed for patients with acute HF.

Reasons for the lower death rates at various time points after hospital discharge among patients with chest pain or orthopnea at the time of hospital presentation are likely to be multifactorial but are unknown. These may be statistically significant but clinically unimportant findings based on the large number of clinical signs and symptoms examined relative to the postdischarge prognosis. Patients with acute HF who have chest pain at the time of hospital presentation may be more likely to have underlying coronary ischemia; because some of these patients may subsequently undergo coronary revascularization, their prognosis may be better than that of patients without chest pain or underlying myocardial ischemia. It is more difficult to postulate a rationale for the better long-term survival of patients who had orthopnea at the time of hospital presentation in our study. This symptom may occur more frequently among patients with mild disease or among those who seek early medical care, but this observation requires additional study before any working hypotheses about its clinical significance may be proffered.

More than a decade ago, Stewart and colleagues suggested that HF was more “malignant” than cancer in a study of Scottish patients who had an incident admission for HF, myocardial infarction, or the 4 most common types of cancers at any Scottish hospital in 1991.²⁵ Among more than 16 000 men and almost 15 000 women studied, the 5-year death rates from HF lagged only behind lung cancer in terms of prognostic importance.²⁵ Our data suggest that this observation remains true today. Although remarkable advances have been made and will continue to be made in the management of patients with HF, it is important that clinicians have a realistic understanding of the long-term prognosis associated with this clinical syndrome. Clinicians also need to be aware of factors that adversely affect patients' long-term survival, which may facilitate targeted treatment toward specific high-risk groups.^4,26 Decisions about the use of more aggressive interventions for patients with HF (eg, surgery, automatic implantable defibrillators, or multivessel percutaneous coronary intervention) must also take into account the presence and the effect of various comorbidities on the long-term survival outlook.

With respect to medical treatment among the Olmsted County cohort,¹⁶ angiotensin-converting enzyme inhibitors were virtually unavailable in the early to mid 1980s but were used by slightly more than half of the patients with HF from 1996 to 2000; the use of β-blockers increased from 10% during the early period (1979-1984) to 31% during the later period (1996-2000) among residents with documented HF.¹⁶ These encouraging increases in the use of effective cardiac regimens among patients with HF indicate a trend toward compliance with current guidelines.^1,19 Diuretics were prescribed to most patients with HF in accord with current recommendations,^1,19 a finding that we also observed in the present investigation. Although one should exercise appropriate caution in interpreting the results from nonrandomized observational studies, our findings support the notion that the use of angiotensin-converting enzyme inhibitors and β-blockers can improve the long-term prognosis among patients with HF, as postdischarge survivors were significantly more likely to have been prescribed each of these effective cardiac medications compared with decedents.

The primary strength of our study is the inclusion of a large sample of patients hospitalized with independently validated HF from a well-characterized northeast metropolitan area. In addition, we collected information about different factors that could affect the long-term prognosis after HF, and our rates of successful follow-up were excellent. The limitations of our study include the inability to systematically examine the long-term prognosis of patients with HF according to the type of left ventricular dysfunction (systolic vs diastolic) given the extent of missing data with regard to ejection fraction findings. We also did not collect information on cause-specific mortality from death certificates.

The results of our study among residents of a large northeast metropolitan area suggest that the long-term survival outlook for patients with HF continues to be poor. We identified multiple characteristics associated with a poor long-term prognosis, several of which might be effectively intervened on. The present results provide further impetus for current efforts to improve the management of patients with HF and for further development and use of long-term strategies to enhance patients' adherence to effective medical and nonpharmacologic treatment regimens.

Correspondence: Robert J. Goldberg, PhD, Department of Community Health, Brown University, Room G-S121, 121 S Main St, Second Floor, Providence, RI 02912 (Robert_Goldberg@Brown.edu).

Accepted for Publication: November 22, 2006.

Author Contributions:Study concept and design: Goldberg and Meyer. Acquisition of data: Goldberg. Analysis and interpretation of data: Goldberg, Ciampa, Lessard, Meyer, and Spencer. Drafting of the manuscript: Goldberg, Ciampa, Lessard, Meyer, and Spencer. Critical revision of the manuscript for important intellectual content: Ciampa, Meyer, and Spencer. Statistical analysis: Goldberg and Lessard. Obtained funding: Goldberg and Spencer.

Financial Disclosure: None reported.

Funding/Support: This study was supported by grant R37 HL69874 from the National Institutes of Health.

Acknowledgment: This research was made possible by the cooperation of the medical records, administration, and cardiology departments of participating hospitals in the Worcester metropolitan area.