Interactive information visualization for exploring and querying electronic health records: A systematic review

To overcome the complexity and scale of making medical decisions based on electronic health records (EHRs) a variety of visual methods have been proposed. This paper surveys twelve state-of-the-art information visualization systems described in the scientific literature and compares them based on a set of carefully selected criteria. It aims to systematically examine the systems’ strengths and weaknesses to inform future information visualization designs.We select twelve state-of-the-art information visualization systems from information visualization, medical information systems and human-computer interaction conferences and journals. We compare the systems using the following criteria: (1) data types covered, (2) multivariate analysis support, (3) number of patients records used (one or many), and (4) user intents addressed. The review describes the twelve systems in detail and evaluates the systems using the aforementioned criteria.
We discuss how the systems differ in their features and highlight how these differences are related to their design and affect the user intent model. Examples of findings include: (1) most systems handle numerical or categorical data but not both, (2) most systems are specifically designed for looking at a single patient or multiple patients but not both, (3) most systems utilize horizontal time lines to represent time, (4) only systems that handle multiple patient records have good support for Filter, and (5) some specific user intents (e.g. the Encode and Connect intents) are rarely addressed.
Based on our review results, we believe that effective information visualization can facilitate analysis of patient records, and we encourage the information visualization community to study the application of their systems and conduct more in depth evaluations. We identify potential future research topics in interactive support for data abstraction and medical tasks that involve looking at a single or multiple records. Finally, we propose to create a repository for data and tasks so benchmarks can be established for both academic and commercial patient record visualization systems.