Among Medicare beneficiaries hospitalized for heart failure between 2001 and 2005, acute and long-term outcomes were poor and did not improve appreciably over time. Within 30 days of hospitalization for heart failure, more then 1 in 10 Medicare beneficiaries died and more than 1 in 5 were readmitted to the hospital. Nearly half of the readmissions were for cardiovascular reasons. Given the paucity of therapeutic options with demonstrated benefit for acute outcomes, these results are not surprising. To date, no placebo-controlled trial in acute heart failure has shown a short-term survival benefit or decreased hospitalizations.14,15
Identifying therapeutic approaches that improve acute outcomes should remain a top priority.
Chronic outcomes were similarly poor. During 1 year of follow-up, more than 1 in 3 Medicare beneficiaries died and two thirds were readmitted. Nearly 40% of patients were admitted at least twice. At first glance, these findings may seem surprising, given the demonstrated survival benefit associated with ACE inhibitors and beta-blockers in clinical trials of patients with heart failure.1–6
Several factors likely explain the discrepancy. First, clinical trials often exclude elderly patients,16
and databases in which to examine the effectiveness of therapies for treating heart failure in the elderly are limited. There is evidence, however, that patients who may benefit the most from beta-blockers, ACE inhibitors, and angiotensin receptor blockers (ARBs) may be least likely to receive them. Compared to high-risk patients, Lee et al17
found that low-risk patients were more likely to receive these therapies after controlling for survival time and potential contraindications.
Second, the analysis population, which was 60% women and had a mean age of 80 years, may disproportionately represent patients with preserved systolic function,18
and the evidence base for these patients is limited.19–21
Moreover, the use of evidence-based therapies in patients with systolic dysfunction is suboptimal. An analysis of the Medicare Current Beneficiary Survey suggests that the prevalence of ACE inhibitor or ARB use was only 50% among beneficiaries with congestive heart failure.22
Smith et al23
found that the prevalence of beta-blocker use after onset of congestive heart failure increased by 2.4 percentage points per year from 1989 through 2000, but the prevalence was only 29% in 2000. Even among patients with low ejection fraction, the prevalence of beta-blocker use was only 43%.
Other findings are also noteworthy. Cardiovascular and respiratory DRGs dominated readmissions, but renal failure and gastrointestinal hemorrhage were not uncommon. Moreover, only a quarter of readmissions were specifically for heart failure. Consistent with the high prevalence of comorbid conditions at baseline, the readmissions likely reflect the high burden of coexisting disease in patients with heart failure. As work by Setoguchi et al24
has shown, the number of heart failure hospitalizations is an important predictor of mortality. Strategies designed to reduce readmissions must reflect the clinical complexity of patients with heart failure. Second, as shown in an earlier analysis,25
we found that black Medicare patients were less likely than other patients to die in the year after the index admission but more likely to be hospitalized. The data do not allow us to explore possible explanatory factors, including medication adherence,26
and socioeconomic status.28
Third, the volume of heart failure discharges at the hospital level was significantly related to mortality and readmission, but the magnitude of the effect was small. Birkmeyer et al29
found a strong and significant relationship between hospital volume and short-term mortality in several surgical cohorts. More recent evidence suggests that a strong volume–outcome relationship exists in inpatient care for patients with stroke.30
In some ways, the modest volume–outcome relationship we observed is unsurprising. With a high burden of coexisting illness, patients with heart failure are often cared for by multiple specialists in a heterogeneous hospital service, and coordination of such care can be challenging. Moreover, the modest relationship may reflect unmeasured clinical heterogeneity and substantial variation in processes of care.
Combined with an analysis of Medicare beneficiaries hospitalized for heart failure in the 1990s,9,31
our findings suggest that survival following an index hospitalization for heart failure has changed little in 13 years. Kosiborod et al9
reported 30-day mortality of 10% to 11% and 1-year mortality of 32% from 1992 through 1999. Similarly, in an analysis of data from the National Heart Failure Project, a quality-of-care initiative for Medicare beneficiaries hospitalized with heart failure in the late 1990s, Rathore et al32
found 1-year mortality of 36%.
Consistent with the 1-year readmission rates from the National Heart Failure Project,33
we found that two thirds of patients were readmitted within a year of the index hospitalization. It is noteworthy that the 30-day and 6-month readmission rates were markedly higher than those reported by Kosiborod et al.9
Specifically, Kosiborod et al9
found 30-day all-cause readmission rates ranging from 10.2% to 13.8%, whereas we found an all-cause readmission rate of about 23%. However, the rates reported by Kosiborod et al9
are similar to the cardiovascular readmission rates we report (about 13%). The source of this difference is unclear.
Our study has some limitations. First, we relied on ICD-9-CM
diagnosis codes from Medicare claims data, not medical chart review, to identify index heart failure admissions.33
Previous studies suggest that a single inpatient diagnosis of heart failure (ICD-9-CM
code 428.XX, 402.X1, 404.x1, or 404.X3) has greater than 95% specificity for the diagnosis of heart failure.34–36
Second, data regarding left ventricular function are not available in claims, so we were unable to differentiate between systolic heart failure and diastolic heart failure. ICD-9-CM
diagnosis codes specific to diastolic heart failure were introduced in 2003, but the codes have not yet been validated. Third, the results may not generalize to all patients with heart failure. The analysis population consisted of elderly patients with a high prevalence of comorbid illness, and important clinical data were not available (eg, blood pressure, blood urea nitrogen count, and serum creatinine at admission). However, because the analysis included 100% of Medicare fee-for-service beneficiaries who were admitted with a principal diagnosis of heart failure, the findings are representative of a large population of relevant patients. Fourth, information regarding adherence to evidence-based guidelines and performance measures are not available in these data, so we cannot ascertain the extent of important treatment gaps. Finally, claims data are not available during periods of managed care coverage, so we may have underestimated readmission rates to the extent that fee-for-service beneficiaries switched to managed care and were subsequently hospitalized.
In conclusion, in this longitudinal analysis of Medicare claims for 100% of inpatient, fee-for-service admissions from 2001 through 2005, we found that the prognosis for elderly patients with heart failure was poor and improved little over time. Medicare beneficiaries comprise a large majority of heart failure patients, so these findings are highly representative of the heart failure population in the United States. Heart failure is a leading cause of hospitalization of Medicare beneficiaries and will likely remain so with the aging of the Medicare population. The need to identify optimal management strategies for these clinically complex patients is urgent.