Abstract Body (Do not enter title and authors here): Introduction: Pre-discharge and in-hospital trends of B-type natriuretic peptide (BNP) are predictive of mortality in patients recently discharged after acute decompensated heart failure (ADHF). The Intermountain risk score (IMRS) uses common serum laboratory data to predict outcomes in hospitalized patients. Considering that in-hospital BNP levels differ for HF with preserved (HFpEF) and HF with non-preserved EF (HFnpEF), the goal of this work was to assess whether BNP at admission and discharge provided prognostic information beyond the IMRS in either phenotype.
Methods: In a hospitalized cohort with ADHF in Intermountain Health, adult patients were included if they had 1) a BNP value within 24 hours of admission, and 2) a separate BNP value within 3 days pre-discharge. HF phenotypes were determined using echocardiography within 1 year prior to admission (HFpEF >= 50%, HFnpEF < 50%). IMRS calculated on admission from CBC and CMP components. Within each HF phenotype, time-dependent ROC analyses were used to derive admission and discharge BNP thresholds for 1-year mortality (highest product of sensitivity and specificity). Patients were categorized as above or below their HF-specific admission and discharge thresholds (low-low, low-high, high-low, high-high). Nested multivariable Cox proportional hazard models for 1-year mortality evaluated admission BNP, discharge BNP, relative change in BNP, and admission-discharge BNP profile, in addition to IMRS and comorbidities.
Results: A total of 2,933 patients were included, where 43.9% (1,287) had HFpEF and 56.1% (1,646) HFnpEF. Median BNP was lower in HFpEF compared to HFnpEF (admit: 703 vs 1300 pg/mL; discharge: 406 vs 796 pg/mL, both p<0.001). In HFpEF, admission and discharge BNP thresholds were 763 pg/mL and 474 pg/mL, respectively, while in HFnpEF, the thresholds were 1520 pg/mL, and 1043 pg/mL. The best model above the baseline model used admission-discharge BNP profile (Table 1). In both HF phenotypes, the low-low profile had the lowest 1-year mortality, and the high-high profile had the highest mortality (Figure 1). After multivariable adjustment, each admission-discharge BNP profile was associated with higher 1-year mortality vs. low-low, meeting significance in high-high profile in HFpEF and high-high and low-high profiles in the HFnpEF (Table 2).
Conclusions: The inclusion of admission-discharge BNP profiles was complementary and added significant and incremental value to the IMRS to predict mortality.