Predictive value of hemoglobin-to-red cell distribution width ratio for mortality in new-onset atrial fibrillation after cardiac surgery a machine learning study based on the MIMIC-IV database

Postoperative atrial fibrillation (POAF) is a common complication after cardiac surgery and is associated with increased mortality. Hemoglobin (Hb) and red cell distribution width (RDW), reflecting oxygen-carrying capacity and systemic stress, have prognostic relevance. The hemoglobin-to-RDW ratio (HRR) integrates these parameters, but its prognostic value in POAF after cardiac surgery remains underexplored. This retrospective study included patients from the Medical Information Mart for Intensive Care (MIMIC)-IV 3.1 database. The primary outcomes were 90-day, 180-day, and 360-day all-cause mortality among patients with POAF after cardiac surgery. Cox proportional hazards models and survival analyses were used to evaluate the association between HRR and mortality. Fully adjusted models excluded hemoglobin and RDW to avoid potential overadjustment. Machine learning models were further developed to assess the predictive performance of HRR. Among 9384 patients undergoing cardiac surgery, 3315 developed postoperative atrial fibrillation. Lower HRR was independently associated with increased 90-day, 180-day, and 360-day all-cause mortality after adjustment for potential confounders. A nonlinear, L-shaped relationship between HRR and mortality was observed, with higher HRR values associated with improved survival. Among machine learning models, CatBoost demonstrated the best predictive performance, with area under the curve values of 0.871, 0.855, and 0.801 for 90-day, 180-day, and 360-day mortality, respectively. Lower HRR is independently associated with increased short- and long-term mortality in patients with POAF after cardiac surgery. As a simple and readily available biomarker, HRR may assist early postoperative risk stratification, although prospective multicenter studies are needed for further validation.