Quantitative versus qualitative blood amount assessment as a predictor for shunt-dependent hydrocephalus following aneurysmal subarachnoid hemorrhage

OBJECTIVE: Reliable tools are lacking to predict shunt-dependent hydrocephalus (SDHC) development after aneurysmal subarachnoid hemorrhage (aSAH). Quantitative volumetric measurement of hemorrhagic blood is a good predictor of SDHC but might be impractical in the clinical setting. Qualitative assessment performed using scales such as the modified Fisher scale (mFisher) and the original Graeb scale (oGraeb) is easier to conduct but provides limited predictive power. In between, the modified Graeb scale (mGraeb) keeps the simplicity of the qualitative scales yet adds assessment of acute hydrocephalus, which might improve SDHC-predicting capabilities. In this study the authors investigated the likely capabilities of the mGraeb and compared them with previously validated methods. This research also aimed to define a tailored mGraeb cutoff point for SDHC prediction. METHODS: The authors performed retrospective analysis of patients admitted to their institution with the diagnosis of aSAH between May 2013 and April 2016. Out of 168 patients, 78 were included for analysis after the application of predefined exclusion criteria. Univariate and multivariate analyses were conducted to evaluate the use of all 4 methods (quantitative volumetric assessment and the mFisher, oGraeb, and mGraeb scales) to predict the likelihood of SDHC development based on clinical data and blood amount assessment on initial CT scans. RESULTS: The mGraeb scale was demonstrated to be the most robust predictor of SDHC, with an area under the curve (AUC) of 0.848 (95% CI 0.763-0.933). According to the AUC results, the performance of the mGraeb scale was significantly better than that of the oGraeb scale (χ2 = 4.49; p = 0.034) and mFisher scale (χ2 = 7.21; p = 0.007). No statistical difference was found between the AUCs of the mGraeb and the quantitative volumetric measurement models (χ2 = 12.76; p = 0.23), but mGraeb proved to be the simplest model since it showed the lowest Akaike information criterion (66.4), the lowest Bayesian information criterion (71.2), and the highest R2Nagelkerke coefficient (39.7%). The initial mGraeb showed more than 85% specificity for predicting the development of SDHC in patients presenting with a score of 12 or more points. CONCLUSIONS: According to the authors' data, the mGraeb scale is the simplest model that correlates well with SDHC development. Due to limited scientific evidence of treatments aimed at SDHC prevention, we propose an mGraeb score higher than 12 to identify patients at risk with high specificity. This mGraeb cutoff point might also serve as a useful prognostic tool since patients with SDHC after aSAH have worse functional outcomes.