The application of health recommender systems based on knowledge graph: a scoping review

Abstract

Abstract Background: Tailored knowledge graph-based recommender systems (KGRSs) have been demonstrated to be able to provide accurate and effective health recommendations to users, and thus significantly reduce health care costs. They are now strongly recommended to be applied in the health care field. Objective: This scoping review aims to identify the current application of KGRSs, their target users and performance metrics, and the potential limitations of implementing health recommender systems in clinical practice. Methods: A review of the studies published from inception to November 1, 2022 was conducted, using key search terms in 6 scientific databases to identify health recommender systems based on knowledge graph technology. Key information from the included studies was extracted and charted. The scoping review was reported following the PRISMA Extension for Scoping Reviews. Result: We included 16 studies and 5 grants totally about the health recommender systems based on knowledge graph technology. They were used in different health areas: traditional Chinese medicine, health management, disease-related decision support, diet, and nutrition recommendations. Among them, 6 studies were for the general public and 6 were for physicians. A total of 13 (81.25%) studies evaluated the KGRS using performance metrics, such as accuracy, recall, F1 score, and area under the curve. All studies pointed out the limitations of the recommender systems and provided directions for their subsequent optimization and improvement. Conclusion: This review describes the state-of-the-art and potential limitations of KGRS used in the health care field. This novel approach has been proven to be effective in overcoming the drawbacks of traditional algorithms, helping users filter massive amounts of data to find out the personalized information they need. Its great potential in digital health needs to be further explored.