Construction and prototype effect evaluation of a multi-agent collaborative system for operating room nursing

Objective To develop an operating room intelligent collaborative management system, define its intelligent auxiliary role for nursing teams, and evaluate its efficacy in process optimization, efficiency improvement, and clinical acceptance. Methods Guided by standards like the Guidelines for Operating Room Nursing Practice, five position-mapped agents (scheduling, resource, early warning, quality control, interaction) were designed. The system integrates a Graph RAG-based knowledge engine, MindsDB-powered AutoML prediction engine, and an innovative function to automatically construct/visualize knowledge graphs from uploaded nursing documents, with a user-friendly human-machine interface adapted to operating room settings. Simulated scenario tests and a 5-point Likert scale survey (86 medical staff) were conducted. Results The system achieved success rates of 95.0% (resource conflicts), 90.0% (emergency insertion), 85.0% (equipment failures), and 90.0% (special coordination), with average solution generation time of 22.3–41.6 s. Overall nursing satisfaction was (4.32 ± 0.51) points, with top scores in “process optimization perception” (4.40 ± 0.48) and “decision support value” (4.35 ± 0.52). Conclusion Integrating knowledge-driven and data-driven intelligence, the system enables automatic knowledge graph construction and updates. As an effective digital assistant for nursing collaboration, its nurse-centric, operating room-adapted design has gained wide clinical recognition, offering a human-machine collaboration solution for intelligent nursing management.