Big Data Analytics in Healthcare

A vast volume of digitized clinical data has been generated and accumulated rapidly since the widespread adoption of Electronic Medical Records (EMRs). These big data in healthcare hold the promise of propelling healthcare to evolve from a proficiency-based art to data-driven science, from a reactive mode to a proactive mode, and from one-size-fits-all medicine to personalized medicine. This chapter first discusses the research background—big data analytics in healthcare, the research framework of big data analytics in healthcare, analysis of the medical process, and the literature summary of diagnosis-treatment pattern mining. Then the challenges for data-driven typical diagnosis-treatment pattern mining are highlighted, including similarity measures between diagnosis and treatment records, typical diagnosis-treatment pattern extraction, prediction, evaluation, and recommendation, when considering the rich temporal and heterogeneous medical information in EMRs. Furthermore, a data-driven unifying diagnosis identification and prediction method (UDIPM) embedding the disease ontology structure is proposed from EMRs to assist in better coding integration of diagnosis. Three categories of typical treatment patterns are mined from doctor order content, duration, and sequence view respectively, which can provide a data-driven guideline to achieve the “5R” goal for rational drug use and clinical pathways.