Throughout history, humans have leveraged technology to enhance healthcare and medical treatment. Traditional approaches relied on limited methods and tools to assess and predict an individual’s health status. However, with recent advancements in pervasive computing, data mining, and deep learning, it has become possible to provide personalized health prediction and assistance with greater accuracy and efficiency. While research on electronic health records (EHRs) has opened new opportunities, it has also introduced significant challenges—particularly in predicting a patient’s condition after hospital discharge or during hospitalization. In this paper, we propose a novel method based on a Supervised Variational Autoencoder (SVAE) for predicting patient health outcomes. The model is designed to address post-discharge and in-hospital prediction tasks while maintaining simplicity in preprocessing and input requirements. Our proposed approach achieves performance comparable to or better than state-of-the-art methods, despite relying on fewer input variables. The results demonstrate the potential of the SVAE framework for real-world healthcare data analysis and its practical applicability in clinical decision-support systems.