Heart failure (HF) is the leading cause of hospitalization among adults over 65 years of age in the United States. Despite advances in medical and device therapy, the rate of rehospitalization remain quite high up to ~50% at 6 months [1], and 70% of these rehospitalizations are related to worsening HF [2, 3]. A major portion of the financial burden in HF is represented by readmission rates. This has stimulated public and private payers to target readmission as a main focus of pay-for-performance initiatives [4] and to apply monetary penalties to hospitals with high readmission rates. This prompted research to study predictors that assist in risk-stratifying patients and identifying those at higher risk of complications. Among these predictors is the B‑type natriuretic peptide (BNP), whose absolute value on admission, on discharge, and its percentage reduction throughout hospitalization is associated with postdischarge outcomes [5,6,7]. We have also previously advocated that the value of longitudinal BNP monitoring [8] is more important for diagnostic and prognostic purposes than utilizing a single BNP measurement [9]. We speculated that a rise in BNP in the early post-discharge period is a sign of increased congestion.

The aim of the current study was to examine whether the magnitude of the rise in BNP from discharge to the 1‑month follow-up of patients hospitalized with decompensated HF is associated with an increased risk of rehospitalization.

Methods

Study population and the ESCAPE trial

The Evaluation Study of Congestive Heart Failure and Pulmonary Artery Catheterization Effectiveness (ESCAPE) trial randomized 433 patients with acute HF to either clinical management guided by pulmonary artery catheterization (PAC) or to clinical management alone. The patients had a left ventricular ejection fraction <30%, 3 months of HF symptoms despite appropriate therapy, a systolic blood pressure <125 mm Hg, and at least one symptom and sign of congestion. Patients with creatinine levels >3.5 mg/dl, and those who required dobutamine or dopamine 3 g/kg/min or milrinone before randomization, were excluded. The study showed that PAC did not affect outcomes, which was the number of days alive outside the hospital at 6 months following randomization [10].

BNP measurement

Patients enrolled in the ESCAPE trial had BNP measured at multiple study time points including at admission, discharge, 1 month, 2 months, 3 months, and 6 months. BNP was measured using the Shionogi assay as previously described. NT-proBNP was not measured in the ESCAPE trial. We hypothesized that, in patients hospitalized with decompensated HF, the magnitude of the rise in BNP level from discharge to 1‑month follow-up (BNP at 1 month—discharge BNP) would predict future rehospitalization. Patients who were rehospitalized within the first 4 weeks of discharge were thus excluded from the current analysis. The study endpoint was all-cause rehospitalization up to 6 months after randomization.

Statistical analysis

Primary analysis compared patients who were rehospitalized versus those who were not rehospitalized following the index hospitalization with decompensated HF. Owing to the small sample size and nonnormality of distribution of most variables, we compared continuous variables using the Mann–Whitney U test and presented the results as median and interquartile (IQR) range. Categorical variables were compared using the Chi-square test or Fisher’s exact test as appropriate, and the data are presented as counts and percentages. The paired-sample t test was used to compare longitudinal BNP values at 1 month relative to discharge. The ability of the magnitude of BNP increase from discharge to the 1‑month follow-up to predict all-cause rehospitalization was assessed using receiver operating characteristics (ROC) curves. Optimum BNP cut-off values were those that provided the highest combined sensitivity and specificity for predicting the study outcome. Statistical analysis was performed using IBM SPSS 21.0 statistical software (version 21.0. Armonk, NY, USA) and the MedCalc software (version 16.8, bbva, Ostend, Belgium). Statistical significance was assessed using two-sided p values and was set at p ≤0.05.

Results

Among the 433 patients enrolled in the ESCAPE trial, 51 (12%) had BNP levels recorded both at hospital discharge and at the 1‑month follow-up. Seven of 51 patients were excluded from the current analysis since they were rehospitalized within the first 4 weeks of discharge. The remaining 44 patients had a mean age of 56 years, 71% were men, 66% were Caucasian, and 91% were classified as New York Heart Association (NYHA) class IV at baseline. At 6 months following randomization, 61% (27/44) of those patients were rehospitalized, and the 6‑month mortality rate was 13.6% (6/44). The average BNP value on discharge for these 44 patients was 467 pg/ml (median: 330 pg/ml; IQR: 142–599 pg/ml), which increased to 919 pg/ml (median: 565 pg/ml; IQR: 176–1,392 pg/ml) at the 1‑month follow-up (p = 0.001). There were no significant differences in rehospitalized and non-rehospitalized patients with regards to markers of congestion on hospital discharge such as the frequency of NYHA class IV (p = 1.000), the presence of rales (p = 0.634), a positive hepatojugular reflux (p = 0.749), at least 1+ lower extremity edema (p = 1.000), jugular venous distension >8 cm (p = 0.724), or BNP level (p = 0.294). Patients who were rehospitalized after 1 month of hospital discharge had a significantly higher magnitude of rise in BNP from discharge to the 1‑month follow-up compared with those who were not rehospitalized (median [IQR]: 329 [11, 956] vs. 44 [−90, 316] pg/ml, p = 0.039, in both groups, respectively). ROC showed that the absolute rise in BNP from discharge to the 1‑month follow-up had an area under the curve (AUC) of 0.686 (95% confidence interval [CI]: 0.529–0.8181, p = 0.0255) in predicting rehospitalization, and an absolute value of BNP rise >191 pg/ml had the highest combined sensitivity (67%) and specificity (71%) for predicting rehospitalization (Fig. 1). Among the 27 patients who were rehospitalized, the average time to rehospitalization was 79 days (median [IQR]: 75 [45–101] days). Table 1 shows the results of the comparison of the baseline characteristics of ESCAPE trial patients according to whether or not they were rehospitalized after the index hospitalization.

Fig. 1
figure 1

a Box and whisker plot illustrating the relationship between the magnitude of BNP change from discharge to 1‑month follow-up and all-cause rehospitalization among ESCAPE trial patients. The five horizontal lines represent the 10th, 25th, 50th, 75th, and 90th percentiles of each variable, from bottom to top, excluding outliers shown as circles and extreme outliers shown as an asterisk. b Receiver operator characteristics curve showing the ability of BNP rise from discharge to 1‑month follow-up to predict all-cause rehospitalization. BNP B‑type natriuretic peptide

Full size image
Table 1 Demographic, clinical, laboratory, echocardiographic, and central hemodynamic characteristics of ESCAPE trial patients according to whether or not they were rehospitalized
Full size table

Discussion

We have shown by univariate comparison that the magnitude of the rise in BNP level from hospital discharge to the 1‑month follow-up was significantly higher in patients who were rehospitalized later on compared with those who were not rehospitalized. An absolute increase in BNP of ~200 pg/ml from discharge to the 1‑month follow-up predicted rehospitalization; this highlights its value for outpatient evaluation of patients recently discharged after HF decompensation, even in those whose symptoms do not suggest overt HF. These findings were observed despite similar degree of clinical congestion and comparable BNP values on hospital discharge of those who will be rehospitalized or not. It is known that rehospitalizations for HF are preceded by a gradual rise in ventricular filling pressure that begins more than 2 weeks before any detectable changes in weight or clinical symptoms [11]. Data from trials of implantable hemodynamic monitors show that the risk for HF events is related to the degree of filling pressure elevation, with progressively higher risk once the median 24-h pulmonary artery diastolic pressure rises >18 mm Hg [12]. In the ESCAPE trial, among the 244 patients who were rehospitalized, 163 (67%) were rehospitalized more than 30 days after discharge, and therefore there is potential for prevention of a portion of these hospitalizations by utilizing the findings in the current study. The magnitude of the rise in BNP level from discharge to the 1‑month follow-up, therefore, can identify those at risk for readmission who may require medication adjustments, more frequent follow-up visits, or outpatient-focused strategies known to minimize hospital readmissions.

We have not come across any study that examined (a) the ability of the rise in BNP from discharge to the 1‑month follow-up to predict rehospitalization and (b) the magnitude of BNP elevation that is clinically relevant. These results also further confirm our previous findings that longitudinal follow-up of BNP levels between two points in time is more valuable for its diagnostic and prognostic implications compared to the utility of single BNP measurement.

BNP is a known marker of volume overload. It is widely used in various cardiovascular conditions, primarily in HF, thanks to its diagnostic and prognostic value [13,14,15]. A rise in BNP levels results from an elevation in end-diastolic wall stress, stiffness, and pressure load [15, 16]. When utilized in conjunction with clinical findings, BNP may guide the initial diagnostic work-up and treatment of patients with suspected HF [14, 17, 18]. Moreover, a BNP-guided strategy reduced the risk of HF-related death or hospital stay for HF in the Systolic Heart Failure Treatment Supported by BNP (STARS-BNP) study [19]. In the ProBNP Outpatient Tailored Chronic HF Therapy (PROTECT) study, the substantial reduction in NT-proBNP was accompanied by a significant reduction in a composite outcome, including worsening HF, hospitalization for HF, and cardiovascular death [20]. Similarly, in patients with HF due to left ventricular systolic dysfunction, NT-proBNP-guided therapy was superior to clinical management, with lower event rates, better quality of life, and cardiac remodeling.

It is therefore logical to expect that an increase in BNP levels after discharge can predict rehospitalization. Studies have shown that BNP values measured after treatment were more predictive of post-discharge outcomes than were values on presentation. Many actually consider discharge BNP or BNP checked shortly after discharge as “baseline,” and treatment should be targeted to maintain the BNP level close to this value.

BNP is influenced by multiple other factors—independent of the severity of HF—such as age [21], body mass index [22], sex, chronic kidney disease [23], and congenital factors [24]. Also, BNP levels correlate poorly with filling pressures [25], therefore, utilizing single BNP measurements to predict rehospitalization can be challenging [9]. The interpatient and intrapatient variability in BNP levels has minimized the value of single measurements, and added more weight to the diagnostic and prognostic ability of individualized longitudinal BNP measurements, which acknowledges the inherent between-patient variability in BNP levels.

Study limitations

There are several limitations to the current study. The main limitation is the small number of cases, which allowed only for univariate comparison. Adjustment for covariates known to affect BNP level—other than severity of HF—was not possible because of the small study population. The ESCAPE study included patients with severe systolic HF and advanced symptoms that do not represent the general HF population, and therefore our findings should not be extrapolated to the general HF population. Whether therapeutic interventions initiated after monitoring BNP level at the 1‑month follow-up relative to discharge are accompanied by decreasing readmission rates needs to be investigated in prospective trials. Because of the small number of cases, our results can best be viewed as hypothesis generating and should be confirmed in larger studies.

Conclusion

The increase in BNP level from hospital discharge to the 1‑month follow-up was significantly higher in HF patients who were rehospitalized later on compared with those who were not rehospitalized. The magnitude of the rise in BNP level from discharge to the 1‑month follow-up is therefore a useful prognostic tool for predicting rehospitalization in patients recently discharged after HF decompensation.