Description
We designed and developed an automated system that generates personalized nutritional goal recommendations for individuals with type 2 diabetes (T2D) using machine learning with patient-generated health data (PGHD) and a rule-based expert system. In a controlled evaluation with 15 participants with T2D from an underserved population, participants could understand recommendations and assemble meals consistent with recommended goals. This study demonstrates the feasibility of using computational methods to generate actionable recommendations from PGHD.
Learning Objective: Understand the challenges in algorithmically generating nutritional recommendations from patient-generated health data, especially for patients from disadvantaged communities with low nutrition and health literacy.
Authors:
Elliot Mitchell (Presenter)
Columbia University
Marissa Burgermaster, The University of Texas at Austin
Elizabeth Heitkemper, Columbia University
Matthew Levine, California Institute of Technology
Yishen Miao, Columbia University
Esteban Tabak, Courant Institute of Mathematical Sciences
Arlene Smaldone, Columbia University School of Nursing
David Albers, University of Colorado
Lena Mamykina, Columbia University
Presentation Materials: