Accurate outcome prediction after aneurysmal subarachnoid hemorrhage (aSAH) remains challenging due to the heterogeneity of clinical and imaging data. This study compares four machine learning approaches for three-month survival prediction using volumetric CT data and clinical metadata from 733 patients across nine Spanish centers. Using the AUCMEDI framework, we implemented: (1) an image-only 3D DenseNet121 CNN, (2) a CNN with metadata integration via feature-level late fusion, (3) a metadataonly XGBoost model, and (4) a fusion XGBoost model combining CNN predictions with metadata. The image-only CNN achieved the best performance (F1-score = 0.78, AUC = 0.84), confirming the strong prognostic value of CT imaging. Adding metadata slightly reduced performance, while the fusion XGBoost reached a competitive F1-score (0.76) and improved interpretability. These results demonstrate that CT-based deep learning enables reliable survival prediction in aSAH, and that model-level fusion provides aninterpretable alternative.