Predicting survival in elderly patients with heart failure

Heart failure is the leading cause of hospitalization in people over 65 years of age. More than 5 million Americans have heart failure, resulting in more than 1 million hospitalizations per year, and the condition is the most costly diagnosis-related group.¹ Numerous pharmacologic and nonpharmacologic therapies have become available for the treatment of heart failure patients. Some of these treatments, such as implantable defibrillators and biventricular pacemakers, are costly, and their value in elderly heart failure patients is often uncertain, in part because of the difficulty in reliably determining prognosis in this population. Our study was intended to evaluate independent predictors of long-term survival in a cohort of patients ≥ 70 years old hospitalized with heart failure and to develop a simple risk score for assessing 6-month, 1-year, and 5-year mortality.

Subjects and methods

The study population consisted of 282 subjects ≥ 70 years of age hospitalized at a university teaching hospital from 1990 to 1994 who were enrolled in a prospective randomized trial to evaluate the efficacy of a multidisciplinary heart failure disease management intervention.² All subjects were initially followed for 1 year. The hospital database and National Death Index were used to determine mortality for up to 14 years following the index hospital admission. A Cox proportional hazards model was constructed to identify clinically relevant independent predictors of the duration of survival, and receiver-operator characteristic curves were used to assess the goodness-of-fit of the model (C statistic) for predicting 6-month, 1-year, and 5-year mortality. A bootstrap procedure was performed to assess the internal validity of the model. A simple risk score was then developed, assigning 1 point for each of the 7 independent predictors of survival, and Kaplan-Meier curves were plotted to illustrate survival according to the number of risk factors present at the time of initial hospitalization.

Results

The mean age of subjects was 79 years; 63.5% of subjects were women, and 55.3% were nonwhite. The majority had hypertension, and more than one fourth had diabetes. The mean New York Heart Association functional class was 2.4, and the mean left ventricular (LV) ejection fraction was 42%; 47% of patients had preserved LV systolic function, defined as an ejection fraction ≥ 45%. Mild anemia and mild-to-moderate chronic kidney disease were also highly prevalent. Baseline characteristics of the study population are shown in Table 1.

Table 1. Baseline characteristics of 282 patients with heart failure.

Characteristic

Age, years

79.2±6.1

Female, n (%)

179 (63.5)

Nonwhite, n (%)

156 (55.3)

Hypertension, n (%)

214 (75.9)

Diabetes mellitus, n (%)

80 (28.4)

Prior myocardial infarction, n (%)

121 (42.9)

NYHA functional class

2.4±1.0

Mean systolic blood pressure, mm Hg

158±37

Mean hemoglobin, mg/dL

12.1±1.9

Mean serum urea nitrogen, mg/dL

30±18

Mean creatinine, mg/dL

1.7±0.9

Mean serum sodium, meq/L

139±4

Mean ejection fraction, %*

42±14

NYHA indicates New York Heart Association.

Arch Intern Med.

*Data on ejection fraction available for 222 patients (79%). (Adapted with permission from Huynh BC, Rovner A, Rich MW. Long-term survival in elderly patients hospitalized for heart failure: 14-year follow-up from a prospective randomized trial. 2006;166:1892-1898.

Table 2. Independent predictors of shorter survival time.

Predictor

Hazard ratio

P value

Age, per 5 years

1.14

.01

Serum sodium < 135 meq/L

1.67

.003

Coronary heart disease

1.51

.002

Dementia

2.02

.02

Peripheral artery disease

1.75

.004

Systolic blood pressure, per 10 mm Hg

0.95

.004

Serum urea nitrogen, per 10 mg/dL

1.20

<.001

Arch Intern Med.

(Adapted with permission from Huynh BC, Rovner A, Rich MW. Long-term survival in elderly patients hospitalized for heart failure: 14-year follow-up from a prospective randomized trial. 2006;166:1892-1898.

The median survival for the study population was 894 days (interquartile range, 355-1821 days). Only 13 patients (5%) were still alive at the end of the follow-up period. Bivariate analysis identified 18 variables associated with survival; of these, 7 remained as independent predictors by Cox regression analysis (Table 2). Note that higher systolic blood pressure was associated with longer survival. The C statistics for all-cause mortality at 6 months, 1 year, and 5 years were 0.84, 0.79, and 0.75, respectively, signifying excellent predictive value. To calculate the risk score, continuous variables in the Cox model were dichotomized as follows: age < 75 years vs ≥ 75 years; serum sodium < 135 meq/L vs ≥ 135 meq/L; systolic blood pressure < 120 mm Hg vs ≥ 120 mm Hg; and serum urea nitrogen < 30 mg/dL vs ≥ 30 mg/dL.

The Figure demonstrates survival according to risk score category, where 0 to 1 risk factors were classified as low risk (n = 89), 2 to 3 risk factors were classified as intermediate risk (n = 153), and 4 or more risk factors were classified as high risk (n = 37). One-year mortality rates for the 3 risk categories were 9.0%, 22.2%, and 73.0%, respectively. Five-year mortality rates by category were 57.3%, 79.1%, and 100%. Compared with patients in the low-risk category, the odds ratio for 1-year mortality was 2.9 in the intermediate-risk group and 27.3 in the high-risk group.

Figure. Probability of survival based on the number of prognostic

P

risk factors (< .001). (Reprinted with permission from from

Huynh BC, Rovner A, Rich MW. Long-term survival in elderly

patients hospitalized for heart failure: 14-year follow-up from a

Arch Intern Med.

prospective randomized trial. 2006;166:1892-1898).

Discussion

This study demonstrates that although the long-term prognosis of elderly patients hospitalized with heart failure is poor, there is considerable heterogeneity in survival rates. Thus, although median survival is about 2.5 years, 25% of the population dies within the first year following admission and another 25% survive for 5 years or longer. Moreover, a relatively simple 7-item risk score, derived from clinical data readily available early during the course of hospitalization, effectively stratifies patients into low-, intermediate-, and high-risk categories for mortality within 6 to 12 months of hospital discharge. This information, in turn, can be used to counsel patients and their families and to guide therapeutic decision making. For example, patients in the high-risk category are unlikely to benefit from an implantable defibrillator because these devices have not been shown to reduce mortality within the first year after insertion.^3,4 Instead, focusing on palliative care and compassionate end-of-life planning would be appropriate, while continuing efforts to optimize quality of life.

Our study has several strengths. First, with a follow-up period of up to 14 years, this study provides the longest prospective follow-up of elderly heart failure subjects reported to date. Second, unlike many prior studies, we did not exclude patients with very advanced age, preserved LV systolic function, dementia, advanced renal insufficiency, prior stroke, or other major comorbid conditions^5-7; as a result, our risk score should be applicable to elderly heart failure patients encountered in routine clinical practice. Third, each of the 7 risk factors identified in our study has been shown to confer an unfavorable prognosis in one or more prior reports, thus providing reassurance that the variables incorporated into our model do not merely reflect "chance" associations.^7-11 Finally, all of the variables in our model are derived from the medical history, physical examination, and routine laboratory evaluation. Thus, the risk score can be calculated within hours of the patient's arrival in the emergency department or on the hospital ward, thereby serving as an aid in decision making. Moreover, the risk score does not rely on the results of sophisticated or costly tests (eg, echocardiography) that may not be readily available at the time important therapeutic decisions are being made.

Among the limitations of our study is the small number of subjects enrolled in this prospective clinical trial, which was conducted at a single academic medical center. The results of our analysis therefore require validation in larger and more diverse populations at other institutions. In addition, management of heart failure has evolved considerably since the early 1990s, when patients were originally recruited for this study, and there is evidence that the prognosis for patients with heart failure has improved significantly since the introduction of beta blockers and other therapies.^12,13 However, it is not clear that the prognostic variables identified in our study would be affected by these new therapeutic interventions because none of the 7 factors are directly modified by current treatment modalities.¹⁴ Our study was also unable to provide insight into other outcomes of interest, such as quality of life, rehospitalizations, or cause of death; additional study of factors affecting these outcomes is clearly needed.

Conclusions

Elderly patients hospitalized with heart failure have a highly variable prognosis. A simple 7-item risk score based on data readily available early in the hospital course can be used to effectively estimate short-, intermediate-, and long-term prognosis. This score may be useful in counseling patients and their families, and as an aid in making diagnostic and therapeutic decisions.