Comparison of Mortality Prediction Models on Long-Term Mortality in Hospitalized Patients With Acute Heart Failure　― The Importance of Accounting for Nutritional Status ―

Abstract

Background: The ideal mortality prediction model (MPM) for acute heart failure (AHF) patients would have sufficient and stable predictive ability for long-term as well as short-term mortality. However, published MPMs for AHF predominantly predict short-term mortality up to 90 days, and their prognostic performance for long-term mortality remains unclear.

Methods and Results: We analyzed 609 AHF patients in a prospective registry from January 2013 to May 2016. We compared the prognostic performance for long-term mortality among 8 systematically identified MPMs for AHF that predict short-term mortality up to 90 days from admission. The PROTECT 7-day model showed the highest c-index for long-term as well as short-term mortality among the studied MPMs. Sensitivity analyses revealed serum albumin and total cholesterol to be the most important variables, as dropping these variables resulted in a significant decline in c-index, when compared with other variables specific to the PROTECT 7-day model. Furthermore, significant improvements in c-index and net reclassification were observed when serum albumin or serum albumin plus total cholesterol was added to the studied MPMs, other than the PROTECT 7-day model.

Conclusions: The PROTECT 7-day model demonstrated the highest predictive performance for long-term as well as short-term mortality in AHF patients among the published MPMs. Our findings indicate the importance of accounting for nutritional status such as serum albumin and total cholesterol in AHF patients when developing a MPM.

The increased incidence of chronic heart failure (HF) has resulted in an increase in acute HF (AHF) hospitalizations, accounting for over 1 million admissions annually in the USA and more than 200,000 admissions in Japan.^1,2 Despite advances in the management of HF, the mortality rate for patients hospitalized for AHF remains strikingly high,^1–3 and recent clinical trials have failed to improve the outcome for such patients.^4,5

Prognostication of patients with AHF could help to improve outcomes by identifying high-risk patients who might potentially benefit from intensive inpatient and outpatient monitoring and early referral for advanced HF therapies, including palliative care. A mortality prediction model (MPM) that can estimate an individual’s risk of death in AHF patients could be a useful tool for clinical decision-making at the time of admission. Recently published MPMs developed from AHF patients enrolled in surveys, registries, and randomized clinical trials have provided insights into significant patient factors that predict death.^6–16 Importantly, the ideal MPM for AHF patients would have sufficient and stable predictive ability for long-term as well as short-term mortality; however, most published MPMs demonstrate predictive ability only for specific time points (either predominantly short-term i.e., ≤90 days, or long-term i.e., ≥1 year). In addition, it can be hypothesized that if a specific model were found to be superior to other MPMs, several important prognostic factors could be identified and be imputed into future MPMs.

Accordingly, the aims of our study were to investigate and compare the predictive ability of published MPMs originally derived for short-term mortality prediction up to 90 days, for long-term mortality (180 and 360 days) in patients with AHF.

Methods

Data Source

Data from the NaDEF (National cerebral and cardiovascular center acute DEcompensated heart Failure) registry, which were obtained between January 2013 and May 2016, were retrospectively analyzed. The design of the NaDEF registry has been described previously.¹⁷ Briefly, it is a single-center, observational, prospective cohort that includes all consecutive patients aged >20 years requiring hospitalization for the first time with a diagnosis of AHF from January 2013. This study was approved by the Institutional Review Board of the National Cerebral and Cardiovascular Center (M22-025) and registered under the Japanese UMIN Clinical Trials Registration (UMIN000017024).

Study Population

A total of 850 consecutive patients with AHF were registered in NaDEF. Patients with acute coronary syndrome (n=38) and those without complete data for calculating risk scores/log-odds for mortality in the 8 studied MPMs (n=203) were excluded. A total of 609 patients were ultimately included in this study (Figure 1).

Figure 1.

Flow chart of the study of 8 published mortality prediction models for risk stratification of acute heart failure patients. ADHERE, Acute Decompensated Heart Failure National Registry; ASCEND-HF, Acute Study of Clinical Effectiveness of Nesiritide in Decompensated Heart Failure; EFFECT, Enhanced Feedback for Effective Cardiac Treatment; GWTG-HF, Get With the Guidelines-Heart Failure; OPTIME-CHF, Outcomes of a Prospective Trial of Intravenous Milrinone for Exacerbations of Chronic Heart Failure; OPTIMIZE-HF, Organized Program to Initiate Lifesaving Treatment in Hospitalized Patients with Heart Failure; PROTECT, Placebo-Controlled Randomized Study of the Selective A1-Adenosine Receptor Antagonist Rolofylline for Patients Hospitalized with Acute Decompensated Heart Failure and Volume Overload to Assess Treatment Effect on Congestion and Renal Function.

Mortality Prediction Models

We compared 8 published MPMs for AHF originally derived for short-term mortality prediction up to 90 days (Table 1, Supplementary Tables 1 – 7). We carried out a detailed search using the MEDLINE/PubMed and EMBASE search engines, and identified all the MPMs based on a specific search strategy (Supplementary Appendix). This search strategy has been previously validated with high sensitivity for finding prediction research in MEDLINE.¹⁸ Among the 28 models, 8 were chosen according to the following criteria: (1) all variables imputed into the model were available in our datasets, (2) derived for short-term mortality prediction up to 90 days after admission, and (3) calculable risk scores were available. The models identified were: (1) The Placebo-Controlled Randomized Study of the Selective A1-Adenosine Receptor Antagonist Rolofylline for Patients Hospitalized with Acute Decompensated Heart Failure and Volume Overload to Assess Treatment Effect on Congestion and Renal Function (PROTECT) risk score for 7-day mortality,⁶ (2) The Get With The Guidelines-Heart Failure (GWTG-HF) risk score for in-hospital mortality,⁷ (3) The Organized Program to Initiate Lifesaving Treatment in Hospitalized Patients with Heart Failure (OPTIMIZE-HF) risk score for in-hospital mortality,⁸ (4) The Acute Decompensated Heart Failure National Registry (ADHERE) model for in-hospital mortality,¹⁰ (5) The Enhanced Feedback for Effective Cardiac Treatment (EFFECT) risk score for 30-day mortality,¹⁴ (6) The Acute Study of Clinical Effectiveness of Nesiritide in Decompensated Heart Failure (ASCEND-HF) risk score for 30-day mortality,¹⁶ (7) The Outcomes of a Prospective Trial of Intravenous Milrinone for Exacerbations of Chronic Heart Failure (OPTIME-CHF) risk score,¹⁵ and (8) The OPTIMIZE-HF risk score post-discharge for up to 90-day mortality.⁹

Table 1. Published Mortality Prediction Models and the Imputed Variables

	PROTECT	GWTG-HF	OPTIMIZE-HF	ADHERE	EFFECT	ASCEND-HF	OPTIME-CHF	OPTIMIZE-HF
Reference	(6)	(7)	(8)	(10)	(14)	(16)	(15)	(9)
Type of study	RCT	Registry	Registry	Registry	AC	RCT	RCT	Registry
No. of patients	2,015	27,850	37,548	33,046	2,624	7,141	949	4,402
No. of hospitals	173	287	259	263	34	398	80	91
Country	International	USA	USA	USA	Canada	International	USA	USA
Mortality	7-day	In-hospital	In-hospital	In-hospital	30-day	30-day	60-day	60~90-day
Variables
Age		✓	✓	✓	✓	✓	✓	✓
Body weight								✓
Systolic BP	✓	✓	✓	✓	✓	✓	✓	✓
Sodium		✓	✓		✓	✓	✓	✓
Blood urea nitrogen	✓	✓		✓	✓	✓	✓
Creatinine			✓					✓
ALB	✓
TC	✓
Heart rate	✓	✓	✓	✓
Respiratory rate	✓				✓
NYHA functional class						✓	✓
LVEF			✓
Black race		✓
COPD		✓			✓
Reactive airway disease								✓
Prior HF admission	✓
Primary cause for admission (HF/other)			✓
Cerebrovascular accident					✓
Cancer					✓
Dementia					✓
Depression								✓
Liver disease					✓			✓
Diabetes	✓

AC, administrative claim; ADHERE, Acute Decompensated Heart Failure National Registry; ALB, albumin; ASCEND-HF, Acute Study of Clinical Effectiveness of Nesiritide in Decompensated Heart Failure; BP, blood pressure; COPD, chronic obstructive pulmonary disease; EFFECT, Enhanced Feedback for Effective Cardiac Treatment; GWTG-HF, Get With the Guidelines-Heart Failure; HF, heart failure; LVEF, left ventricular ejection fraction; NYHA, New York Heart Association; OPTIME-CHF, Outcomes of a Prospective Trial of Intravenous Milrinone for Exacerbations of Chronic Heart Failure; OPTIMIZE-HF, Organized Program to Initiate Lifesaving Treatment in Hospitalized Patients with Heart Failure; PROTECT, Placebo-Controlled Randomized Study of the Selective A1-Adenosine Receptor Antagonist Rolofylline for Patients Hospitalized with Acute Decompensated Heart Failure and Volume. Overload to Assess Treatment Effect on Congestion and Renal Function; RCT, randomized clinical trial; TC, total cholesterol.

In the ADHERE model, we used the coefficients of blood urea nitrogen (BUN), systolic BP, heart rate (HR) and age from their multivariable model to calculate the log-odds of in-hospital death (0.0212×BUN−0.0192×systolic BP+0.0131×HR+0.0288×age−4.72), as shown in the original paper.¹⁰

Statistical Analysis

Continuous variables are presented as mean±SD when normally distributed, and as median and interquartile range (IQR) when non-normally distributed. After defining predicted mortality for each model using the risk scores or the log-odds of mortality in the Cox regression model, we assessed the predictive performance for each model. Model predictive ability was assessed by calculating Harrell’s c-index using bootstrap resampling (200 times). To identify important prognostic factors, which were specifically imputed into the best prognostic model among the studied MPMs, sensitivity analyses were performed by dropping each specific factor from the model. Thereafter, we added these prognostic parameters or the Controlling Nutritional Status (CONUT) score (Supplementary Table 8)¹⁹ to each model other than the best prediction model, and compared the c-index and the category-free net reclassification improvement described by Penicina et al,²⁰ for 180-day and 360-day mortality. All tests were two-tailed, and P<0.05 was considered statistically significant. All analyses were performed with Stata MP64 version 15 (StataCorp, College Station, TX, USA).

Results

Baseline Characteristics

The baseline clinical characteristics of our cohort are shown in Table 2. Mean age was 74.7 years, 61% were male, and 52% had HF with reduced ejection fraction. Mean systolic BP, HR and respiratory rate were 139.6 mmHg, 93 beats/min and 22.4 breaths/min, respectively. Almost half of the patients had a history of HF hospitalization, the prevalence of ischemic heart disease was 24%, and that of cerebrovascular accident was 24%. Median BUN, serum creatinine and B-type natriuretic peptide levels on admission were 24 mg/dL, 1.1 mg/dL and 619pg/mL, respectively.

Table 2. Baseline Characteristics of AHF Patients

Variable	Study population (n=609)
Age, years	74.7±12.4
Male sex, n (%)	373 (61)
Body mass index	22.8 (20.3~25.6)
NYHA III or IV, n (%)	545 (89)
Systolic BP, mmHg	139.6±31.1
Heart rate, beats/min	93.0±28.8
Respiratory rate, breaths/min	22.4±6.5
LVEF, %	37 (23~55)
LVEF ≤40%, n (%)	318 (52)
Comorbidities, n (%)
Prior HF admission	273 (45)
Ischemic heart disease	145 (24)
Diabetes mellitus	225 (37)
Hypertension	438 (72)
Depression	17 (3)
Dementia	18 (3)
COPD/asthma	56 (9)
Liver disease	7 (1)
Chronic kidney disease	314 (52)
Cancer	81 (13)
Cerebrovascular accident	147 (24)
Laboratory data
Hemoglobin, g/dL	12.0±2.2
Sodium, mEq/L	139.6±4.2
ALB, mg/dL	3.8 (3.5~4.1)
TC, mg/dL	153 (131~179)
Blood urea nitrogen, mg/dL	24 (17~33)
Lymphocyte count, /mL	1,102 (743~1,606)
CONUT score	3 (2~5)
Creatinine, mg/dL	1.1 (0.9~1.6)
BNP, pg/mL	619 (327~1,156)
Medications on admission, n (%)
ACEIs or ARBs	312 (51)
β-blockers	298 (49)
MRAs	152 (25)
Loop diuretics	328 (54)
Digitalis	82 (14)
Statins	213 (35)
Medications at discharge, n (%)
ACEIs or ARBs	424 (71)
β-blockers	427 (71)
MRAs	279 (47)
Loop diuretics	488 (81)
Digitalis	77 (13)
Statins	264 (44)

Continuous variables are presented as mean±SD if normally distributed, and median (interquartile range) if not normally distributed. Categorical variables are presented as number of patients (%). ACEI, angiotensin converting enzyme; AHF, acute HF; ARB, angiotensin receptor blocker; BNP, B-type natriuretic peptide; CONUT, Controlling Nutritional Status; COPD, chronic obstructive pulmonary disease; MRA, mineralocorticoid-receptor antagonist. Other abbreviations as in Table 1.

Comparison of Discriminative Ability Among Published MPMs

Among the study population of 609 patients, the number of deaths at 30 days after admission was 7 (1.2%), 22 at 90 days (3.6%), 34 at 180 days (5.6%) and 66 at 360 days (10.8%). The PROTECT model had the best and most stable predictive ability, with a c-index of 0.73 or higher compared with the other 7 MPMs at each follow-up time point after admission; the GWTG-HF and ADHERE models each demonstrated a lower c-index of 0.66 (Figure 2).

Figure 2.

Predictive ability of 8 published mortality prediction models for acute heart failure patients. Abbreviations as in Figure 1.

Sensitivity Analysis for Identifying Important Prognostic Factors in the PROTECT Model

Sensitivity analyses were performed by removing each specific variable, including serum albumin, total cholesterol, history of HF admission and diabetes, which were specifically imputed in the PROTECT model compared with the other MPMs (Table 1). Of these 4 variables, removing serum albumin or total cholesterol from the model resulted in a significant decline in c-index at 180 and 360 days after admission (Figure 3).

Figure 3.

Sensitivity analysis for PROTECT 7-day mortality prediction model. HF, heart failure.

Improvement of Predictive Performance by Adding Serum Albumin and Cholesterol and Nutritional Score to MPMs Other Than the PROTECT Model

According to the sensitivity analyses, we found that nutritional markers such as serum albumin and total cholesterol were important for long-term mortality prediction. Consequently, in addition to these markers, we examined the CONUT score, which contains serum albumin and total cholesterol as major components and is useful for assessing the nutritional status in AHF patients. Table 3 and Table 4 show comparisons of the models for 180-day and 360-day mortality; one being the original score/log-odds for mortality from each MPM other than the PROTECT model, and the other having the addition of serum albumin, total cholesterol, serum albumin plus total cholesterol and the CONUT score to each MPM. The c-indices were significantly increased, and the net reclassification improvements were significant in most MPMs other than the PROTECT model when adding serum albumin, serum albumin plus total cholesterol, and the CONUT score, especially for 360-day mortality.

Table 3. Significance of Adding Nutritional Markers/Score to Mortality Prediction Models Other Than PROTECT 7-Day Model for Long-Term Mortality Prediction

Models	Original	+ALB	P value	+ TC	P value	+ALB+TC	P value	+CONUT	P value
180-day mortality: C-index (95% CI)
GWTG-HF in-hospital	0.69 (0.60~0.77)	0.76 (0.60~0.78)	0.019	0.73 (0.65~0.81)	0.102	0.78 (0.70~0.86)	0.008	0.76 (0.68~0.84)	0.037
OPTIMIZE-HF in-hospital	0.70 (0.61~0.78)	0.75 (0.66~0.85)	0.017	0.73 (0.64~0.82)	0.142	0.77 (0.68~0.86)	0.004	0.76 (0.67~0.85)	0.029
ADHERE in-hospital	0.69 (0.60~0.78)	0.76 (0.67~0.85)	0.031	0.73 (0.65~0.82)	0.132	0.78 (0.70~0.86)	0.015	0.76 (0.68~0.84)	0.062
EFFECT 30-day	0.72 (0.65~0.80)	0.77 (0.69~0.85)	0.069	0.74 (0.66~0.83)	0.327	0.79 (0.71~0.86)	0.024	0.77 (0.70~0.85)	0.103
ASCEND 30-day	0.70 (0.63~0.78)	0.76 (0.68~0.85)	0.039	0.73 (0.65~0.82)	0.305	0.77 (0.70~0.85)	0.028	0.76 (0.68~0.84)	0.061
OPTIME-CHF 60-day	0.74 (0.67~0.82)	0.79 (0.71~0.87)	0.075	0.77 (0.69~0.85)	0.162	0.80 (0.73~0.88)	0.047	0.79 (0.72~0.86)	0.077
OPTIMIZE-HF 90-day	0.69 (0.62~0.77)	0.74 (0.65~0.83)	0.077	0.72 (0.63~0.80)	0.372	0.76 (0.68~0.84)	0.028	0.74 (0.66~0.83)	0.155
360-day mortality: C-index (95% CI)
GWTG-HF in-hospital	0.66 (0.60~0.73)	0.73 (0.67~0.79)	<0.001	0.70 (0.64~0.76)	0.046	0.74 (0.68~0.80)	0.002	0.73 (0.67~0.79)	0.005
OPTIMIZE-HF in-hospital	0.69 (0.63~0.75)	0.74 (0.67~0.80)	0.012	0.71 (0.65~0.77)	0.115	0.75 (0.69~0.81)	0.003	0.74 (0.68~0.80)	0.019
ADHERE in-hospital	0.66 (0.60~0.73)	0.73 (0.67~0.79)	0.005	0.69 (0.63~0.76)	0.089	0.74 (0.68~0.80)	0.002	0.73 (0.66~0.79)	0.011
EFFECT 30-day	0.73 (0.68~0.79)	0.77 (0.71~0.83)	0.014	0.75 (0.69~0.80)	0.196	0.78 (0.72~0.83)	0.006	0.77 (0.71~0.82)	0.035
ASCEND 30-day	0.71 (0.66~0.77)	0.75 (0.70~0.81)	0.021	0.73 (0.67~0.79)	0.174	0.76 (0.71~0.82)	0.011	0.75 (0.70~0.81)	0.032
OPTIME-CHF 60-day	0.73 (0.67~0.78)	0.77 (0.71~0.82)	0.028	0.74 (0.68~0.80)	0.267	0.77 (0.72~0.83)	0.021	0.76 (0.70~0.82)	0.068
OPTIMIZE-HF 90-day	0.71 (0.65~0.76)	0.75 (0.69~0.81)	0.020	0.72 (0.67~0.78)	0.239	0.76 (0.70~0.81)	0.011	0.74 (0.68~0.80)	0.081

All P values vs. original model. CI, confidence interval; CONUT, Controlling Nutritional Status. Other abbreviations as in Table 1.

Table 4. NRI of Adding Nutritional Markers/Score to Mortality Prediction Models Other Than PROTECT 7-Day Model for Long-Term Mortality Prediction

Models	+ALB	P value	+TC	P value	+ALB+TC	P value	+CONUT	P value
180-day mortality: NRI (95% CI), %
GWTG-HF in-hospital	56 (21~90)	0.002	39 (5~74)	0.026	45 (10~79)	0.012	34 (−1~68)	0.057
OPTIMIZE-HF in-hospital	53 (19~88)	0.003	36 (1~70)	0.044	46 (12~81)	0.009	36 (1~70)	0.044
ADHERE in-hospital	58 (23~92)	0.001	39 (4~74)	0.027	45 (10~79)	0.012	30 (−4~65)	0.085
EFFECT 30-day	48 (13~82)	0.007	32 (−3~66)	0.073	47 (12~82)	0.008	36 (2~71)	0.040
ASCEND 30-day	51 (16~85)	0.004	39 (5~74)	0.026	44 (10~79)	0.012	51 (16~85)	0.004
OPTIME-CHF 60-day	50 (16~85)	0.005	29 (−5~64)	0.097	48 (13~83)	0.007	56 (21~90)	0.002
OPTIMIZE-HF 90-day	55 (20~90)	0.002	40 (5~75)	0.023	48 (13~83)	0.007	32 (2~67)	0.068
360-day mortality: NRI (95% CI), %
GWTG-HF in-hospital	38 (12~64)	0.004	28 (2~54)	0.032	41 (16~67)	0.002	50 (24~75)	<0.001
OPTIMIZE-HF in-hospital	38 (13~64)	0.003	25 (−0.1~51)	0.051	42 (17~68)	0.001	46 (20~71)	<0.001
ADHERE in-hospital	41 (15~66)	0.002	25 (−1~51)	0.055	41 (16~67)	0.002	48 (23~74)	<0.001
EFFECT 30-day	42 (16~67)	0.001	20 (−6~46)	0.123	42 (16~67)	0.001	40 (14~65)	0.002
ASCEND 30-day	42 (17~68)	0.001	23 (−2~49)	0.076	41 (16~67)	0.002	53 (27~78)	<0.001
OPTIME-CHF 60-day	46 (20~71)	<0.001	19 (−7~44)	0.153	43 (17~68)	0.001	52 (26~78)	<0.001
OPTIMIZE-HF 90-day	45 (19~70)	<0.001	23 (−2~49)	0.076	40 (14~66)	0.002	47 (22~73)	<0.001

NRI, net reclassification improvement. Other abbreviations as in Tables 1,3.

Discussion

The present findings demonstrated that the PROTECT risk score for 7 days was the best and most stable among the 8 studied MPM for long-term as well as short-term mortality after admission of patients with AHF. We also identified serum albumin and total cholesterol, which are nutrition-related markers, as important prognostic variables compared with the other 2 specific variables imputed in the PROTECT model based on our sensitivity analyses. We further highlighted that the addition of nutritional markers/score including serum albumin, serum albumin plus cholesterol, and the CONUT score to each MPM other than the PROTECT model significantly increased their predictive ability for long-term mortality in AHF patients, which was shown not only by comparison of c-indices but also by net reclassification improvement. These findings suggested that nutrition-related parameters may need to be considered in developing an MPM that is more accurate and stable for long-term as well as short-term mortality of patients with AHF.

AHF is a life-threatening condition and both short-term and subsequent mortality rates are still high. Multiple MPMs for short-term mortality have been proposed, and improving their predictive performance for patients hospitalized with AHF has important clinical and research implications. For instance, identification of high-risk patients would enable effective triage (e.g., transfer from a small county hospital to a tertiary referral hospital with a cardiac intensive care facility, earlier referral for advanced HF therapy and palliative care etc.). Additionally, risk stratification would help with designing clinical trials where newer therapeutic modalities are tested across patients with varying prognoses (e.g., testing drugs in patients with high risk vs. low risk for in-hospital death). In order to improve the predictive performance of MPMs, it could be a good methodology to find the model that has the best and stable predictive performance for not only short-term but also long-term mortality and to identify the reasons why the best model is superior to other MPMs. In the present study, we found the PROTECT model had higher predictive performance for short- and long-term mortality in AHF patients when compared with 7 other MPMs, and furthermore, nutritional markers such as serum albumin and total cholesterol were important for the higher c-index for long-term mortality in the PROTECT model.

Malnutrition is associated with worse clinical outcomes, so current guidelines recognize nutritional management as an important therapeutic approach in HF patients.^21,22 Indeed, advanced HF patients with a high risk of death are characterized by decreased cardiac output and subsequent hypoperfusion in peripheral tissue, with enhanced inflammatory and neurohormonal activities such as pro-inflammatory cytokine production, tumor necrosis factor-α/nuclear factor-κ B signaling pathway, renin-angiotensin-aldosterone system, and sympathetic nervous system, as systemic compensatory mechanisms.^23–26 These adverse reactions affect immunity, insulin resistance, and the anabolic-catabolic balance, resulting in impaired protein and lipid metabolism, which ultimately leads to malnutrition and cachexia.^27,28 Furthermore, these high-risk patients exhibit significant malabsorption because of gut edema, increased energy expenditure with dyspnea and the increased effort of breathing, and anorexia caused by gut and hepatic congestion.²⁴

Importantly, these mechanisms are accelerated in AHF patients, resulting in rapid progression of malnutrition and further exacerbation of the HF syndrome. Emerging evidence has elucidated the consistent usefulness of several nutritional parameters, including serum albumin and total cholesterol, as determinants of clinical outcomes for such patients.^29–31 For instance, hypoalbuminemia on admission, which is mainly caused by severe congestion, inflammation and malnutrition, is related to impaired oncotic pressure and is strongly associated with short-term as well as long-term mortality.^30,32–34 We have also reported the importance of nutritional assessment on admission based on the geriatric nutritional risk index (GNRI) and the CONUT score, which include serum albumin and total cholesterol as major components for stratifying the subsequent long-term risk of mortality in patients with AHF.^30,35,36

Taking all this evidence together, nutrition-related parameters on admission would be useful for predicting short- and long-term mortality of AHF patients, and accounting for them could lead to the development of more accurate MPMs. Our present findings indicated that adding nutritional markers/score, including serum albumin, serum albumin plus cholesterol, and the CONUT score, to each MPM that does not contain these nutritional markers significantly increases the model’s predictive ability for long-term mortality in AHF patients. To establish more accurate MPMs that simultaneously predict short- and long-term mortality in real-world AHF patients, well-designed and updated AHF registries with a large sample size and which have variables of nutritional markers, including at least serum albumin and total cholesterol, are warranted.

Study Limitations

First, it was conducted in a single-center registry and had a small sample size and number of events, thereby limiting the ability to generalize the findings and the statistical power for detecting differences in negative data, and external validity should be examined by expanding the study sites. In the similar context, the differences in c-indices between the PROTECT original model and the model without albumin or without total cholesterol were actually small (around 0.025) as a consequence of the small sample size, although the differences were statistically significant; therefore, confirming our findings in larger, multicenter cohorts is necessary for strengthening clinical relevance. Second, we could not confirm the validity of our findings in AHF patients without prior HF hospitalization (i.e., de novo AHF) because of the limited sample size and events (among 336 patients without prior HF hospitalization, the number of deaths at 30 days after admission was 2; at 90 days was 10; at 180 days was 15 and at 360 days was 24). Further study with a larger number of patients with a first AHF hospitalization is warranted. Finally, we could evaluate the additional value of nutritional parameters and the CONUT score to the studied MPMs other than the PROTECT model for outcome prediction only at 180 and 360 days because of the limited number of events. However, we believe these limitations are outweighed by the strength of the study design, showing robustness of the importance of accounting for nutrition-related markers when developing more accurate MPMs for long-term mortality in AHF patients, and the novelty of the research idea.

Conclusions

The PROTECT 7-day model demonstrated the highest predictive performance for long-term mortality in AHF patients among 8 published MPMs derived for short-term mortality prediction. Serum albumin and total cholesterol may contribute to the accuracy and stability of the PROTECT 7-day model for long-term outcome prediction, and the addition of these markers improved most MPMs other than the PROTECT model. These findings indicate the importance of accounting for the nutritional status of AHF patients when developing a MPM.

Acknowledgments

The authors are grateful for the contributions of all the investigators, clinical research coordinators, data managers, and laboratory technicians involved in the NaDEF registry.

Funding

This work was supported by a grant from the Japan Cardiovascular Research Foundation (T. Anzai, 24-4-2), and a grant-in-aid for Young Scientists from the Japan Society for the Promotion of Science (T. Nagai, 15K19402).

Conflict of Interest

None.

Supplementary Files

Please find supplementary file(s);

http://dx.doi.org/10.1253/circj.CJ-18-1243

References

• 1.   Munir MB, Sharbaugh MS, Thoma FW, Nisar MU, Kamran AS, Althouse AD, et al. Trends in hospitalization for congestive heart failure, 1996–2009. Clin Cardiol 2017; 40: 109–119.
• 2.   Yasuda S, Nakao K, Nishimura K, Miyamoto Y, Sumita Y, Shishido T, et al. The current status of cardiovascular medicine in Japan: Analysis of a large number of health records from a nationwide claim-based database, JROAD-DPC. Circ J 2016; 80: 2327–2335.
• 3.   Chioncel O, Mebazaa A, Harjola VP, Coats AJ, Piepoli MF, Crespo-Leiro MG, et al. Clinical phenotypes and outcome of patients hospitalized for acute heart failure: The ESC Heart Failure Long-Term Registry. Eur J Heart Fail 2017; 19: 1242–1254.
• 4.   O’Connor CM, Starling RC, Hernandez AF, Armstrong PW, Dickstein K, Hasselblad V, et al. Effect of nesiritide in patients with acute decompensated heart failure. N Engl J Med 2011; 365: 32–43.
• 5.   Packer M, O’Connor C, McMurray JJV, Wittes J, Abraham WT, Anker SD, et al. Effect of ularitide on cardiovascular mortality in acute heart failure. N Engl J Med 2017; 376: 1956–1964.
• 6.   O’Connor CM, Mentz RJ, Cotter G, Metra M, Cleland JG, Davison BA, et al. The PROTECT in-hospital risk model: 7-day outcome in patients hospitalized with acute heart failure and renal dysfunction. Eur J Heart Fail 2012; 14: 605–612.
• 7.   Peterson PN, Rumsfeld JS, Liang L, Albert NM, Hernandez AF, Peterson ED, et al. A validated risk score for in-hospital mortality in patients with heart failure from the American Heart Association Get with the Guidelines program. Circ Cardiovasc Qual Outcomes 2010; 3: 25–32.
• 8.   Abraham WT, Fonarow GC, Albert NM, Stough WG, Gheorghiade M, Greenberg BH, et al. Predictors of in-hospital mortality in patients hospitalized for heart failure: Insights from the Organized Program to Initiate Lifesaving Treatment in Hospitalized Patients with Heart Failure (OPTIMIZE-HF). J Am Coll Cardiol 2008; 52: 347–356.
• 9.   O’Connor CM, Abraham WT, Albert NM, Clare R, Gattis Stough W, Gheorghiade M, et al. Predictors of mortality after discharge in patients hospitalized with heart failure: An analysis from the Organized Program to Initiate Lifesaving Treatment in Hospitalized Patients with Heart Failure (OPTIMIZE-HF). Am Heart J 2008; 156: 662–673.
• 10.   Abraham WT, Adams KF, Fonarow GC, Costanzo MR, Berkowitz RL, LeJemtel TH, et al. In-hospital mortality in patients with acute decompensated heart failure requiring intravenous vasoactive medications: An analysis from the Acute Decompensated Heart Failure National Registry (ADHERE). J Am Coll Cardiol 2005; 46: 57–64.
• 11.   Cleland JGF, Teerlink JR, Davison BA, Shoaib A, Metra M, Senger S, et al. Measurement of troponin and natriuretic peptides shortly after admission in patients with heart failure: Does it add useful prognostic information? An analysis of the Value of Endothelin Receptor Inhibition with Tezosentan in Acute heart failure Studies (VERITAS). Eur J Heart Fail 2017; 19: 739–747.
• 12.   Lee DS, Stitt A, Austin PC, Stukel TA, Schull MJ, Chong A, et al. Prediction of heart failure mortality in emergent care: A cohort study. Ann Intern Med 2012; 156: 767–775, W-261, W-262.
• 13.   Cleland JG, Chiswell K, Teerlink JR, Stevens S, Fiuzat M, Givertz MM, et al. Predictors of postdischarge outcomes from information acquired shortly after admission for acute heart failure: A report from the Placebo-Controlled Randomized Study of the Selective A1 Adenosine Receptor Antagonist Rolofylline for Patients Hospitalized With Acute Decompensated Heart Failure and Volume Overload to Assess Treatment Effect on Congestion and Renal Function (PROTECT) Study. Circ Heart Fail 2014; 7: 76–87.
• 14.   Lee DS, Austin PC, Rouleau JL, Liu PP, Naimark D, Tu JV. Predicting mortality among patients hospitalized for heart failure: Derivation and validation of a clinical model. JAMA 2003; 290: 2581–2587.
• 15.   Felker GM, Leimberger JD, Califf RM, Cuffe MS, Massie BM, Adams KF Jr, et al. Risk stratification after hospitalization for decompensated heart failure. J Card Fail 2004; 10: 460–466.
• 16.   Khazanie P, Heizer GM, Hasselblad V, Armstrong PW, Califf RM, Ezekowitz J, et al. Predictors of clinical outcomes in acute decompensated heart failure: Acute Study of Clinical Effectiveness of Nesiritide in Decompensated Heart Failure outcome models. Am Heart J 2015; 170: 290–297.
• 17.   Nagai T, Nishimura K, Honma T, Higashiyama A, Sugano Y, Nakai M, et al. Prognostic significance of endogenous erythropoietin in long-term outcome of patients with acute decompensated heart failure. Eur J Heart Fail 2016; 18: 803–813.
• 18.   Geersing GJ, Bouwmeester W, Zuithoff P, Spijker R, Leeflang M, Moons KG. Search filters for finding prognostic and diagnostic prediction studies in Medline to enhance systematic reviews. PLoS One 2012; 7: e32844.
• 19.   Ignacio de Ulibarri J, Gonzalez-Madrono A, de Villar NG, Gonzalez P, Gonzalez B, Mancha A, et al. CONUT: A tool for controlling nutritional status. First validation in a hospital population. Nutr Hosp 2005; 20: 38–45.
• 20.   Pencina MJ, D’Agostino RB Sr, Steyerberg EW. Extensions of net reclassification improvement calculations to measure usefulness of new biomarkers. Stat Med 2011; 30: 11–21.
• 21.   Ponikowski P, Voors AA, Anker SD, Bueno H, Cleland JG, Coats AJ, et al. 2016 ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure: The Task Force for the diagnosis and treatment of acute and chronic heart failure of the European Society of Cardiology (ESC). Developed with the special contribution of the Heart Failure Association (HFA) of the ESC. Eur J Heart Fail 2016; 18: 891–975.
• 22.   Yancy CW, Jessup M, Bozkurt B, Butler J, Casey DE Jr, Drazner MH, et al. 2013 ACCF/AHA guideline for the management of heart failure: Executive summary: A report of the American College of Cardiology Foundation/American Heart Association Task Force on practice guidelines. Circulation 2013; 128: 1810–1852.
• 23.   Levine B, Kalman J, Mayer L, Fillit HM, Packer M. Elevated circulating levels of tumor necrosis factor in severe chronic heart failure. N Engl J Med 1990; 323: 236–241.
• 24.   Rahman A, Jafry S, Jeejeebhoy K, Nagpal AD, Pisani B, Agarwala R. Malnutrition and cachexia in heart failure. J Parenter Enteral Nutr 2016; 40: 475–486.
• 25.   Tsutamoto T, Hisanaga T, Wada A, Maeda K, Ohnishi M, Fukai D, et al. Interleukin-6 spillover in the peripheral circulation increases with the severity of heart failure, and the high plasma level of interleukin-6 is an important prognostic predictor in patients with congestive heart failure. J Am Coll Cardiol 1998; 31: 391–398.
• 26.   Anker SD, Egerer KR, Volk HD, Kox WJ, Poole-Wilson PA, Coats AJ. Elevated soluble CD14 receptors and altered cytokines in chronic heart failure. Am J Cardiol 1997; 79: 1426–1430.
• 27.   Heck PM, Dutka DP. Insulin resistance and heart failure. Curr Heart Fail Rep 2009; 6: 89–94.
• 28.   Anker SD, Chua TP, Ponikowski P, Harrington D, Swan JW, Kox WJ, et al. Hormonal changes and catabolic/anabolic imbalance in chronic heart failure and their importance for cardiac cachexia. Circulation 1997; 96: 526–534.
• 29.   Uthamalingam S, Kandala J, Daley M, Patvardhan E, Capodilupo R, Moore SA, et al. Serum albumin and mortality in acutely decompensated heart failure. Am Heart J 2010; 160: 1149–1155.
• 30.   Bonilla-Palomas JL, Gamez-Lopez AL, Moreno-Conde M, Lopez-Ibanez MC, Anguita-Sanchez M, Gallego de la Sacristana A, et al. Hypoalbuminemia in acute heart failure patients: Causes and its impact on hospital and long-term mortality. J Card Fail 2014; 20: 350–358.
• 31.   Horwich TB, Hernandez AF, Dai D, Yancy CW, Fonarow GC. Cholesterol levels and in-hospital mortality in patients with acute decompensated heart failure. Am Heart J 2008; 156: 1170–1176.
• 32.   Nohria A, Tsang SW, Fang JC, Lewis EF, Jarcho JA, Mudge GH, et al. Clinical assessment identifies hemodynamic profiles that predict outcomes in patients admitted with heart failure. J Am Coll Cardiol 2003; 41: 1797–1804.
• 33.   Anand IS, Latini R, Florea VG, Kuskowski MA, Rector T, Masson S, et al. C-reactive protein in heart failure: Prognostic value and the effect of valsartan. Circulation 2005; 112: 1428–1434.
• 34.   Bonilla-Palomas JL, Gamez-Lopez AL, Anguita-Sanchez MP, Castillo-Dominguez JC, Garcia-Fuertes D, Crespin-Crespin M, et al. Impact of malnutrition on long-term mortality in hospitalized patients with heart failure. Rev Esp Cardiol 2011; 64: 752–758 (in Spanish).
• 35.   Honda Y, Nagai T, Iwakami N, Sugano Y, Honda S, Okada A, et al. Usefulness of Geriatric Nutritional Risk Index for assessing nutritional status and its prognostic impact in patients aged ≥65 years with acute heart failure. Am J Cardiol 2016; 118: 550–555.
• 36.   Iwakami N, Nagai T, Furukawa TA, Sugano Y, Honda S, Okada A, et al. Prognostic value of malnutrition assessed by Controlling Nutritional Status score for long-term mortality in patients with acute heart failure. Int J Cardiol 2017; 230: 529–536.