PubMedFrontiers in medicine2026-10-15
The red blood cell distribution width-to-albumin ratio and 28-day mortality in patients with severe cardiogenic shock: a retrospective cohort study based on MIMIC-IV with single-center external validation.
Ou Yi Y, Meng Xiongwei X, Zhang Dan D, Wu Junhong J et al.
Cardiogenic shock (CS) is a critical syndrome with a very high fatality rate, and in-hospital mortality ranges from 27% to 51%. Among the various prognostic approaches, serum biomarkers have attracted considerable attention because they are easy to measure, inexpensive, and can be monitored dynamically. The red cell distribution width-to-albumin ratio (RAR) is a novel composite marker that integrates red blood cell heterogeneity (reflecting inflammation and oxidative stress) with albumin levels (indicating nutritional and anti-inflammatory capacity). It has shown initial promise for risk stratification in critically ill patients. However, the independent prognostic value of RAR specifically in CS remains unclear. Therefore, this study aims to systematically evaluate the ability of RAR to predict 28-day mortality in these patients.
This study used a two-cohort retrospective design. The internal training cohort was derived from CS patients in the MIMIC-IV v2.2 database, and the external validation cohort came from the medical record system of the Affiliated Hospital of Guizhou Medical University. To explore the association between RAR and short-term poor outcomes in severe CS, we applied multivariable Cox regression, restricted cubic spline (RCS) regression, subgroup analysis, and Kaplan-Meier survival curves. To develop a simple prediction model, the internal cohort was randomly split into a training set and a test set in a 7:3 ratio. In the training set, we used five methods-Boruta algorithm, LASSO-Cox regression, random forest (RF), gradient boosting (GBDT), and support vector machine (SVM)-to select key features associated with poor CS prognosis, covering laboratory indicators, comorbidities, demographic characteristics, and vital signs. Based on the selected feature variables, we built prognostic prediction models using multivariable Cox regression in each of the three cohorts. The discriminative ability of the model was compared with that of traditional disease severity scores using receiver operating characteristic (ROC) curves, with the area under the curve (AUC) as the measure of predictive performance.
A total of 1,314 CS patients were included in this study. In the fully adjusted model (Model 3), both continuous RAR and RAR quartiles were independently associated with 28-day adverse outcomes. For 28-day ICU mortality, each unit increase in RAR was associated with a 13.4% higher risk (HR 1.134, 95% CI 1.076-1.197, p < 0.001); compared with the lowest quartile (Q1), the highest quartile (Q4) had a significantly increased risk of 51.7% (HR 1.517, 95% CI 1.123-2.049, p = 0.007). For 28-day in-hospital mortality, the HR for continuous RAR was 1.085 (95% CI 1.031-1.141, p = 0.002), and the HR for Q4 vs. Q1 was 1.576 (95% CI 1.196-2.076, p = 0.001). Restricted cubic spline analysis further confirmed a significant positive dose-response relationship between RAR and mortality (overall p < 0.05). In the external validation cohort, the predictive value of RAR for 28-day ICU mortality remained consistent and was even stronger (HR for continuous RAR: 1.359, 95% CI 1.177-1.568, p < 0.001). the RAR-based risk prediction model showed moderate but consistent discriminative performance, with areas under the receiver operating characteristic curve of 0.71, 0.70, and 0.75 in the training, test, and external validation sets, respectively, all of which were higher than those of traditional severity scores (all DeLong tests p < 0.05).
This study demonstrated in both cohorts that RAR is an independent associated with 28-day mortality risk in CS patients. The prediction model built on RAR and five other clinical features exhibited good clinical performance. With further prospective validation, this model may become a valuable exploratory tool for risk stratification in CS patients.