Stress hyperglycemia ratio as a predictor of hydrocephalus in critically ill patients with intracerebral hemorrhage: a retrospective study and machine learning-based model

Hydrocephalus is a severe complication of intracerebral hemorrhage (ICH), and its early prediction remains challenging. The stress hyperglycemia ratio (SHR), reflecting acute glucose elevation relative to baseline, has been associated with poor outcomes in ICH, but its relationship with hydrocephalus is unclear. We retrospectively analyzed 1,383 critically ill ICH patients from the MIMIC-IV database and evaluated associations between SHR and hydrocephalus using multivariable logistic regression, restricted cubic spline analysis, and subgroup analyses. Hydrocephalus occurred in 8.97% of patients, with higher incidence in those with elevated SHR (11.91% vs. 5.59%, P < 0.001). SHR was independently associated with hydrocephalus as both a continuous (OR = 1.72, 95% CI: 1.20–2.48, P = 0.003) and categorical variable (OR = 1.75, 95% CI: 1.15–2.69, P = 0.01), with linear risk increase above 1.05. A logistic regression model combining SHR with six clinical variables (mean SpO₂, intraventricular hemorrhage, mannitol use, mechanical ventilation, sepsis, and age) achieved best predictive performance (AUC = 0.81) and maintained accuracy in an external cohort (AUC = 0.76, n = 513). These findings indicate that SHR is a valuable predictor of hydrocephalus after ICH and may facilitate early risk stratification and individualized clinical management.