Triage After Hospitalization with Advanced Heart Failure: The ESCAPE Risk Model and Discharge Score

Abstract

Objectives

Identifying high-risk heart failure (HF) patients at hospital discharge may allow more effective triage to management strategies.

Background

HF severity at presentation predicts outcomes, but the prognostic importance of clinical status changes due to interventions is less well described.

Methods

Predictive models using variables obtained during hospitalization were created using data from ESCAPE and internally validated by bootstrapping method. Model coefficients were converted to an additive risk score. Additionally, data from the FIRST (Flolan International Randomized Survival Trial) was used to externally validate this model.

Results

Patients discharged with complete data (n=423) had 6-month mortality and death or rehospitalization rates of 18.7% and 64%. Discharge risk factors for mortality included BNP, per doubling (Hazard Ration [HR]: 1.42, 95% confidence interval [CI]: 1.15–1.75), cardiopulmonary resuscitation or mechanical ventilation during hospitalization (HR: 2.54, 95% CI: 1.12–5.78), blood urea nitrogen, per 20-U increase) (HR: 1.22, 95% CI: 0.96–1.55), serum sodium, per unit increase (HR: 0.93, 95% CI: 0.87–0.99), age >70 (HR: 1.05, 95% CI: 0.51–2.17), daily loop diuretic, furosemide equivalents >240 mg (HR: 1.49, 95% CI: 0.68–3.26), lack of beta-blocker (HR: 1.28, 95% CI: 0.68–2.41), and 6-minute walk, per 100 feet increase (HR: 0.955, 95% CI: 0.99–1.00; c index 0.76. A simplified discharge score discriminated mortality risk from 5% (score=0) to 94% (score =8). Bootstrap validation demonstrated good internal validation of the model (c index 0.78, 95% CI: 0.68–0.83).

Conclusions

The ESCAPE discharge risk model and score refine risk assessment after inhospital therapy for advanced decompensated systolic HF, allowing clinicians to focus surveillance and triage for early life-saving interventions in this high-risk population.

One of the difficult clinical challenges in the management of heart failure (HF) is objectively defining the patient population at highest risk of adverse events. The vast majority of the systolic HF population consists of clinically stable outpatients with mild-moderate symptoms who are treated with neurohormonal antagonists and antitachycardia electrical therapies (1). The short-term risk for hospitalization and death in these patients is relatively low. The development of clinical decompensation, which typically includes the signs and symptoms of congestion, is the trigger that brings this population to medical attention. In fact, 75% of HF admissions occur in patients with a preexistent diagnosis (2). The adverse event risk likely varies during a HF hospitalization depending on response to therapy. However, there are currently no risk stratification tools that triage patients following a period of relative decompensation, a vulnerable time in the natural history of this disease when the short-term risks for rehospitalization and mortality are high (3,4).

Several risk stratification approaches in patients with HF have been established (5–11). However, few patients in these analyses had advanced functional limitations (New York Heart Association [NYHA] functional class IV symptoms), the cohort at highest risk of events (10). Additionally, in-hospital mortality has been included in many prior risk models and may limit the predictive accuracy in chronic, ambulatory advanced HF. Thus, factors portending increased risk of recurrent events after discharge in patients with recent NYHA functional class IV symptoms who are discharged alive are currently less well known. Furthermore, patients may be better characterized by clinical status after systematic optimization of therapy than at the time of admission from varied outpatient regimens.

Effective risk stratification also impacts decisions regarding resource allocation such as high level surveillance, early intensive disease management, or aggressive pursuit of high-risk interventions such as left ventricular (LV) assist devices or cardiac transplantation (1). This is because patients at high risk of events are most likely to derive maximum clinical benefits from many of the newer resource-intensive strategies.

Using the unique population of the North American Evaluation Study of Congestive Heart Failure and Pulmonary Artery Catheterization Effectiveness (ESCAPE) trial (12), we provide a comprehensive analysis of the relationships between discharge clinical factors and 6-month mortality after optimization of therapy during a hospitalization for patients with severe advanced HF. The aim of this investigation was to develop a model and risk score using patient data routinely available at discharge that would be clinically useful in triage decisions for patients with recent HF decompensation requiring hospitalization. Further, using these identifiers, the goal of this analysis was to create a simple risk score based on discharge variables that could be used as an initial screen to identify patients at high and low risk of recurrent events.

METHODS

Patient population

The ESCAPE trial enrolled 433 patients hospitalized with advanced HF at 26 sites in the United States and Canada between January 2000 and November 2003. The design, primary endpoints, and results of the ESCAPE trial have been previously published (12). Briefly, patients hospitalized with severe symptomatic HF despite recommended therapies were randomly assigned to receive clinical assessment or pulmonary artery catheter (PAC)-guided therapy. Patients met the following inclusion criteria within the past year: 1) an urgent visit to the emergency department,;2) treatment during the proceeding month with more than 160 mg of furosemide daily; 3) therapy with an angiotensin-converting enzyme (ACE) inhibitor and diuretic for at least 3 months; 4) LV ejection fraction of ≤0.30%; 5) systolic blood pressure of ≤125 mm Hg; and 6) at least 1 sign and 1 symptom of congestion. The exclusion criteria included: 1) serum creatinine >3.5 mg/dL; 2) prior use of dobutamine or dopamine >3 μg/kg/min; or 3) any prior use of milrinone during the current hospitalization. The therapeutic target in both groups was resolution of clinical symptoms and signs of congestion (orthopnea, edema, and jugular venous pressure elevation) with the additional goals in the PAC group of achieving a pulmonary capillary wedge pressure of ≤15 mm Hg and a right atrial pressure of ≤8 mm Hg. Medication use was not specified, but use of intravenous inotropic agents was discouraged. ACE inhibitor and beta-blocker doses were titrated in the outpatient HF programs at these selected centers during the 6 months after randomization according to patient tolerability and current guidelines (1). Diuretics were adjusted both during and after hospitalization to optimize fluid balance without progressive deterioration in renal function.

Data collection and definitions

Selected demographics, baseline characteristics, laboratory values, quality of life indices, and physiologic parameters were collected at baseline and throughout the hospitalization as well as at several time periods during follow-up using standard data collection forms. Specific instructions and definitions for all variables were provided to the investigative sites to assist with form completion. Clinical variables used in the trial were defined in the Manual of Operations. The survival status of each patient was determined. The all-cause mortality and recurrent hospitalization data were ascertained by the site investigators.

Statistical methods

The baseline characteristics of the study patients were summarized as frequencies and percentages for categorical variables and by the median of the 25th and 75th percentiles for continuous variables.

During the conduct of ESCAPE, B-type natriuretic peptide (BNP) levels were measured using 2 distinct assays (Shinogi, Osaka, Japan, and Biosite, San Diego, California). While BNP assays were available on nearly everyone, only data from patients in whom BNP levels were measured by the Shinogi assay and with no missing variables (n = 255) were used for mortality modeling in this analysis (Table 2).

Table 2.

Multivariate Discharge Predictors of Death in ESCAPE

Variables	HR	95 % CI	Chi-square	P
Discharge BNP (per doubling)	1.408	1.174, 1.689	13.57	0.0002
Cardiac arrest or mechanical ventilation (yes/no)	2.755	1.225, 6.197	6.00	0.0143
Discharge BUN (per 20 mg/dL increase)	1.295	1.044, 1.605	5.65	0.0175
Discharge sodium (per unit mEq/L increase)	0.952	0.896, 1.011	2.54	0.1113

The c-index for this model is 0.758. HR, hazard ratio; CI, confidence interval; BNP, brain natriuretic peptide; BUN, blood urea nitrogen.

Using Cox proportional hazards method, a model was created that was derived from a group of univariately predictive discharge variables. The end point was time to death at 6 months. A bootstrapping method was then used to determine which variables would be included in the model. The data were re-sampled 1000 times, fitting a Cox proportional hazards model that selected variables in a stepwise manner. The baseline data were initially evaluated to determine whether or not a non-linear form of the variable would provide better predictability. The following variables were used as linear terms, unless otherwise indicated. The candidate baseline clinical predictor variables were age (=0 if age <65, =age –65 if age >65), systolic blood pressure (SBP [0 if SBP >120 mm Hb, =120 –SBP otherwise), blood urea nitrogen (BUN), 6-minute walk (feet), serum sodium, hemoglobin, creatinine, and ischemic etiology (yes/no). As the discharge model included events and/or alterations in clinical parameters that occurred during the hospitalization, additional variables were found to be relevant and also included in the model: cardiopulmonary resuscitation or mechanical ventilation (1 if the patient required, 0 otherwise), beta-blocker use at discharge, ACE inhibitor use at discharge, high diuretic dose (1 if patient was treated with diuretic doses equivalent to furosemide >240 mg/day, 0 otherwise). Loop diuretic equivalents were calculated as follows: 1 mg furosemide = 0.5 mg torsemide = 0.025 mg bumetanide. Only variables with a P value ≤0.10 were included and those with a P value ≤0.05 remained in the model. The frequency with which each variable fit in the final model was determined and those variables that remained in >50% of the 1000 final models were included.

We also constructed a simplified score model for mortality based on discharge data. For this model, the significance levels of the variables were ignored. A multivariable model was constructed based on dichotomized measures (BNP, BUN >40 mg/dl) so that the estimated regression coefficients were similar, or were simple multiples of each other. This score included 8 clinical variables with scores of 1 point possible for each, except for BUN and BNP, to which additional points were assigned for the highest values. A maximum of 13 points was possible.

Finally, we used the data from FIRST (Flolan International Randomized Survival Trial) to externally validate the ESCAPE 6-month mortality model developed in this study (13). Details of FIRST have been previously published (13). Briefly, this trial evaluated the effects of epoprostenol (Flolan) in patients with NYHA functional class IIIB/IV HF and decreased LV ejection fraction. Patients were eligible for enrollment if severely compromised hemodynamics were documented while the patient was receiving a regimen of digoxin, diuretics, and an ACE inhibitor. Patients were randomly assigned to receive epoprostenol infusion or standard care. The trial was terminated early because of a strong trend toward decreased survival in the patients treated with epoprostenol. Chronic intravenous epoprostenol therapy was not associated with improvement in distance walked, quality of life, or morbid events. Of the variables in the ESCAPE model, 2 (diuretic dose and BNP levels) were not available in the FIRST dataset. The ESCAPE model for 6-month mortality was redeveloped in the absence of these 2 variables, and the new ESCAPE model (without diuretic dose and BNP level) was validated in the FIRST study dataset. All analyses were performed using SAS statistical software (SAS Institute, Inc., Cary, NC).

RESULTS

Among 433 patients enrolled in the ESCAPE trial, 18.7% died and 64% had death or rehospitalization at 6-month follow-up. The baseline demographics and clinical characteristics of patients enrolled in the ESCAPE trial stratified by the outcome of a 6-month death are shown in Table 1. Patients who died were more likely to be older, have previous coronary disease (prior myocardial infarction or prior coronary revascularization), and lower systolic and diastolic blood pressure. Additionally, BUN, serum creatinine, and BNP were significantly higher in patients who died compared with those who did not. Finally, the median 6-minute walk distance was significantly shorter with a strong trend towards lower peak V02 in patients who died.

Table 1.

Baseline clinical characteristics

Characteristics	Death		P
	No (n=338)	Yes (n=83)
Age, yrs	56 (46, 65)	60 (50, 74)	0.010*
Females, No. (%)	91 (27)	19 (23)	0.454
Race, nonwhite, No. (%)	133 (39)	35 (42)	0.638
Weight, kg	85 (71, 99)	80 (66, 93)	0.130*
Etiology, No. (%)
Ischemic	159 (47)	54 (65)	0.003
Non-ischemic	177 (53)	29 (35)	0.004
Hypertension, No. (%)	165 (49)	34 (41)	0.183
Diabetes mellitus, No. (%)	108 (33)	29 (36)	0.526
Current smoking, No. (%)	41 (14)	11 (15)	0.798
Prior MI, No. (%)	138 (41)	49 (59)	0.003
Prior stroke, No. (%)	31 (9.2)	9 (11)	0.653
Prior PCI, No. (%)	74 (22)	24 (29)	0.184
Prior CABG, No. (%)	92 (27)	30 (36)	0.116
Heart rate, bpm	81 (70, 91)	80 (71, 94)	0.342*
Blood pressure, mm Hg
Systolic	105 (94, 117)	99 (90, 113)	0.016*
Diastolic	68 (60, 74)	62 (60, 70)	0.019*
LVEF, %	20 (15, 25)	19 (15, 25)	0.297*
Serum sodium, mEq/L	137 (135, 140)	136 (132, 138)	0.001*
BUN, mg/dL	27 (19, 38)	41 (24, 65)	<0.001*
Serum creatinine, mg/dL	1.3 (1.0, 1.7)	1.7 (1.3, 2.3)	<0.001*
Baseline BNP, pg/mmol	518 (194, 1052)	994 (371, 1579)	<0.001*
Peak VO2	9.9 (8.1, 11.7)	8.9 (7.0, 10.2)	0.080*
6-minute walk, ft	367 (0, 792)	121 (0, 532)	0.011*
Any ACE-I at baseline? n (%)	270 (79.9)	60 (72.3)	0.132
Did pt receive Beta blocker baseline? n (%)	220 (65.3)	43 (51.8)	0.023
Base Angiotensin II antagonist drugs n (%)	59 (17.5)	16 (19.3)	0.698
Diuretic at baseline > 240 mg n (%)	111 (32.8)	43 (51.8)	0.001

BNP=b-type natriuretic peptide; bpm= beats per minute; BUN=blood urea nitrogen; CABG=coronary artery bypass grafting; CHF=congestive heart failure; LVEF=left ventricular ejection fraction; MI=myocardial infarction; PCI=percutaneous coronary intervention; VO2=oxygen consumption.

Continuous variables are presented as median (25th, 75th) unless otherwise indicated.

^*P value based on non-parametric test.

Variables associated with 6-month mortality in patients with advanced HF discharged from the hospital in descending order of their model chi-square values are shown in Table 2. The strongest association with death was higher discharge BNP levels. The model c-index was 0.76, suggesting reasonably good ability of the model to discriminate between those who died and 6-month survivors. Furthermore, bootstrap methods demonstrated good internal validation of the model (c-index 0.78, 95% confidence interval: 0.68–0.83).

This model was validated externally in the FIRST study dataset. Since the FIRST study did not have information on diuretic dose and BNP level, a new ESCAPE model was developed without these 2 factors (c-index 0.74). This new ESCAPE model (without diuretic dose and BNP level) showed modest discriminatory ability to identify patients who died compared with those who survived (c-index 0.65).

Figure 1 shows the reliability of the model predictions for 6-month mortality. The model was reasonably accurate in estimation of events among patients at varying risks.

Figure 1.

Observed versus predicted mortality for the score model. Sample size in parentheses.

Risk score

The multivariable model for 6-month mortality was converted into a coefficient-based simple additive risk score (Table 3). This score included 8 clinical variables with scores of 1 point possible for each, except for BUN and BNP where additional points were assigned for the highest values. A maximum of 13 points was possible. The summation of points assigned for each predictor led to the prediction of overall 6-month mortality risk. The use of categorized measures (rather than of continuous measures) for the ease of calculating the score resulted in additional variables, which were not significant in the original model (Table 2) to be included in the risk score (age >70 years, discharge 6 minute walk test, diuretic dose measures >240 mg/day and no beta-blocker therapy). Table 4 demonstrates estimated probability of dying within 6 months of hospitalization based on the discharge score model. The majority of patients had a risk score ≤5. As shown, the simplified discharge score allowed discrimination of the wide gradient of risks (5% risk of death with score=0 to 94% with score =8).

Table 3.

Simplified Discharge Score Model of Mortality from ESCAPE

Criteria (based on discharge measurements)	Score if Yes, (No = 0)
Age >70	1
BUN >40	1
BUN >90*	1
6-minute walk < 300 feet	1
Sodium <130 mEq/L	1
CPR/mechanical ventilation (yes/no)	2
Diuretic dose >240 at discharge (yes/no)	1
No beta-blocker at discharge	1
Discharge BNP >500	1
Discharge BNP >1300	3
TOTAL OF COLUMN 2 (SCORE)

BNP=b-type natriuretic peptide; BUN=blood urea nitrogen; CPR=cardiopulmonary resuscitation.

This model includes terms which are not statistically significant. However, all terms in the model have approximately the same estimated effect (except for assist which has about twice the effect).

This model has a c-index of 0.739, which is comparable to the continuous multivariate models.

Diuretic dose is represented in furosemide equivalents.

^*If the BUN level is greater than 90 then both BUN variables are coded as 1.

Table 4.

Estimated Probability of Dying in 6 Months Based on the Discharge Score Model

SCORE	No. of patients	Observed deaths	Observed mortality	Estimated probability of dying (no BNP)	Estimated probability of dying (with BNP)
0	91	7	0.077	0.053	0.033
1	125	13	0.104	0.103	0.065
2	114	19	0.167	0.189	0.123
3	53	14	0.264	0.322	0.223
4	29	13	0.448	0.492	0.368
5	15	12	0.800	0.664	0.543
6	4	3	0.750	0.801	0.708
7	1	1	1.000	0.891	0.831
>/=8	1	1	1.000	0.943	0.909

BNP=b-type natriuretic peptide.

The estimated probabilities for death with and without BNP are shown.

In this model, one point is added for BNP >500 and 3 additional points for BNP >1300.

DISCUSSION

Despite advances in the management of advanced acute decompensated HF, patients have a very high risk of mortality and rehospitalization during the early period following hospitalization. In the ESCAPE trial, 1 of every 5 patients died, and 2/3 reached the combined endpoint of death or rehospitalization at 6 months despite a relatively young median age (60 years). Risk stratification using clinical and laboratory information at the time of discharge may allow more precise risk estimation, closer follow-up of the high-risk group, and timely allocation of limited resources of advanced device therapies (including LV assist devices) and cardiac transplantation before patients have deteriorated to become ineligible.

The clinical determinants of cardiac events and mortality in patients with severe advanced decompensated HF due to systolic dysfunction are complex. Nonetheless, as shown in our study, much of the important prognostic information is contained in the clinical characteristics representative of the severity of LV dysfunction. After age, these include circulatory-renal compromise as indicated by hypotension, elevated natriuretic peptide levels, high BUN levels, and the need for high doses of diuretics to maintain a normal volume status. Clinical instability defined by cardiac arrest, the need for mechanical ventilation, intolerance to beta-blocker therapy, and significant functional limitations, defined in this analysis by a short 6-minute walk distance, were also adverse prognostic factors. Taken together, these variables were nearly complete in explaining 76% of variation in the incidence of death at 6-month follow-up and were reliable in accurately differentiating various risk categories. We further developed a simple bedside risk prediction score that can be used to accurately estimate the probability of death for patients with NYHA functional class IV HF. This discharge risk score encompasses a wide range of risks in this advanced HF population, making this tool valuable for clinical decision making and patient counseling. Furthermore, the ESCAPE study model is equally predictive as the widely validated hospitalization heart failure risk models the ADHERE (Acute Decompensated Heart Failure National Registry) study (2) and the EFFECT (Enhanced Feedback for Effective Cardiac Treatment) study (15), as shown in Table 5.

Table 5.

ESCAPE, EFFECT, ADHERE Model Comparison

Trial	n (derivation cohort)	Predictors	Endpoints	c-index
ESCAPE	423	BNP Cardiac arrest or mechanical ventilation Sodium level	6-month mortality	0.76
EFFECT	2624	Age Systolic BP Respiratory Rate Sodium level BUN Comorbid conditions: cerebrovascular disease, dementia, COPD, hepatic cirrhosis, cancer	30-day mortality	0.8
			1 year mortality	0.77
ADHERE	33046	BUN Systolic BP Age Heart Rate Serum Creatinine	In-hospital mortality	0.759*

^*AUC

The ESCAPE risk score extends the existing literature evaluating risk factors for death in patients presenting with ambulatory HF into the spectrum of advanced decompensated disease. The major independent risk factors for death in most of these studies are nonetheless similar and include older age (9–11, 14–17), chronic lung or liver disease (14), depression (18), NYHA functional class IV (16), higher heart rate (6, 15,17), lower systolic blood pressure (6, 9–11, 14–17,19), lower LV ejection fraction (9–11, 14), lower serum sodium (6,11, 14–17), higher BUN or creatinine (2, 11, 14–17), lack of beta-blocker use (11, 20,21), and requirement of high diuretic dose (equal to 240 mg of furosemide) (11, 22,23).

While these studies and registries demonstrate the consistency of many of the important variables, they were limited in their inclusion of patients with advanced systolic dysfunction and severe decompensation. The ESCAPE trial captured this important population and characterized the clinical, laboratory, physiologic, and functional status of these patients.

Relative systemic hypotension as well as high BUN and high doses of loop diuretics are all markers of significantly reduced LV systolic function associated with hemodynamic compromise. Previous studies and registries have demonstrated the importance of renal dysfunction in predicting outcome (24,25). In particular, BUN may not only be a marker of advanced renal dysfunction but also may indicate the combined deterioration of cardiac and renal function. Similarly, a low systolic blood pressure may represent low forward output. Thus, it is not surprising that it has been shown to be associated with increased mortality consistently in all previous studies (6–12,14–17). Increased requirement for loop diuretics (ie, >240 mg) suggests an important population that is severely congested, resistant to diuretic therapy, and has an increased risk of events at follow-up (22,23). The presence of respiratory failure requiring mechanical ventilation and cardiac arrest during a HF hospitalization signifies a tenuous clinical profile associated with prolonged lengths of stay, concomitant increased morbidities, associated infections, cognitive impairment, and subsequently reduced clinical outcomes.

The finding that beta-blocker therapy influences outcome is not novel and has previously been demonstrated in the OPTIMIZE-HF (Organized Program to Initiate Lifesaving Treatment in Hospitalized Patients With Heart Failure) study, the IMPACT-HF (Initiation Management Predischarge: Process for Assessment of Carvedilol Therapy in Heart Failure) registry and trial, and ESCAPE (21,26,27), and is consistent with benefits of these agents shown in randomized clinical trials of advanced HF. Whether these agents improve survival or merely identify patients with less decompensation who can tolerate beta-blockade is not clear, but their use continues to be a favorable factor in the overall risk assessment.

The natural history of HF challenges all models of risk determination. HF patients typically experience prolonged periods of stability interrupted by acute events that result in volume overload or reduced perfusion with associated end-organ dysfunction. Further, disease progression is common over a variable period of time that is patient-specific. Thus, the risk of adverse events, hospitalization, and death fluctuates but tends to be highest during the period surrounding a hospitalization. Further, it should be anticipated that risks change during a HF hospitalization depending on response to therapy. To date, the ESCAPE Risk Model is the only strategy that addresses hospitalized patients with advanced HF and accounts for alterations in clinical status, physiology, and functional status in the prediction of short- and intermediate-term risk.

Clinical importance of risk stratification

Patients with advanced HF consume considerable resources and have a high risk of morbidity and mortality. Rapid advances in treatment strategies have resulted in several new therapeutic options available to clinicians managing these patients. However, the major limitations in ubiquitous use of these newer technologies include lack of widespread availability and prohibitive cost. The ability to identify the high-risk subgroups should allow better resource use by targeting the intensive strategies to those at particularly high-risk. However, this hypothesis remains to be tested because it is also possible that these patients may have advanced beyond the benefit of our current strategies. Prospective experience will be required to determine whether such a focused risk-based strategy of resource allocation (ie, implantable cardioverter defibrillators, LV assist devices, cardiac transplantation, cellular regeneration) is an effective use of limited resources to improve overall outcomes.

Study limitations

This analysis has several important limitations. The ESCAPE study population does not represent the general population of hospitalized HF patients, but rather those with severe LV dysfunction and advanced symptoms. The study population was comprised of patients at centers with expertise and extensive experience managing advanced HF. However, these are the pools from which most patients are currently selected for advanced management strategies and replacement therapies. Additionally, patients who had baseline characteristics shown to be associated with worse outcomes were excluded, such as those with severe renal dysfunction (creatinine >3.5 mg/dL) or those requiring high doses of inotropes during hospitalization. While our model was internally validated using bootstrap method, we were unable to find a dataset that had all the elements available in the ESCAPE dataset for external validation. As a result, as discussed above, we validated the ‘reduced’ ESCAPE model (without diuretic dose and BNP level) in the FIRST study data. We support further external validation in the future to further confirm the predictive accuracy and reliability of the ESCAPE model prior to broad adoption for clinical use.

Conclusions

This model and risk score in advanced decompensated HF patients, developed from easily determined clinical characteristics and hospital course characteristics, can be used in patients to identify patient cohorts with high, medium or low risk for death at 6 months. Thus, this risk score allows clinicians to intensify monitoring follow-up and therapeutic strategies. The mortality model provides a high degree of discriminatory power using the clinical variables. Prospective validation of the models in independent databases and using the model to triage risk, in both clinical practice and in clinical trials, should improve the care of these patients with advanced HF.

Abbreviations

ACE

angiotensin-converting nzyme

BNP

B-type natriuretic peptide

BUN

blood urea nitrogen

ESCAPE

Evaluation Study of Congestive Heart Failure and Pulmonary Artery Catheterization Effectiveness

FIRST

Flolan International Randomized Survival Trial

HF

heart failure

LV

left ventricular

NYHA

New York Heart Association

PAC

pulmonary artery catheter