PEI-YUN S. HSUEH

PEI-YUN S. HSUEH

PEI-YUN S. HSUEH

IBM Academy of Technology Member, and Research Scientist, Center for Computational Health, IBM T.J. Watson Research Center

Integrating Data Science with Science of Care for Precision
Behavioral factors are the key contributors to mental health risk and morbidity, accounting for 41 percent of global disease burden. Recent studies documented the importance of accounting for individuality and heterogeneity in human health behavior through personalized approaches. In practice, varying behavioral responses are often revealed in patient care history. The rise of consumer awareness and the prevalence of personal health technologies (e.g., mobiles, sensors, wearables) have further enabled the accumulation of personal health data for interpretation. However, today’s care programs are structured around population-level evidence, but not personal understanding. What if healthcare professionals can take advantage of the revealed behavioral understanding to further engage target patients and personalize their care plans? To address the multi-level challenge, recently, in addition to traditional clinical and epidemiological methods, novel AI and machine learning algorithms are being proposed. The goal of this talk is to review the development of an interpretable behavioral learning pipeline that captures individual predictive pathways from observational behavior data. As the black-box nature of AI/ML has widened the gap between how humans and machines make decisions, we will also outline the lessons underlying current practice for making AI/ML more interpretable and actionable in health informatics. Example showcases will help illustrate how to support precision health applications that are maximally patient-centric yet minimally disruptive.