Which Doctor to Trust: A Recommender System for Identifying the Right Doctors
Date Submitted: May 22, 2016
Open Peer Review Period: May 24, 2016 - Jun 2, 2016
Background: In the field of medical/health informatics, it is critical to identify key opinion leaders (KOLs) on different diseases. KOLs refer to the people who have the ability in influencing public opinions on the subject matter for which the opinion leaders have the authority. This is important for both health industry and patients. However, no study has tried to solve this issue systematically. Objective: We aim to develop a recommender system for identifying KOLs for any specific disease with the methods of healthcare data mining methods. Methods: In this paper, an unsupervised aggregation approach is exploited for integrating these ranking features to find the right doctors who have the reputation as KOLs on different diseases. Furthermore, we introduce the design, implementation, and deployment details of the recommender system. In this system, the professional footprints of doctors, such as articles in scientific journals, presentation activities, patient advocacy, and media exposure, are collected and used as ranking features for KOL identification. Results: We collect and exploit a large amount of public information (as much as 3,657,797 medical journal articles) related to almost all doctors (as much as 2, 381, 750 doctors) in China, including their profiles, academic publications, funding, etc. Comparative experiments have been done on the datasets and demonstrate that the proposed system outperforms several benchmark systems with similar purpose with a significant margin. Moreover, a case study is performed in real-world system to verify the reasonableness of our proposed method. Conclusions: The results suggest that doctor profile and academic publication are useful data sources for identifying key opinion leaders in the field of medical/health informatics. Moreover, the recommender system has been deployed and applied the data service for a recommender system of NetEase. Patients could obtain authority ranking lists of doctors with this system. Clinical Trial: As this was a research on method and application of doctor recommendation, no patients were involved and no intervention was performed, thus trial was not registered.