Development of the IS-Score: A Novel Predictive Model for Shunt Requirement Following Aneurysmal Subarachnoid Hemorrhage.

BACKGROUND

Chronic hydrocephalus after aneurysmal subarachnoid hemorrhage is a common condition that can require shunting of cerebrospinal fluid to the peritoneal cavity [ventriculoperitoneal shunting (VPS)]. Several predictive scoring systems have been developed to estimate the risk of requiring VPS, most recently the Mayo age, grade, EVD (MAGE) score and the SAH-VP score. The aim of this study was to identify predictors of ventriculoperitoneal shunt dependency and to develop a new predictive score to compare with the MAGE and SAH-VP scores.

METHODS

A retrospective single-center cohort study of 220 patients admitted with computed tomography (CT)-verified aneurysmal subarachnoid hemorrhage (aSAH), of whom 114 required extraventricular drainage, between 1 September 2018 and 21 January 2025, was conducted. They were categorized into two groups: those requiring VPS (n = 62) and those who did not (n = 52). Univariate and multivariate analyses (least absolute shrinkage and selection operator logistic regression combined with the extended Bayesian information criterion for the latter) were performed. Statistically significant variables were entered into a generalized linear model. A clinical risk score was developed by scaling and rounding the bootstrapped regression coefficients relative to the smallest significant effect.

RESULTS

Least absolute shrinkage and selection operator (LASSO) regression identified seven predictors: intensive care unit (ICU) stay ≥10 days (OR 10.85), World Federation of Neurological Surgeons (WFNS) score of 5 (OR 1.35), EVD duration≥14 days (OR 1.73), two or more failed weaning attempts (OR 5.38), clamp failure (OR 2.99), cerebrospinal fuid (CSF) output>200 mL in the frst 3 days (OR 3.67), and failed closure after weaning (OR 4.42). The final model, named the Islamagič Shunt score (IS-score) demonstrated a bootstrapped area under the receiver operating characteristic curve (AUC) of 0.914 (95% CI 0.87-0.95). In comparison, the SAH-VP yielded an AUC of 0.736 (95% CI 0.66-0.80) and the MAGE score an AUC of 0.791 (95% CI 0.73-0.85).

CONCLUSIONS

The IS-score demonstrated superior predictive performance for VPS following aSAH compared with the MAGE and SAH-VP scores, but further external validation is needed.