Heart failure is a common chronic disease marked by frequent exacerbations often resulting in hospitalization and death (1- 2). At 40 years of age, the lifetime risk of developing heart failure is 1 in 5 (1). It has been the number one reason for admission among Medicare patients and those in the Veterans Affairs (VA) Health Care System (3). Readmission for heart failure occurs in 20% of patients within 30 days of discharge in those over age 65 years in the Medicare health care system. The high rate of hospitalization is a major contributor to the estimated $37.2 billion in cost of heart failure care in the U.S. for 2009 (1).

There are multiple treatments known to prolong survival for patients with heart failure, and many of these interventions have been incorporated into the management of these patients (2). In the VA Health Care System, use of certain guideline-recommended treatments is now at a high level (4). However, it is unclear if the increased performance on process of care measures for hospitalized patients has led to improvements in outcome.

The purpose of this study was to determine if recent mortality and readmission rates have improved within the VA Health Care System.

We used data from the VA's National Patient Care Database, which includes both inpatient (discharge diagnoses, major procedures) and outpatient data (diagnoses, location of encounter). We included the first admission per patient with the principal discharge diagnosis of heart failure (International Classification of Diseases-9th Revision [ICD-9] 428.xx, 429.3, 402.01, 402.11, 402.91, 425.xx, or diagnosis-related group of 127), the admission occurred from 2002 to 2006, and the patient was at least 18 years of age. We examined data back to 1999 to exclude prior heart failure admissions. Hospitalizations at non-VA facilities but paid by the VA (8% of all heart failure admissions) were included.

In secondary analyses, we included all heart failure hospitalizations from 2002 to 2006 (incident and prevalent cases combined).

In secondary analyses, hospitalizations were excluded if the patient was admitted to a non-VA facility (but care paid by the VA, 8%). This exclusion did not importantly change the results, and therefore, all first admissions were included in the primary analysis.

Prior diagnoses were defined as any inpatient discharge diagnosis (primary or secondary) or at least 2 outpatient encounters with the diagnosis in the year prior to admission. Comorbid conditions were defined using ICD-9-Clinical Modification diagnostic coding of the Charlson comorbidity criteria (5). Additional comorbidities (not included in the Charlson score) were defined as follows: atrial fibrillation or flutter (ICD-9 427.3x), hypertension (ICD-9 401 to 405), ischemic heart disease (ICD-9 410 to 414), and depression (ICD-9 300.4, 301.12, 309.0, 309.1, 311).

Laboratory data were obtained for serum creatinine, blood urea nitrogen, sodium, blood hemoglobin, hemoglobin A1C, B-type natriuretic peptide, and troponin T or I. We used the value closest to admission that was obtained during hospitalization or during the week before admission for all values except troponin and hemoglobin A1C. For troponin, we used the highest value during admission. For hemoglobin A1C levels, we used the value closest to admission that was obtained during hospitalization or during the month before admission. Troponin T or I values were classified as below or above the detection threshold.

Medication data were obtained using VA pharmacy records for filled prescriptions. We examined 2 periods: 180 days before admission, and 90 days following discharge. Patients were included in the analysis of medication use for a given period if they had at least 1 filled prescription for any medication during that time.

We used the VA's Beneficiary Identification Records Locator Subsystem death file and the Social Security Death Index to determine survival. The primary outcome was survival at 30 days following admission. Secondary mortality outcomes included in-hospital survival, and survival at 1 year. Other secondary outcomes included readmission rates at 30 days. Hospitalizations were categorized as all cause, heart failure (primary or secondary diagnosis at discharge), or primary heart failure (primary diagnosis at discharge). We determined VA office (face-to-face) visits during 14 days following discharge for those alive at discharge.

For comparing 2002 and 2006 results, Pearson chi-square analysis was used to evaluate categorical variables at and t tests were used to evaluate differences in continuous variables. A test for trend was used to evaluate differences across all years, 2002 to 2006. We used logistic regression and generalized estimating equations to adjust differences in time-specific mortality and readmission for patient characteristics while controlling for clustering of patients within facilities. A 2-sided p value <0.05 was considered statistically significant.

The study population consisted of 50,125 patients hospitalized with a primary discharge diagnosis of heart failure from 2002 through 2006. The admission rate remained relatively constant at about 5 heart failure admissions per 1,000 patients per year that received treatment within the VA (Figure 1). However, the total number of patients hospitalized with a primary discharge diagnosis of heart failure increased from 9,655 in 2002 to 10,151 in 2006, which was proportional to the increase in patients seeking care within the VA system.

Changes in patient characteristics of those hospitalized with heart failure over time are shown in (Table 1). The majority of patients (>90%) were men and 22% to 24% of those were reported to be black. However, race was available for only 80% of patients. Approximately one-half of the patients had a prior diagnosis of heart failure as an outpatient, close to 40% had a prior diagnosis of ischemic heart disease, and one-third had diabetes. Increases in comorbid disease coding over time was noted for most diagnoses, though prior ischemic heart disease and chronic obstructive pulmonary disease remained stable and a history of myocardial infarction decreased from 2002 to 2006. The summary Charlson comorbidity score, based on comorbidities during the index admission, increased from 1.72 to 1.89 (p < 0.0001). In the year before the first heart failure admission, 27% were hospitalized for nonheart failure diagnoses in 2002 compared with 26% in 2006 (p = 0.27).

Laboratory data (Table 2) were available for the vast majority of patients (>80%) for sodium, hemoglobin, creatinine, blood urea nitrogen and troponin. Values for hemoglobin A1C were only available in one-third and B-type natriuretic peptide values in a little less than one-half of patients in 2006. From 2002 to 2006, there were slight decreases in the fraction of patients with elevated blood urea nitrogen and hemoglobin A1C. Of those tested in 2006, one-half had elevated troponin and 72% had markedly elevated B-type natriuretic peptide (>400 pg/ml). Although data on weight were not available, coded obesity increased from 10% in 2002 to 12% in 2006 (p < 0.0001).

Medication data were available for 87% of patients during the 6 months before admission and in 89% during the 3 months following admission (Table 3). Large increases were seen for beta-blockers, largely due to an increase in those recommended by guidelines (carvedilol, metoprolol succinate, or bisoprolol). Use of angiotensin-receptor blockers increased whereas use of angiotensin-converting enzyme (ACE) inhibitors and loop diuretics changed minimally. In contrast, use of digoxin dropped markedly. During the 3 months following discharge, similar trends in medication use were noted.

Heart failure hospitalization outcomes are shown in (Table 4). Length of stay decreased by approximately 1 day (7.1 to 5.9 days) from 2002 to 2006 (Table 4), though right heart catheterization, coronary angiography, implantable cardioverter-defibrillator placement, and cardiac resynchronization therapy during hospitalization all increased slightly. The fraction of patients discharged home was over 85%, and this value increased over time. Approximately one-half of patients had a face-to-face follow-up visit within 14 days, and this number increased only slightly from 2002. Rehospitalization at 30 days in 2006 was 5.6% for heart failure (primary diagnosis) and 16.5% for any cause. Rehospitalization rates for heart failure increased slightly but significantly (p < 0.001) over time from 2002 to 2006 (Figure 2). Total hospitals days over the 1 year following admission were not significantly different (Table 4). A length of stay above 5 days was significantly associated with readmission for any cause at 30 days (18% vs. 14% for those with a length of stay ≤5 days, p < 0.0001) but not for readmission with heart failure as the primary diagnosis (5.8% vs. 5.5%, p = 0.33).

When all heart failure admissions were included in the denominator, readmissions were much higher (22.5%), indicating that those hospitalized for heart failure for the first time are less likely to be readmitted than those with prior heart failure admissions.

In 2002, mortality was 7.1% at 30 days and 28% at 1 year following an admission for heart failure. By 2006, survival following a heart failure admission improved at multiple time points, including in-hospital, 30 days, and 1 year following admission ((Table 4), Figure 3). To determine if the prognostic value of coded diagnoses changed from 2002 to 2006 (which might occur if coding became more aggressive), we determined the odds ratio for an increase in the Charlson comorbidity score adjusted for age separately for 2002 and 2006. The odds ratio for 1-year mortality for a 1 point increase in the Charlson score was similar in 2002 (1.16, 95% confidence interval [CI]: 1.12 to 1.19) and 2006 (1.15, 95% CI: 1.11 to 1.19).

After adjustment for patient characteristics including laboratory values, a significant reduction in mortality was observed from 2002 to 2006 (Figure 4). After adjustment for patient characteristics, laboratory values, and clustering of patients within hospitals, the odds ratio for a heart failure rehospitalization at 30 days increased slightly over time (1.21, 95% CI: 1.04 to 1.41 for 2006 vs. 2002). In contrast, the risk of death at 30 days following admission dropped markedly in 2006 versus 2002 (odds ratio: 0.54, 95% CI: 0.47 to 0.61) (Figure 4). Further adjustment for the increase in use of potentially life-prolonging medication (beta-blockers, ACE inhibitors, angiotensin-receptor blockers) increased the odds ratio for mortality by 5% (p > 0.1).

Heart failure readmission and survival by age group is shown in (Figure 5)A and (Figure 5)B. Readmission rates for heart failure were slightly higher in the older and younger age groups, but there were no clear differences over time. Mortality was reduced from 2002 to 2006 in all age groups.

The VA Health Care System has been known to provide high levels of recommended process of care for heart failure as determined by performance measures from the Center for Medicare and Medicaid Services (4). However, it has been less clear if outcomes such as mortality or rehospitalization have improved. Our study finds that survival has improved following a VA heart failure hospitalization from 2002 to 2006 but with a probable increase in rehospitalization rates. It should be noted that the rehospitalization data are incomplete because we could not determine trends in non-VA readmissions for heart failure that were not paid by the VA.

Our finding of reduced mortality following an admission for heart failure is consistent with improved in-hospital and discharge treatment for heart failure. Other studies of VA populations have noted increased use of beta-blockers and ACE inhibitors over time (6- 7), and similar improvements in heart failure care have been seen in the general population. The CHS (Cardiovascular Health Study) of older Americans with heart failure found that use of ACE inhibitors increased 2.3% per year and beta-blocker use increased 2.4% per year from 1989 to 2000 (8). Additional studies are needed to determine how much of the decline in observed mortality rates can be explained by increased use of guideline-recommended medications.

Several potential explanations for the observed decrease in mortality are unrelated to improved care. First, patients may be less ill. Laboratory data of renal function, hemoglobin, and sodium suggested that severity of disease may have decreased slightly. However, the vast majority of coded comorbidities increased over time. The Charlson comorbidity score, a measure of comorbidities during the index admission increased significantly from 2002 to 2006. Reduced mortality could also be explained by a higher fraction of elective hospitalizations for devices such as defibrillators. Although we could not determine the reason for admission, few patients underwent device placement during hospitalization.

Our finding that comorbidities increased over time suggests that patients hospitalized in 2006 were more ill than those hospitalized earlier. This increase in risk could be explained by providers managing more heart failure as an outpatient, thus raising the illness threshold for admission. Chart review data from non-VA U.S. populations have also found that hospitalized patients with heart failure are increasingly at higher risk of death (9). Our results are also consistent with data from Minnesota that indicate an increase in hypertension and diabetes among hospitalized patients with heart failure (9). However, it is also possible that the increase was due to more aggressive coding without a true change in comorbidity. The latter was suggested by the slight decrease in abnormal lab values from 2002 to 2008. To further evaluate this issue, we compared the predictive value of the Charlson comorbidity score for patients hospitalized in 2002 with those hospitalized in 2006 (more coded diagnoses). If coding practice became more aggressive (e.g., less severe renal dysfunction now coded as renal failure), then we would expect that coded diagnoses would become less predictive of mortality. We found that Charlson score remained similarly predictive of mortality (odds ratio: 1.15 to 1.16 for 30-day mortality), suggesting that coding practices were similar in 2002 and 2006. Laboratory evidence for severity of illness (hyponatremia, anemia, renal dysfunction) was not markedly changed from 2002 to 2006, suggesting that the severity of heart failure was similar over this period.

Surprisingly, the trends in outcome for rehospitalization and mortality showed divergent patterns. Even though we observed clear decreases in mortality over time, readmission rates actually increased slightly. An increase over time was more evident after adjustment for patient characteristics. We would have expected both mortality and rehospitalization to decrease if increases in guideline-recommended treatment with beta-blockers and ACE inhibitors occurred because these medications have led to reduced mortality and hospitalizations in clinical trials (2). It is also possible that these recommended therapies have a more robust impact on mortality than on readmission for heart failure. Whereas increased implantable cardioverter-defibrillator use would be expected to lead to improved mortality without an impact on heart failure hospitalizations, the increase use observed was not great enough to explain the decline in mortality.

Studies in non-VA populations have found similar outcome trends. Mortality for patients with heart failure has decreased in U.S. Medicare (10), other U.S. (11- 14), and non-U.S. (15- 17) populations up through the early 2000s. Our findings indicate that this decline in mortality is also present in the VA and is continuing at least through 2006. Our observation that rehospitalization rates are flat or increasing is also consistent with most trend studies including U.S. Medicare (10) and non-U.S. studies (15,17). Other data from U.S. male physicians (11) and data from Minnesota (14) report a stable incidence of heart failure, and recent data from the National Hospital Discharge Survey indicate a slight increase in hospitalization rates from 1995 to 2004 (18). Other outcome studies of hospitalized heart failure patients (19- 21) have noted higher mortality rates. However, these studies examined populations with a higher mortality with older age (mid-70s vs. 69.5 years for our study) and with a higher fraction of white patients (shown to have worse survival than nonwhite patients [22]).

The explanation for the lack of decline in heart failure rehospitalization is unclear. Length of stay has declined markedly, and it is possible that patients are less stable at the time of discharge than in prior years. Although recommended use of medical therapies at discharge have improved (ACE inhibitors and beta-blockers), other interventions may have more of an impact on rehospitalization such as intensive patient education before discharge, early follow-up, and medication reconciliation. It is unclear to what extent the use of these interventions has changed in the VA Health Care System. Finally, the increase in survivors following discharge increased the opportunity for readmission. Patients alive in 2006 that would have died in 2002 are likely to have advanced disease and a high rate of readmission.

Our study raises concerns that use of rehospitalization for heart failure as a marker of poor care may be flawed, as it did not track with mortality in our study. It is possible that many of the rehospitalizations were not “preventable.” A prior randomized trial of intensive primary care follow-up following a discharge for chronic disease (including heart failure) noted that admissions actually increased though patients rated their health better (23). It is also likely that some heart failure readmissions were lifesaving, and more readmissions or a longer length of stay for heart failure may improve the probability of long-term survival.

Elderly veterans are eligible for Medicare and an unknown number were receiving dual care (VA and non-VA). Thus, the VA rehospitalization rate should be considered a lower bound of the actual rehospitalization rate. However, the relative changes over time in readmission rates should not be affected by dual care unless the proportion of dual care changed substantially over time. We limited our cohort to patients initially hospitalized with a primary discharge diagnosis of heart failure. Whereas this definition has been shown to be highly specific (24), it will not include heart failure episodes that were not the primary reason for admission. We did not have data on body mass index, though changes in coded obesity were small. We also did not have data on left ventricular ejection fraction and the relative changes in mortality for those with and without reduced left ventricular ejection fraction are unknown. Finally, we did not have data on cause of death, and it is possible that trends in mortality from heart failure may have differed from trends in mortality from other causes.

In summary, we found that the VA hospitalization rate for heart failure has remained stable from 2002 to 2006. Although patients have increasingly more comorbidities, their mortality has declined significantly. In contrast, rehospitalization rates for heart failure have not fallen and may have increased. Additional studies are needed to determine the reasons for the decline in mortality and to determine which fraction of hospitalizations are preventable.