A Personal Health Recommender System Incorporating Personal Health Records, Modular Ontologies, and Crowd-Sourced Data

We present an architecture for a Personal Health Recommender System (PHRS) that begins with a person's personal electronic health record (PEHR) and augments it by combining crowd-sourced data mined for symptoms, diseases, treatments and best practices, together with authoritative sources and curated domain ontologies. This novel approach is patient-centric in that the PEHR contains the patient's health stats, specific symptoms, diagnoses, illnesses, medications and treatment plans. By accessing anonymized crowd-sourced data for similar cases, the patient and healthcare providers can ascertain the best course of treatment. In addition, the data in the PEHR can be classified according to authoritative ontologies using Semantic Web services. A modular ontology may be constructed for each of the patient's illnesses, and then they may be combined into an integrated ontology by patient or illness. As the PEHR is populated with more data, the ontology may evolve, guided by resources and services available on the Semantic Web.