In large hospitals, manual quality control of medical records is labor-intensive, lacks unified standards, and is difficult to integrate with other data platforms. This project develops a comprehensive intelligent outpatient electronic medical record (EMR) quality control system. At its core is an AI-driven quality control engine that applies natural language processing to perform semantic parsing and retrieval of medical entities and events, enabling intrinsic quality assessment. The system can extract key information from EMR texts—such as “elevated serum creatinine, eyelid edema, bilateral kidney shrinkage”—and provide diagnostic predictions with confidence levels. A large-scale medical knowledge graph has been constructed, containing over 100,000 entities and 1.4 million relationships, covering symptoms, diseases, examinations, and treatments across multiple specialties. Leveraging graph neural network computation on this knowledge base significantly improves both accuracy and interpretability, reducing biases of purely data-driven models and ensuring reliable quality control for multi-disease outpatient records.