<p>Abdominal symptoms are a common reason for emergency department visits, yet early admission decisions remain challenging due to limited diagnostic information available at the initial stage of evaluation. Abdominal radiography is frequently performed as a first-line imaging modality, but its role in supporting early hospitalization decisions remains underexplored. In this study, we propose a multimodal deep learning framework that integrates abdominal radiographs with early clinical information to predict hospitalization outcomes. Convolutional neural networks were used to extract imaging features, while language models were applied to encode clinical text information. These features were combined through a fusion layer to perform admission versus discharge classification. The proposed framework was evaluated using multiple deep learning architectures paired with classical machine learning classifiers. Experimental results demonstrate that the multimodal approach outperforms single-modality models. The best-performing configuration achieved area under the ROC curve (AUC) values of 0.7958 and 0.8638 before and after data augmentation, respectively. These findings suggest that integrating abdominal imaging with early clinical information can improve predictive performance and may support clinical decision-making in emergency care settings.</p>