Validation Of The BAN-ADHF Risk Score And Risk Of Long-term Mortality In Patients With Acute Decompensated Heart Failure

Abstract

Introduction Heart failure (HF) remains one of the leading causes of mortality, especially after patients were admitted for acute decompensated HF (ADHF). However, risk scores for long-term risk prediction after developing ADHF using common clinical parameters have not been well studied. Recently, the BAN-ADHF risk score was developed using machine learning-based variable importance metrics in a cohort of patients with ADHF to predict diuretic resistance. However, its ability to predict long-term outcomes in real world cohort of patients with ADHF has not been established. Hypothesis We hypothesized that the BAN-ADHF risk score will accurately stratify risk of long-term all-cause mortality in patients with ADHF. Methods Electronic medical records of 965 patients admitted with ADHF at the Cleveland Clinic were reviewed. We excluded 70 patients with missing data for calculating the risk score for a final analysis cohort of 895 patients. The BAN-ADHF risk score were calculated using clinical variables and weighted scores, which included baseline BUN, Cr, NT-proBNP, diastolic blood pressure (DBP), age, history of diabetes, atrial fibrillation, and years of HF. Participants were categorized into quintile groups of BAN-ADHF scores (≤5, 6-7, 8, 9-11, and ≥12). Kaplan Meier and Cox proportional hazard models were used to evaluate the association between the BAN-ADHF score groups and mortality with differences determined using the log-rank test. Results Mean age was 69.8±15.6 years, 49.9% were men, 71.6% had AF, 58.4% had diabetes, and 85% had CKD. Median (IQR) was 71 (62-80) mmHg for DBP, 21 (15-10) mg/dL for BUN, 1.10 (0.87-1.50) mg/dL for Cr, and 2,856 (1,042.5-7,025.5) pg/mL for NT-proBNP. During median follow up of 1,467 days, 467 (52.2%) patients died. A step-wise increase in all-cause mortality was observed across increasing BAN-ADHF quintiles (Figure, log-rank p<0.001). Compared to Quintile 1, participants with BAN-ADHF scores in Quintile 5 had a 192% higher risk of all-cause mortality (HR 2.92, 95% CI 2.18-3.92, p<0.001). Conclusions Higher BAN-ADHF risk scores are significantly associated with an increased risk of long-term all-cause mortality in patients with ADHF. Heart failure (HF) remains one of the leading causes of mortality, especially after patients were admitted for acute decompensated HF (ADHF). However, risk scores for long-term risk prediction after developing ADHF using common clinical parameters have not been well studied. Recently, the BAN-ADHF risk score was developed using machine learning-based variable importance metrics in a cohort of patients with ADHF to predict diuretic resistance. However, its ability to predict long-term outcomes in real world cohort of patients with ADHF has not been established. We hypothesized that the BAN-ADHF risk score will accurately stratify risk of long-term all-cause mortality in patients with ADHF. Electronic medical records of 965 patients admitted with ADHF at the Cleveland Clinic were reviewed. We excluded 70 patients with missing data for calculating the risk score for a final analysis cohort of 895 patients. The BAN-ADHF risk score were calculated using clinical variables and weighted scores, which included baseline BUN, Cr, NT-proBNP, diastolic blood pressure (DBP), age, history of diabetes, atrial fibrillation, and years of HF. Participants were categorized into quintile groups of BAN-ADHF scores (≤5, 6-7, 8, 9-11, and ≥12). Kaplan Meier and Cox proportional hazard models were used to evaluate the association between the BAN-ADHF score groups and mortality with differences determined using the log-rank test. Mean age was 69.8±15.6 years, 49.9% were men, 71.6% had AF, 58.4% had diabetes, and 85% had CKD. Median (IQR) was 71 (62-80) mmHg for DBP, 21 (15-10) mg/dL for BUN, 1.10 (0.87-1.50) mg/dL for Cr, and 2,856 (1,042.5-7,025.5) pg/mL for NT-proBNP. During median follow up of 1,467 days, 467 (52.2%) patients died. A step-wise increase in all-cause mortality was observed across increasing BAN-ADHF quintiles (Figure, log-rank p<0.001). Compared to Quintile 1, participants with BAN-ADHF scores in Quintile 5 had a 192% higher risk of all-cause mortality (HR 2.92, 95% CI 2.18-3.92, p<0.001). Higher BAN-ADHF risk scores are significantly associated with an increased risk of long-term all-cause mortality in patients with ADHF.