A Knowledge Graph–Driven Dietary Recommendation App for Chronic Kidney Disease: Development and Preliminary Accuracy Validation

AIMS: This study aimed to construct a chronic kidney disease (CKD) dietary knowledge graph and develop a WeChat-based mini-program for personalized dietary recommendations. METHODS: Using a seven-step ontology method, a schema was built and real-world data integrated to form the graph. The mini-program was developed and preliminarily validated by comparing its nutrient recommendations with those of two renal dietitians using data from 40 CKD patients. RESULTS: The knowledge graph comprised 1825 entities and 15,141 semantic relations. The developed mini-program included two core functions: user information input and dietary recommendation. The recommendation module integrated personalized features such as allergy history collection and ingredient substitution, aiming to balance health requirements with individual preferences. Accuracy for core nutrients ranged from 90% to 100%. CONCLUSION: The system offers a practical, accurate, and personalized digital tool for CKD dietary management with promising clinical applicability. IMPACT: This knowledge graph-driven app provides nurses with a reliable digital tool to deliver consistent, evidence-based dietary education while offering patients personalized and accessible guidance that improves adherence, strengthens self-management, and supports better outcomes. PATIENT OR PUBLIC CONTRIBUTION: No patient or public contribution.