Related Experiment Video
Updated: Sep 19, 2025
Endovascular Perforation Model for Subarachnoid Hemorrhage Combined with Magnetic Resonance Imaging MRI
Published on: December 16, 2021
External Validation of the ARISE Prediction Models for Aneurysmal Rebleeding After Aneurysmal Subarachnoid Hemorrhage
Hendrik-Jan Mijderwijk1, Jordi de Winkel2,3, Daan Nieboer3
1Department of Neurosurgery, Medical Faculty Düsseldorf, Heinrich Heine University, Düsseldorf , Germany.
Background And Objectives:
To externally validate the Aneurysmal RebleedIng after Subarachnoid hEmorrhage (ARISE) prediction models that predict preinterventional aneurysmal rebleeding within 24 and 72 hours after aneurysmal subarachnoid hemorrhage (aSAH).
Methods:
We pooled data from two international hospital registries from University Hospital Oslo, Norway, and University Hospital Rotterdam, The Netherlands, to validate the ARISE base model (including patient age, sex, hypertension, World Federation of Neurological Surgeons grade, Fisher grade, aneurysm size, and cerebrospinal fluid diversion) and the ARISE extended model (adding aneurysm irregularity to the base model). Model performance was assessed with discrimination (Harrell c-statistic, model-based c-statistic) and calibration (calibration-in-the-large, calibration slope, and calibration plots). After validation, we updated the ARISE models as appropriate.
Results:
The combined cohort consisted of 1467 patients, of whom 143 (10%) suffered preinterventional rebleeding. In the University Hospital Oslo, Norway cohort, the externally validated c-statistics were 0.75 (95% CI: 0.71-0.80) for the ARISE base model and 0.71 (0.66-0.76) for the ARISE extended model. In the University Hospital Rotterdam, The Netherlands cohort, the c-statistics were 0.70 (0.64-0.76) for the ARISE base model and 0.64 (0.57-0.72) for the ARISE extended model. Calibration-in-the-large was poor; the average predicted risks were lower than the average observed risk for both models in both centers. After updating the baseline hazard, the base model calibrated excellently over the range of clinically relevant probabilities of rebleeding.
Conclusion:
The ARISE base model had good discriminative ability for the prediction of preinterventional rebleeding in patients suffering from aSAH. Updating the baseline hazard for each center was needed to improve calibration. After local validation and adjustment of the baseline hazard if required, the ARISE baseline model may well be used for risk prediction in patients with aSAH in other settings. The ARISE extended model needs further modification before reliable application can take place.
