Visual Approaches to Bring Population Data Insights at Your Fingertips

Background/Aims: The data surge in electronic medical records offers a tremendous potential to leverage it to impact population/patient care delivery. However, to make sense of large data volume requires significant time commitment and attention –– both of which are scarce resources for stakeholders in decision-making roles. Thus, a need exists to use technology that improves our ability to make faster inferences and decisions from data. At Kaiser Permanente Mid-Atlantic States, we are exploring novel ways to combine computational cycles and human intuitions in a learning loop to enhance care delivery experience. Our aim is simple: Instead of spending hours trying to gather information, we wanted to come up with a more user-intuitive approach that offers an easy way to sift through multidimensional data –– to have the information right there at your fingertips for insights. Methods: We focused on human immunodeficiency virus (HIV) cohort data from Kaiser Permanente Mid-Atlantic States. In contrast to our current approach of disseminating the report in tabular format, we adopted a visual data-driven strategy to best support user’s tasks and to yield more information in less time. For rapid development, we used Tableau to build visual interface. From project inception, stakeholders were part of the design process to bring important perspectives. Data views that can be snapped together in a dashboard interface were built. Visual data encoding was used to reduce user’s cognitive effort, and filters were added to help customize views to user’s needs. Results: The built HIV dashboard offered at a glance overview of population. Supporting interaction offered flexibility to instantly update data view to different inclusion/exclusion combinations as they came to user’s mind. It offers decision makers macro-to-micro level population insights like: how the population is doing at the regional/service area/medical center level, how the population is distributed among different sites, and others. Conclusion: The designed, visual, data-driven tool attempts to augment decision makers’ cognition, support their ability to gather data insights and facilitate decision-making. Leveraging computational tools and designs, we show an approach to divide and conquer data, where the best of both human intuition and knowledge can be augmented by interaction with tools to boost performance