Dynamic sodium dysregulation predicts mortality in intracerebral hemorrhage: A retrospective cohort study of temporal variability and absolute thresholds.

Dysnatremia and sodium variability are emerging prognostic markers in neurocritical care, yet their role in intracerebral hemorrhage (ICH) remains underexplored. This study investigates the independent and synergistic effects of serum sodium variability (coefficient of variation, CV) and absolute sodium levels on mortality in ICH patients. We conducted a retrospective cohort study using data from the Medical Information Mart for Intensive Care (MIMIC-IV) database. Sodium variability was quantified as CV, while mean sodium levels were categorized into normonatremia (135-145 mmol/L), hyponatremia (< 135 mmol/L), and hypernatremia (> 145 mmol/L). Outcomes included 30-day, in-hospital, and ICU mortality. Multivariable Cox models adjusted for demographics, disease severity, and interventions. A total of 2044 ICH patients were included in the study. Increased sodium variability exhibited a robust linear association with mortality, with each 1% rise in CV independently predicting a 38% higher 30-day mortality risk (adjusted HR 1.38, 95% CI 1.31-1.46; P < 0.001). Patients in the highest CV quartile (Q4) faced a 2.48-fold mortality risk compared to the lowest quartile (Q1). Hypernatremia (> 145 mmol/L) emerged as a potent independent predictor of mortality (HR 2.93, 95% CI 2.43-3.52; P < 0.001), whereas hyponatremia showed non-significant trends (HR 1.17, 0.85-1.61). Restricted cubic spline analyses revealed a U-shaped relationship between mean sodium levels and mortality, with 140.7 mmol/L as the optimal threshold, while sodium variability demonstrated a linear dose-response effect. Both sodium variability and hypernatremia independently predict mortality in ICH, with CV demonstrating a linear dose-response relationship and absolute levels showing a U-shaped risk curve. These findings advocate for dual monitoring of sodium homeostasis to guide precision fluid management in ICH.