Beyond the scores: identifying key predictors of in-hospital mortality in critically ill patients with acute heart failure

Abstract Introduction Heart failure (HF) frequently complicates the course of critical illness and remains a leading cause of death among ICU patients. Despite advances in organ support and monitoring, predicting outcomes in acute HF remains challenging. Identifying early clinical predictors of in-hospital mortality may help clinicians recognize high-risk patients and tailor interventions accordingly. Purpose To identify independent clinical and severity-related predictors of in-hospital mortality among critically ill patients with acute heart failure (AHF). Methods Single-center retrospective cohort of consecutive ICU patients who had decompensated HF at admission or during ICU between 2020 and 2023. Disease severity was assessed using the SAPS II, APACHE II, and SOFA scores at 24 hours after admission. The following variables were analyzed as potential predictors of in-hospital mortality: left ventricular ejection fraction (LVEF) category, serum creatinine, serum sodium, lactate, invasive mechanical ventilation, cardiogenic shock, renal replacement therapy (RRT), extracorporeal membrane oxygenation (ECMO). Variables significantly associated with mortality in univariate analysis were entered into a multivariable logistic regression (backward LR) to identify independent predictors. Model performance was assessed using ROC curve analysis. Results A total of 113 ICU patients with AHF were included. Overall in-hospital mortality was 33.6% (n = 38). The median age was 69 years, and 60.5% were male. Non-survivors presented significantly higher severity scores: SAPS II 61.1 ± 20.5 vs. 52.5 ± 17.3 (p = 0.026), APACHE II 28.1 ± 9.7 vs. 23.5 ± 9.1 (p = 0.019), and SOFA 24-hour 10.4 ± 3.2 vs. 9.0 ± 3.2 (p = 0.038). Renal replacement therapy (RRT) was required in 26 patients (23.0%) and was more frequent among non-survivors (42.1% vs. 13.3%, p 0.001). Lactate levels were higher in non-survivors (1.97 1.33 - 4.13 vs. 1.46 1.00 - 2.25 mmol/L, p = 0.047). Other variables were not significantly associated with in-hospital mortality. In the multivariable logistic regression model, higher APACHE II score and the need for renal replacement therapy were independently associated with in-hospital mortality. Each additional point in APACHE II increased the odds of death by 5.2% (OR = 1.05, 95% CI 1.00 - 1.10, p = 0.040), while RRT conferred a 4.8-fold higher risk of death (OR = 4.85, 95% CI 1.85 - 12.73, p = 0.001). The final model showed good discriminative ability, with an area under the ROC curve (AUC) of 0.738 (95% CI 0.63 - 0.85, p = 0.003). Conclusion In critically ill patients with acute heart failure, higher disease severity (APACHE II) and need for renal replacement therapy were independent predictors of in-hospital mortality. These findings highlight the prognostic value of physiological severity at admission and the importance of renal dysfunction as a marker of poor outcome.