Associations of Red Blood Cell Distribution Width-Derived Indicators and Their Longitudinal Dynamic Trajectories With...

Red blood cell distribution width (RDW)-derived indicators have increasingly been recognized as biomarkers reflecting systemic inflammation and hematological disorders. However, the prognostic value of these indicators in critically ill patients with pulmonary hypertension (PH), especially the dynamic implications of their temporal changes, remains insufficiently elucidated. A total of 990 adult patients diagnosed with PH in the intensive care unit (ICU) based on ICD-9/10 codes were enrolled from the MIMIC-IV database. The association between baseline levels of three key indicators and the 1-year all-cause mortality risk was assessed using Cox proportional hazards models and restricted cubic spline (RCS) analysis, and the consistency of this association was assessed for external consistency in the eICU-CRD cohort (n = 298) and NWICU cohort (n = 182). In the subgroup of patients with ≥ 3 consecutive laboratory tests (n = 316), group-based multi-trajectory modeling (GBTM) was employed to identify the joint longitudinal trajectory patterns of these three indicators. The robustness of the prediction results was then evaluated using machine learning models as secondary exploratory analysis. In the MIMIC-IV cohort, higher baseline RAR (aHR = 1.17, 95% CI: 1.08-1.27) and RPR (aHR = 2.31, 95% CI: 1.38-3.86) were significantly associated with increased 1-year mortality risk, whereas higher HRR exerted a protective effect (aHR = 0.31, 95% CI: 0.17-0.58). Across the multiple cohorts, only RAR remained statistically significant in the multivariate model of the eICU-CRD cohort (p = 0.009); although RPR and HRR showed a consistent association, the results did not reach statistical significance. GBTM identified three distinct joint trajectory patterns. Compared with the reference group (Class 2), the high-risk trajectory group (Class 3, n = 35) exhibited a significantly elevated mortality risk (aHR = 2.48). In the exploratory analysis, the support vector machine (SVM) demonstrated moderate discriminative ability (AUC = 0.782). RDW-derived indicators are independently associated with the mortality risk in critically ill patients with ICD-code defined PH. Both their baseline levels and short-term change trajectories may reflect the severity of the underlying disease and systemic stress response. These findings suggest that RDW-derived indicators play an important role in risk phenotyping; however, prospective studies are warranted to confirm their clinical significance.