ADHERE Tree Model for Acute HF Mortality Calculator

ADHERE Classification and Regression Tree (CART) — explanation and clinical context The ADHERE CART is a simple bedside decision tree developed from the Acute Decompensated Heart Failure National Registry to estimate the risk of in-hospital mortality for patients hospitalized with acute decompensated heart failure. The CART analysis identified three sequential admission variables that provided the best discrimination: admission blood urea nitrogen (BUN), admission systolic blood pressure (SBP), and serum creatinine. Specifically, the published branch points are BUN ≥ 43 mg/dL, SBP < 115 mmHg, and serum creatinine ≥ 2.75 mg/dL; using these splits the original derivation produced five terminal risk groups with predicted in-hospital mortality that ranged from approximately 2.1% (low risk) to 21.9% (high risk).

These thresholds permit rapid stratification at the bedside without calculator apps or complex scoring. Clinical interpretation: the ADHERE tree is intended as a pragmatic risk-stratification aid at hospital presentation. A patient with low BUN (<43 mg/dL) and SBP ≥115 mmHg is classified as low risk (predicted in-hospital mortality ≈ 2.1% in the original cohort), whereas the combination of high BUN (≥43), low SBP (<115) and markedly elevated creatinine (≥2.75) identifies the small high-risk subgroup with substantially higher observed mortality (≈21.9% in the derivation cohort).

Intermediate groups (for example BUN ≥43 but SBP ≥115, or BUN <43 but SBP <115) have intermediate predicted mortality (originally reported ~5.6% and ~4.9% respectively, and one intermediate node ~13.2%). These node probabilities were reported in ADHERE and reproduced in multiple external validations and comparisons. Limitations & recommended use: the ADHERE tree was derived from a large US registry and provides quick prognostic insight, but calibration and absolute probabilities vary by population and time: several validation studies found differences between predicted and observed mortalities and advise local validation before using the exact percentages to make treatment-limiting decisions.

Use clinical judgment—treat the ADHERE group as an adjunct (triage / escalation signal), not a replacement for bedside assessment or local prognostic models. For contemporary deployments, consider local recalibration (or use local registry estimates) if you require absolute probability precision.