An eight-variable risk model for short- and long-term mortality in hospitalized heart failure

Abstract

Abstract Heart failure (HF) is a leading cause of hospitalization and death worldwide, yet most prognostic models target long-term outcomes and are not optimized for acute-phase risk stratification. This study developed and externally validated an 8-variable Cox model for predicting 28-day and 1-year all-cause mortality in hospitalized HF patients, following the TRIPOD statement. Using the MIMIC-IV v3.1 database ( n = 12,269) for development and temporal validation and a Chinese cohort ( n = 1284; 1,278 with complete data) for geographic external validation, LASSO-penalized Cox regression was used for variable selection. The 28-day model achieved C-index values of 0.738, 0.736, and 0.830 in the development, temporal validation, and Guangdong cohorts, respectively, and showed better discrimination than the 1-year model (ΔC-index = 0.058 in the development set). The systemic immune-inflammation index (SII) showed a stronger association with 28-day mortality (HR 1.40, 95% CI 1.32–1.48) than with 1-year mortality (HR 1.18, 95% CI 1.14–1.23; bootstrap interaction P = 0.001). A web-based risk calculator was externally validated after slope + intercept recalibration. The 8-variable model and calculator provide practical tools for early HF risk stratification.