External validation of the scoring models to predict shunt insertion after aneurysmal subarachnoid haemorrhage.

BACKGROUND The current study is an external validation of four scoring models proposed in the literature for predicting ventriculoperitoneal shunt insertion after aneurysmal subarachnoid haemorrhage (aSAH) using retrospective patient data from Sheffield Teaching Hospital (STH). METHODS Data was collected on various demographics and patients were individually scored using the four scoring models and the models compared with each other using receiver-operator characteristic curves. The best model had the highest area under the curve. RESULTS A total of 301 aSAH patients were referred to the Neurosurgery department in STH between 1st January 2014 and 31st December 2017. Scoring model 4 also had the largest area under the curve of 0.853 (p<0.001) and scoring model 3 had the lowest area under the curve of 0.654 (p=0.036). CONCLUSIONS Scoring model 4 was found to be the best scoring model out of the four scoring models externally validated to predict shunt dependency after an aSAH in STH patients. Scoring model 4 is less applicable in modern practice due to a higher proportion of coiling and use of the Hunt and Hess grade. There is a need for a new scoring model to predict shunt insertion in modern day practice.