Large-scale repositories of secondary-use patient data are emerging as a critical resource for both clinical and epidemiological research. Motivated by this opportunity, a variety of interactive visual analysis methods have been developed to make the use of this data more efficient and accessible. These techniques often combine interactive filters and on-demand computational analysis to allow ad hoc cohort exploration and refinement. This approach has indeed made it possible to quickly select and revise cohorts during analysis. However, the seemingly simple filters supported by these tools can produce dramatic—and often unseen—confounding effects on the makeup of the cohort across the thousands of variables often found in real-world medical data. This project explores how visual analytics methods can be used to create interfaces which interactively measure and visually convey to users the degree of drift in representation during iterative visual cohort selection.  The methods are being developed as part of a system called Tempo, a visualization-based cohort selection system.

Publications

David Gotz, Shun Sun, and Nan Cao. Adaptive Contextualization: Combating Bias During High-Dimensional Visualization and Data Selection. ACM International Conference on Intelligent User Interfaces (IUI) (2016). Best Paper Award
[PDF, 4.9M]

David Gotz and Shun Sun. Visual Assessment of Cohort Divergence During Iterative Cohort Selection. Visual Analytics in Healthcare Workshop Posters (2015).
[PDF, 1.9M]