Multimodal Data Fusion and Reinforcement Learning-Driven Intelligent Nursing Decision Support System

In response to the problems of traditional nursing decision-making relying on manual experience and insufficient dynamic adaptability, this study proposes an intelligent nursing decision support system (AI CareDSS) that integrates multimodal data perception and reinforcement learning. The system adopts an improved spatiotemporal graph convolutional network (STGCN) to model the dynamic correlation between patient physiological parameters and nursing operations, and designs a decision optimization model based on dual delay depth deterministic policy gradient (TD3). By introducing dynamic priority experience replay mechanism (DPER) and multi-objective constraint conditions, personalized recommendation and risk control of nursing operations can be achieved. The experiment showed that on a simulated dataset containing 12 clinical scenarios, the clinical conformity rate of the system's recommended decisions reached 96.3%, with a response time of less than 0.8 seconds, significantly better than traditional expert systems (p<0.01). This research provides innovative solutions for real-time nursing decision-making in the context of smart healthcare.