Krist Wongsuphasawat

LifeFlow: visualizing an overview of event sequences

Event sequence analysis is an important task in many domains: medical researchers may study the patterns of transfers within the hospital for quality control; transportation experts may study accident response logs to identify best practices. In many cases they deal with thousands of records. While previous research has focused on searching and browsing, overview tasks are often overlooked. We introduce a novel interactive visual overview of event sequences called \emph{LifeFlow}. LifeFlow is scalable, can summarize all possible sequences, and represents the temporal spacing of the events within sequences. Two case studies with healthcare and transportation domain experts are presented to illustrate the usefulness of LifeFlow. A user study with ten participants confirmed that after 15 minutes of training novice users were able to rapidly answer questions about the prevalence and temporal characteristics of sequences, find anomalies, and gain significant insight from the data.

Interactive Information Visualization to Explore and Query Electronic Health Records

Physicians are confronted with increasingly complex patient histories based on which they must make life-critical treatment decisions. At the same time, clinical researchers are eager to study the growing databases of patient histories to detect unknown patterns, ensure quality control, and discover surprising outcomes. Designers of Electronic Health Record systems (EHRs) have great potential to apply innovative visual methods to support clinical decision-making and research. This work surveys the state-of-the-art of information visualization systems for exploring and querying EHRs, as described in the scientific literature. We examine how systems differ in their features and highlight how these differences are related to their design and the medical scenarios they tackle. The systems are compared on a set of criteria: (1) data types covered, (2) multivariate analysis support, (3) number of patient records used (one or multiple), and (4) user intents addressed. Based on our survey and evidence gained from evaluation studies, we believe that effective information visualization can facilitate analysis of EHRs for patient treatment and clinical research. Thus, we encourage the information visualization community to study the application of their systems in health care. Our monograph is written for both scientific researchers and designers of future user interfaces for EHRs. We hope it will help them understand this vital domain and appreciate the features and virtues of existing systems, so they can create still more advanced systems. We identify potential future research topics in interactive support for data abstraction, in systems for intermittent users, such as patients, and in more detailed evaluations.

Finding comparable temporal categorical records: A similarity measure with an interactive visualization

An increasing number of temporal categorical databases are being collected: Electronic Health Records in healthcare organizations, traffic incident logs in transportation systems, or student records in universities. Finding similar records within these large databases requires effective similarity measures that capture the searchers intent. Many similarity measures exist for numerical time series, but temporal categorical records are different. We propose a temporal categorical similarity measure, the M&M (Match & Mismatch) measure, which is based on the concept of aligning records by sentinel events, then matching events between the target and the compared records. The M&M measure combines the time differences between pairs of events and the number of mismatches. To accom-modate customization of parameters in the M&M measure and results interpretation, we implemented Similan, an interactive search and visualization tool for temporal categorical records. A usability study with 8 participants demonstrated that Similan was easy to learn and enabled them to find similar records, but users had difficulty understanding the M&M measure. The usability study feedback, led to an improved version with a continuous timeline, which was tested in a pilot study with 5 participants.

Extracting Insights from Electronic Health Records: Case Studies, a Visual Analytics Process Model, and Design Recommendations

Current electronic health record (EHR) systems facilitate the storage, retrieval, persistence, and sharing of patient data. However, the way physicians interact with EHRs has not changed much. More specifically, support for temporal analysis of a large number of EHRs has been lacking. A number of information visualization techniques have been proposed to alleviate this problem. Unfortunately, due to their limited application to a single case study, the results are often difficult to generalize across medical scenarios. We present the usage data of Lifelines2 (Wang et al. 2008), our information visualization system, and user comments, both collected over eight different medical case studies. We generalize our experience into a visual analytics process model for multiple EHRs. Based on our analysis, we make seven design recommendations to information visualization tools to explore EHR systems.

Visual information seeking in multiple electronic health records: design recommendations and a process model

Current electronic health record (EHR) systems facilitate the storage, retrieval, persistence, and sharing of patient data. However, the way physicians interact with EHRs has not changed much. More specifically, support for temporal analysis of a large number of EHRs has been lacking. A number of information visualization techniques have been proposed to alleviate this problem. Unfortunately, due to their limited application to a single case study, the results are often difficult to generalize across medical scenarios. We present the usage data of Lifelines2,[22] our information visualization system, and user comments, both collected over eight different medical case studies. We generalize our experience into an information-seeking process model for multiple EHRs. Based on our analysis, we make recommendations to future information visualization designers for EHRs on design requirements and future research directions.

Querying event sequences by exact match or similarity search: Design and empirical evaluation

Specifying event sequence queries is challenging even for skilled computer professionals familiar with SQL. Most graphical user interfaces for database search use an exact match approach, which is often effective, but near misses may also be of interest. We describe a new similarity search interface, in which users specify a query by simply placing events on a blank timeline and retrieve a similarity-ranked list of results. Behind this user interface is a new similarity measure for event sequences which the users can customize by four decision criteria, enabling them to adjust the impact of missing, extra, or swapped events or the impact of time shifts. We describe a use case with Electronic Health Records based on our ongoing collaboration with hospital physicians. A controlled experiment with 18 participants compared exact match and similarity search interfaces. We report on the advantages and disadvantages of each interface and suggest a hybrid interface combining the best of both.

Visual Analytics for Transportation Incident Data Sets

Transportation systems are being monitored at an unprecedented scope, which is resulting in tremendously detailed traffic and incident databases. Although the transportation community emphasizes developing standards for storing these incident data, little effort has been made to design appropriate visual analytics tools to explore the data, extract meaningful knowledge, and represent results. Analyzing these large multivariate geospatial data sets is a nontrivial task. A novel, web-based, visual analytics tool called Fervor is proposed as an application that affords sophisticated, yet user-friendly, analysis of transportation incident data sets. Interactive maps, histograms, two-dimensional plots, and parallel coordinates plots are four featured visualizations that are integrated to allow users to interact simultaneously with and see relationships among multiple visualizations. Using a rich set of filters, users can create custom conditions to filter data and focus on a smaller data set. However, because of the multivariate nature of the data, finding interesting relationships can be a time-consuming task. Therefore, a rank-by-feature framework has been adopted and further expanded to quantify the strength of relationships among the different fields describing the data. In this paper, transportation incident data collected by the Maryland State Highway Administrations CHART program are used; however, the tool can be easily modified to accept other transportation data sets.

ICE--visual analytics for transportation incident datasets

Transportation systems are being monitored at an unprecedented scope resulting in tremendously detailed traffic and incident databases. While the transportation community emphasizes developing standards for storing this incident data, little effort has been made to design appropriate visual analytics tools to explore the data, extract meaningful knowledge, and represent results. Analyzing these large multivariate geospatial datasets is a non-trivial task. A novel, web-based, visual analytics tool called ICE (Incident Cluster Explorer) is proposed as an application that affords sophisticated yet user-friendly analysis of transportation incident datasets. Interactive maps, histograms, two-dimensional plots and parallel coordinates plots are four visualizations that are integrated together to allow users to simultaneously interact with and see relationships between multiple visualizations. Accompanied by a rich set of filters, users can create custom conditions to filter data and focus on a smaller dataset. Due to the multivariate nature of the data, a rank-by-feature framework has been expanded to quantify the strength of relationships between the different fields.

Towards event sequence representation, reasoning and visualization for EHR data

Efficient analysis of event sequences and the ability to answer time-related, clinically important questions can accelerate clinical research in several areas such as causality assessments, decision support systems, and retrospective studies. The Clinical Narrative Temporal Reasoning Ontology (CNTRO)-based system is designed for semantically representing, annotating, and inferring temporal relations and constraints for clinical events in Electronic Health Records (EHR) represented in both structured and unstructured ways. The LifeFlow system is designed to support an interactive exploration of event sequences using visualization techniques. The combination of the two systems will provide a comprehensive environment for users to visualize inferred temporal relationships from EHR data. This paper discusses our preliminary efforts on connecting the two systems and the benefits we envision from such an environment.

LifeFlow: visualizing an overview of event sequences (video preview)

Event sequence analysis is an important task in many domains: medical researchers may study the patterns of transfers within the hospital for quality control; transportation experts may study accident response logs to identify best practices. In many cases they deal with thousands of records. While previous research has focused on searching and browsing, overview tasks are often overlooked. We introduce a novel interactive visual overview of event sequences called LifeFlow. LifeFlow is scalable, can summarize all possible sequences, and represents the temporal spacing of the events within sequences. In this video, we show an example of patient transfer data and briefly demonstrate how to analyze them with LifeFlow. Please see [11] or visit http:www.cs.umd.eduhcillifeflow for more detail.