Abstract
Chronic diseases are a critical focus in the management of elderly health. Early disease prediction plays a vital role in achieving disease prevention and reducing the associated burden on individuals and healthcare systems. Traditionally, separate models were required to predict different diseases, a process that demanded significant time and computational resources. In this research, we utilized a nationwide dataset and proposed a multi-task learning approach combined with a multimodal disease prediction model. By leveraging patients' medical records and personal information as input, the model predicts the risks of diabetes mellitus, heart disease, stroke, and hypertension simultaneously. This approach addresses the limitations of traditional methods by capturing the correlations between these diseases while maintaining strong predictive performance, even with a reduced number of features. Furthermore, our analysis of attention scores identified risk factors that align with previous research, enhancing the model's interpretability and demonstrating its potential for real-world applications.