In this observational study, 5,919 heart failure patients, who were hospitalized in Sir Run Run Shaw Hospital and its medical consortium hospitals from January 2009 to April 2019, were recruited. Patients with the following criteria were excluded: (1) NRS-2002 score was not evaluated or documented at admission; (2) missing data on baseline characteristics, laboratory measurements, or past medical history; (3) severe hepatic or renal dysfunction, active malignant tumor, or critical autoimmune disease; (4) pregnant or lactating women during hospitalization or follow-up. Finally, a total of 2,830 patients were enrolled. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting standards were followed (18). Ethical approval was obtained from the ethics committee of Sir Run Run Shaw Hospital (NO.20201217-36).

The estimated number of participants was based on the incidence of re-hospitalization in the previous study and on the principle of 10 outcome events per variable (19). A previous study reported that ~20% patients with heart failure would be re-admitted within 1-year (1). Therefore, 20% was adopted as the estimated incidence of re-hospitalization in patients with heart failure. According to the estimated incidence of re-hospitalization and nine variables included in the log-binomial regression model, at least 450 heart failure patients were needed, which was far less than actual enrollment.

According to the 2016 European Society of Cardiology (ESC), heart failure is defined as an inability of the heart to meet the humans' metabolic demands, accompanied with typical symptoms and signs. The typical symptoms of heart failure mainly include dyspnea, weakness, chest tightness, and cardiogenic shock. The typical signs of heart failure involve lower limb edema, elevated jugular venous pressure, and pulmonary oedema (20). According to the level of left ventricular ejection fraction (LVEF), heart failure is classified into three types: heart failure with reduced ejection fraction (HFrEF) (<40%), heart failure with mid-range ejection fraction (HFmrEF) (40–49%), and heart failure with preserved ejection fraction (HFpEF) (≥50%).

Based on the European Society for Clinical Nutrition and Metabolism (ESPEN), the detailed calculation method of NRS-2002 was listed in Supplementary Table 1 (21). NRS-2002 score is calculated according to patients' impaired nutritional status, the severity of disease and age. The impaired nutritional status and severity of disease are scored between 0 and 3 points, respectively. Patients aged 70 years or older would receive an additional point based on the total score. Age-adjusted total NRS-2002 score ranges on a scale of 0 to 7. A total NRS-2002 score ≥3 points is considered at high nutritional risk, while those with a score below 3 points are considered at low nutritional risk (22).

Patients who were first hospitalized for heart failure were enrolled in this study and were followed up for 1-year, regardless of the classification of heart failure. The primary endpoint was re-hospitalization for heart failure during a 1-year follow-up after discharge. And the secondary endpoint was the length of initial hospital stay, which was defined as the length of stay in patients who were hospitalized for heart failure for the first time and was calculated by subtracting the date of admission from the date of discharge.

Baseline characteristics and clinical parameters related to this study were all derived from the Hospital Information System (HIS). Demographic data, laboratory data, comorbidities, and medication were collected. Fasting venous blood samples were collected on the morning after hospital admission day and then immediately sent to hematological and biochemical laboratory examinations. LVEF was assessed according to the classical Teichholz method. N-terminal pro-B-type natriuretic peptide (NT-proBNP) was analyzed using the NT-proBNP fold-elevation (actual value divided by the upper limit of normal value according to the age stratification). Nutritional status was assessed routinely within 24 h after hospital admission by the NRS-2002, which was applied by the trained nursing staff.

Continuous variables were presented as mean ± standard deviation (SD) if normally distributed, and median (interquartile range) if not. Categorical variables were expressed as numbers and proportions. Comparisons between high and low nutritional risk group were made using Student t-test if normally distributed or Mann–Whitney U test if not. The chi-square test or Fisher's exact test were used to examine the comparability of baseline characteristics for categorical variables according to minimal expected cell value.

The association between nutritional status (low nutritional risk and high nutritional risk) and 1-year re-hospitalization was evaluated using log-binomial regression models. Relative risks (RRs) with 95% CI were calculated. Covariates with potential significance for the prognosis of heart failure were adjusted in adjusted model 1, such as age (<65 or ≥65 years) (23), sex (male or female) (23), diabetes (yes or no) (24), hypertension (yes or no) (25), and estimated glomerular filtration rate (eGFR) (<90 or ≥90 ml/min/1.73 m²) (26). Adjusted model 2 further adjusted LVEF (<40, 40–49, or ≥50%) (27), NT-proBNP fold-elevation (<2 or ≥2) (28) and admission of diuretics (yes or no) (29). The association of NRS-2002 with the length of initial hospital stay was first visualized by loess smooth curve. Kaplan–Meier curve was generated to show the cumulative incidence of discharge in patients stratified by high and low nutritional risk and log-rank test was performed. Cox regression analysis was performed to estimate the hazard of discharge, with the same adjustment mentioned above. Hazard ratios (HRs) >1.0 represented shorter length of hospital stay, while HRs < 1.0 represented longer length of hospital stay. To further explore whether NRS-2002 remained effective in patients at low nutritional risk, the total population was divided into four categories (NRS-2002 score = 0, 1, 2, and ≥3). Log-binomial and Cox regression analysis were performed again and the NRS-2002 score = 0 group was regarded as the reference category. Finally, exploratory analysis was conducted according to the classification of heart failure and the level of NT-proBNP fold-elevation.

All statistical tests were 2-sided and p-value <0.05 was considered significant. All statistical analysis were performed with the Statistical Package for Social Science software version 25.0 (SPSS Inc., Chicago, USA) and R software version 4.0.5 (The R Foundation for Statistical Computing, Vienna, Austria).

As shown in Figure 1, a total of 5,919 patients with heart failure were screened for eligibility and 2,830 patients were finally enrolled. Table 1 presents the baseline demographics and clinical features of enrolled patients. The average age was 64.3 ± 11.9 years and 1,880 (66.4%) were male. A total of 122 (4.3%) patients were at high nutritional risk (NRS-2002 score ≥3). A total of 568 (20.1%) patients were re-hospitalized for heart failure after discharge in 1-year follow-up. The median length of initial hospital stay was 5.0 [3.0, 8.0] days. Compared with low nutritional risk group, patients at high nutritional risk had a higher incidence of re-hospitalization (30.3 vs. 19.6%, p = 0.006), and longer length of initial hospital stay (7.0 [5.3, 12.0] days vs. 5.0 [3.0, 7.0] days, p < 0.001). Moreover, high nutritional risk patients had worse cardiac function with lower LVEF (54.5 [41.1, 60.0]% vs. 57.7 [43.5, 65.0]%, p = 0.016) and higher NT-proBNP fold-elevation (2.23 [1.44, 4.35] vs. 1.86 [1.23, 3.08], p = 0.005), and were more likely to be treated with diuretics (39.3 vs. 22.9%, p < 0.001) and nitroglycerin (68.9 vs. 49.4%, p < 0.001). However, there was no statistical difference in sex, smoking status, eGFR, total cholesterol (TC), low density lipoprotein cholesterol (LDL-C), presence of diabetes, hypertension and ischemic cardiomyopathy, and medication with milrinone and cedi-lanid (all p-values >0.05).

Figure 2A shows the population distribution stratified by nutritional categories (NRS-2002 score = 0, 1, 2, and ≥3), and the increasing trend of the incidence of re-hospitalization with the increase of NRS-2002 score. Log-binomial regression analysis was performed to reveal the association of nutritional risk assessed by NRS-2002 with 1-year re-hospitalization. Supplementary Table 2 shows that high nutritional risk (NRS-2002 score ≥3 vs. <3) was an independent risk factor of re-hospitalization (Adjusted model 2: RR = 1.424, 95% confidence interval (CI) = 1.083 to 1.871). Then, NRS-2002 was classified into four categories and the NRS-2002 = 0 group was considered as reference category. Higher NRS-2002 score was still found to be strongly and independently associated with higher incidence of re-hospitalization (Adjusted model 2: [1 vs. 0]: RR = 1.383, 95% confidence interval (CI) = 1.152 to 1.660; [2 vs. 0]: RR = 1.425, 95% CI = 1.108 to 1.832; [3–7 vs. 0]: RR = 1.770, 95% CI = 1.310 to 2.393) (Table 2).

Loess smooth curve was plotted, showing that with the increase of NRS-2002 score, the length of initial hospital stay increased correspondingly (Figure 2B). The cumulative incidence of discharge was also lower in the high nutritional risk group (Log rank p < 0.001) (Figure 3). Cox regression models verified that a high nutritional risk (NRS-2002 score ≥3 vs. <3) was associated with the increasing length of initial hospital stay (Adjusted model 2: HR = 0.636, 95% CI = 0.529 to 0.764) (Supplementary Table 2). Then, NRS-2002 was classified into four categories and the NRS-2002 = 0 group was regarded as a reference category. It was found that higher NRS-2002 score (2 or ≥3) was still tightly and independently associated with longer length of initial hospital stay (Adjusted model 2: [2 vs. 0]: HR = 0.854, 95% CI = 0.748 to 0.976; [3–7 vs. 0]: HR = 0.609, 95% CI = 0.503 to 0.737) (Table 3).

To further explore the effect of NRS-2002 in different subgroups, an exploratory analysis was carried out according to heart failure classification (HFrEF, HFmrEF, or HFpEF) and the level of NT-proBNP fold-elevation (<2 or ≥2). Results of the association between NRS-2002 categories and 1-year re-hospitalization or the length of initial hospital stay were presented in Figure 4, respectively. Figure 4A shows that the positive association of a high NRS-2002 score with 1-year re-hospitalization still remained regardless of NT-proBNP fold-elevation level, but remained only in patients with HFpEF (all p-values <0.05). The results in Figure 4B showed that a high NRS-2002 score (NRS-2002 = 2 or ≥3) was tightly associated with a longer length of initial hospital stay in HFpEF patients or patients whose NT-proBNP fold-elevation <2, while in patients whose NT-proBNP more than 2, such positive association could only be observed when NRS-2002 ≥3.