Healthcare Analytics and Clinical Intelligence: A Risk Prediction Framework for Chronic Care
54 Pages Posted: 1 Jun 2014 Last revised: 20 Jun 2014
Date Written: June 4, 2014
While recent research has suggested the tremendous potential of electronic health records (EHR) to transform healthcare, there remains a limited understanding of the best ways to utilize EHR data to improve clinical decision-making. Healthcare analytics based on EHR data may be able to offer a solution to the challenging goal of providing effective clinical decision support in chronic care. This paper takes a first step towards data-driven, evidence-based healthcare analytics in information systems research. Following the paradigms of design science and predictive analytics research, we propose, demonstrate and evaluate a design framework of risk prediction in the context of chronic disease management. Our framework draws on a large longitudinal real-world EHR dataset and evidence based guidelines to support data- and science-driven clinical decision making. We choose diabetes and coronary heart disease as our experimental cases, each with thousands of patients in their respective cohorts. The results of the experiments suggest that our design can achieve an accurate and reliable predictive performance and that the design is generalizable across chronic diseases. The design artifact and the experimental results contribute to the IS knowledge base and provide important theoretical and practical implications for design science, predictive analytics, and health IT research.
Keywords: design science, predictive analytics, health informatics, time-to-event modeling, theories of abstraction, electronic health records, EHR, health IT, evidence-based medicine
Suggested Citation: Suggested Citation