A new article in Bio-IT World by DCRI Biostatistics Director and Duke Forge founding faculty member Michael Pencina, PhD, offers a perspective on the challenge and potential of using machine learning to enable learning health systems.
As the healthcare system is increasingly permeated by immense amounts of digital data, metaphors of a health data “deluge” abound. In his commentary, Pencina likens the construct of the learning health system, where data drive feedback loops that enable continuous learning and quality improvement as part of routine care delivery, to a hydroelectric dam. As data accumulate and “pool” behind this dam, they can be used to power insights and enable new methods that would be impossible without sophisticated computing technology and techniques.
However, Pencina also cautions against simplistic applications of machine learning in isolation. He notes that a better approach would be one that incorporates a fusion of relevant expertise, pointing to the example of a Duke Forge demonstration project that combines machine and deep learning approaches with clinical experience and insight to guide optimized medical management for patients with diabetes. The project is one of multiple innovative programs that Forge supports by acting as a convener and catalyst, while also fostering clinical leadership and project team autonomy.
Dr. Pencina will be among the featured speakers and discussants at the upcoming Duke Forge Health Data Science Symposium on May 9th, where he will address quantitative methods to accelerate learning.