Association between RDW-to-albumin ratio and mortality in HFpEF: a retrospective study based on MIMIC-IV and external validation

Objective Heart failure with preserved ejection fraction (HFpEF) in the intensive care unit (ICU) has high mortality, yet reliable bedside prognostic markers remain limited. The red cell distribution width-to-albumin ratio (RDW/Alb), reflecting inflammation and nutrition, has not been validated in this setting. Methods This retrospective cohort study queried the MIMIC-IV (v2.2) database for adults (≥18 years) with first ICU admission and HFpEF (left ventricular ejection fraction ≥50% by ICD coding or echocardiographic narrative). RDW and serum albumin within 24 h of ICU entry were used to calculate RDW/Alb, analyzed as tertiles (T1 ≤ 4.08; T2 4.08–5.13; T3 > 5.14). The primary endpoint was all-cause mortality at 30, 90, and 365 days. Kaplan–Meier curves, multivariable Cox regression, restricted cubic splines (RCS), and subgroup analyses were conducted. Prognostic discrimination of RDW/Alb was compared with the triglyceride-glucose (TyG) index in a biomarker-complete subset. Findings were externally validated in 429 HFpEF patients from general wards at our hospital. Results Among 3,436 ICU-HFpEF patients, 659 (19.2%), 907 (26.4%), and 2,997 (87.3%) deaths occurred at 30, 90, and 365 days, respectively. Mortality rose stepwise across tertiles (30-day: 8.0% vs 16.2% vs 33.3%; log-rank < 0.001). In fully adjusted models, each unit increase in RDW/Alb was associated with 12% higher hazard for 30-day (HR 1.12, 95% CI 1.10–1.15) and 90-day mortality (HR 1.12), and a 10% increase for 1-year mortality (HR 1.10, 95% CI 1.07–1.12). Compared with T1, T3 patients had HRs of 3.13, 3.02, and 1.37 for 30-, 90-, and 365-day mortality (all p < 0.001). RCS revealed a nonlinear risk surge above an RDW/Alb of 4.56. The association remained across subgroups and was stronger in females, non-diabetics, and non-statin users (interaction < 0.01). In 490 patients with glucose and triglyceride data, RDW/Alb outperformed TyG in predicting mortality (AUC 0.67–0.68 vs 0.52–0.54; p < 0.01). External validation confirmed RDW/Alb as a predictor of 1-year mortality (HR for T3 vs T1: 2.90; 95% CI: 1.55–5.41; p < 0.001). Conclusion RDW/Alb is a simple, widely available marker that strongly predicts mortality in ICU patients with HFpEF, outperforming TyG and supporting its role in early risk stratification.