IDMVis: Temporal event sequence visualization for type 1 diabetes treatment decision support

Screenshot of the IDMVis interface.
IDMVis is an interactive visualization tool for showing type 1 diabetes patient data. It is designed to help clinicians perform temporal inference tasks: specifically for recommending adjustments to patient insulin protocol, diet, and behavior. The overview panel displays two weeks of event sequence data at a glance; the detail panel shows additional information for the selected day; and the summary statistics panel displays the distribution of insulin and carbohydrate intake overall and for specific events (e.g., breakfast)
Abstract
Type 1 diabetes is a chronic, incurable autoimmune disease affecting millions of Americans in which the body stops producing insulin and blood glucose levels rise. The goal of intensive diabetes management is to lower average blood glucose through frequent adjustments to insulin protocol, diet, and behavior. Manual logs and medical device data are collected by patients, but these multiple sources are presented in disparate visualization designs to the clinician—making temporal inference difficult. We conducted a design study over 18 months with clinicians performing intensive diabetes management. We present a data abstraction and novel hierarchical task abstraction for this domain. We also contribute IDMVis: a visualization tool for temporal event sequences with multidimensional, interrelated data. IDMVis includes a novel technique for folding and aligning records by dual sentinel events and scaling the intermediate timeline. We validate our design decisions based on our domain abstractions, best practices, and through a qualitative evaluation with six clinicians. The results of this study indicate that IDMVis accurately reflects the workflow of clinicians. Using IDMVis, clinicians are able to identify issues of data quality such as missing or conflicting data, reconstruct patient records when data is missing, differentiate between days with different patterns, and promote educational interventions after identifying discrepancies.
Materials
DOI | Code | Demo | Video Preview | Demo Video | BibTeX
Authors
Citation

Cody Dunne, Vis Lab — Northeastern University
West Village H, Room 302F
440 Huntington Ave, Boston, MA 02115, USA