Postoperative complications remain a critical concern in clinical practice, adversely affecting patient outcomes and contributing to rising healthcare costs. This work presents MIRACLE, a deep learning architecture for predicting postoperative complication risk in lung cancer surgery by integrating preoperative clinical and radiological data. MIRACLE uses a hyperspherical embedding space to fuse heterogeneous inputs and includes an interventional module that provides interpretable, actionable insights that domain experts can adjust based on clinical expertise.