Prognostic performance of the EHMRG score and the EHMRG-BI model in geriatric acute heart failure

Abstract

This study aimed to evaluate the prognostic performance of the Emergency Heart Failure Mortality Risk Grade (EHMRG) score for predicting 30-day mortality in geriatric patients with acute heart failure (AHF) and to determine whether incorporating brain natriuretic peptide (BNP) and endotracheal intubation status could improve risk prediction. This retrospective, single-center observational study included patients aged ≥ 65 years with AHF who presented to the emergency department and were subsequently hospitalized between January 2024 and January 2026. Demographic, clinical, laboratory, and outcome data were obtained from electronic medical records. A multivariable prediction model integrating the EHMRG score, BNP, and endotracheal intubation status (EHMRG-BI) was derived and initially evaluated in the same cohort. Because model development and initial performance assessment were conducted in the same cohort, internal validation with 1,000 bootstrap resamples was performed to quantify optimism and obtain optimism-corrected performance estimates. A total of 348 patients were included, of whom 60 (17.2%) died within 30 days. In the multivariable prediction model, EHMRG score, BNP, and endotracheal intubation were independently associated with 30-day mortality after adjustment for the other variables included in the model. The EHMRG-BI model showed greater apparent discrimination than the original EHMRG score (AUC, 0.777 [95% CI, 0.712–0.843] vs. 0.681 [95% CI, 0.606–0.757]; DeLong p = 0.002). Following bootstrap internal validation, the optimism-corrected AUC was 0.772, indicating limited optimism in model discrimination. The optimism-corrected calibration intercept and slope were − 0.060 and 0.950, respectively, and the optimism-corrected Brier score was 0.126. The EHMRG-BI model demonstrated improved apparent discrimination for 30-day mortality compared with the original EHMRG score in this single-center derivation cohort, with performance largely preserved after bootstrap internal validation. These findings should be considered preliminary, and prospective external validation is required before its routine clinical implementation can be recommended.