We propose a new way of managing chronic diseases that brings together the fields of medicine and data science. We hypothesise that the interaction between individuals with the relapsing and remitting autoimmune kidney disease ANCA vasculitis, and their environment, can be detected and defined by observing the whole system in action and integrating a wide array of data sources. Collaboration with IBM will allow us to take advantage of recent advances in machine learning technology that allows iterative refinement of algorithms to generate a technology readiness level 3 (proof of concept) physician tool, to present analysis results in a clinically meaningful way. The ultimate goal of this approach is to define the signature of vasculitis relapse and use this to aid in planning and delivery of optimum immunosuppressive therapy at the level of the patient. To achieve this, we will use advanced data science methodologies and Bayesian statistical techniques to develop a data architecture that curates and combines from four sources: Fixed patient-level factors (HLA-DP phenotype, granular clinical dataset obtained at diagnosis), External medical influences (maintenance immunosuppression, antibiotic prescriptions, Hospital Inpatient Enquiry records), External environmental influences, linked to patient location through time (meteorological data streams, community pathogen patterns: readily available as online data streams) and Direct patient-derived data sources (location, patient-reported quality of life and accelerometer defined activity). We expect a 50% rate of relapse after 5 years in a cohort of patients derived from the Rare Kidney Disease registry; we shall describe for the first time the relapse prodrome and define in great detail the environmental influences linking to this event. The goal of the current two year project is to establish the infrastructure and obtain one year of data, with a view to obtaining Horizon 2020 support for a full five years of data capture.
