2018/02/23 by Matthew E. Levine, Levine, Matthew E., David J. Albers +9
Medicine · Biochemistry, Genetics and Molecular Biology · #Nutritional Studies and Diet #Nutrition, Genetics, and Disease
paper · pdf · doi:10.48550/arxiv.1802.08761
Objective: To evaluate unsupervised clustering methods for identifying\nindividual-level behavioral-clinical phenotypes that relate personal biomarkers\nand behavioral traits in type 2 diabetes (T2DM) self-monitoring data. Materials\nand Methods: We used hierarchical clustering (HC) to identify groups of meals\nwith similar nutrition and glycemic impact for 6 individuals with T2DM who\ncollected self-monitoring data. We evaluated clusters on: 1) correspondence to\ngold standards generated by certified diabetes educators (CDEs) for 3\nparticipants; 2) face validity, rated by CDEs, and 3) impact on CDEs' ability\nto identify patterns for another 3 participants. Results: Gold standard (GS)\nincluded 9 patterns across 3 participants. Of these, all 9 were re-discovered\nusing HC: 4 GS patterns were consistent with patterns identified by HC (over\n50% of meals in a cluster followed the pattern); another 5 were included as\nsub-groups in broader clusers. 50% (9/18) of clusters were rated over 3 on\n5-point Likert scale for validity, significance, and being actionable. After\nreviewing clusters, CDEs identified patterns that were more consistent with\ndata (70% reduction in contradictions between patterns and participants'\nrecords). Discussion: Hierarchical clustering of blood glucose and\nmacronutrient consumption appears suitable for discovering behavioral-clinical\nphenotypes in T2DM. Most clusters corresponded to gold standard and were rated\npositively by CDEs for face validity. Cluster visualizations helped CDEs\nidentify more robust patterns in nutrition and glycemic impact, creating new\npossibilities for visual analytic solutions. Conclusion: Machine learning\nmethods can use diabetes self-monitoring data to create personalized\nbehavioral-clinical phenotypes, which may prove useful for delivering\npersonalized medicine.\n