BNP-Informed Prediction of In-Hospital Mortality in Acute Decompensated Heart Failure: An Admission-Based Model in a Cardiorenal Syndrome-Enriched Cohort

Background: Admission-based risk stratification in acute decompensated heart failure (ADHF) remains challenging, particularly in cohorts enriched for cardiorenal syndrome type 1 (CRS1). B-type natriuretic peptide (BNP) is the most extensively validated admission biomarker in ADHF, yet its independent contribution alongside heart failure (HF) phenotype and serum albumin within a prespecified multivariable mortality prediction model has not been formally established in CRS-enriched populations. Methods: In a retrospective cohort of consecutive index ADHF admissions (N=220 complete cases) at a single center enriched for CRS1, we developed a prespecified multivariable logistic regression model to predict in-hospital death using: age, sex, HF phenotype (HFpEF/HFmrEF/HFrEF), systolic blood pressure (SBP), estimated glomerular filtration rate (eGFR), serum albumin, and log-transformed BNP [ln(BNP)]. Discrimination was assessed by the area under the receiver operating characteristic curve (AUC) with 200-iteration bootstrap optimism correction. Calibration was assessed across risk deciles, and clinical utility was evaluated by decision curve analysis. Reporting followed the TRIPOD statement. Results: Seventeen patients (7.7%) died during the index hospitalization. ln(BNP) was the sole statistically significant independent predictor of in-hospital mortality (OR 2.39 per ln-unit; 95% CI 1.25–4.59; p=0.009). Albumin and eGFR showed consistent directional associations with mortality. The model demonstrated good apparent discrimination (AUC 0.81), with an optimism-corrected AUC of 0.73. Decision curve analysis indicated net benefit at threshold probabilities of 5–30%. A prespecified two-variable sensitivity model (albumin + ln[BNP]) yielded AUC 0.77, confirming the robustness of these two markers. Conclusions: This exploratory, internally validated model incorporating BNP, albumin, eGFR, and HF phenotype demonstrated promising discrimination for in-hospital mortality in a CRS-enriched ADHF cohort. The principal contribution is the application of a formally prespecified, TRIPOD-reported admission model in a CRS-enriched population, rather than identifying BNP as a novel prognostic marker. ln(BNP) was the sole statistically significant independent predictor. These findings are hypothesis-generating and require external validation before any clinical deployment.