Personal Health Train on FHIR: A Privacy Preserving Federated Approach for Analyzing FAIR Data in Healthcare

Abstract

Big data and machine learning applications focus on retrieving data on a central location for analysis. However, healthcare data can be sensitive in nature and as such difficult to share and make use for secondary purposes. Healthcare vendors are restricted to share data without proper consent from the patient. There is a rising awareness among individual patients as well regarding sharing their personal information due to ethical, legal and societal problems. The current data-sharing platforms in healthcare do not sufficiently handle these issues. The rationale of the Personal Health Train (PHT) approach shifts the focus from sharing data to sharing processing/analysis applications and their respective results. A prerequisite of the PHT-infrastructure is that the data is FAIR (findable, accessible, interoperable, reusable). The aim of the paper is to describe a methodology of finding the number of patients diagnosed with hypertension and calculate cohort statistics in a privacy-preserving federated manner. The whole process completes without individual patient data leaving the source. For this, we rely on the Fast Healthcare Interoperability Resources (FHIR) standard.