Sequen-C: A multilevel overview of temporal event sequences
Jessica Magallanes, Department of Computer Science
Tony Stone, Centre for Urgent and Emergency Care Research
Paul D Morris, Department of Infection, Immunity and Cardiovascular Disease
Suzanne Mason, Centre for Urgent and Emergency Care Research
Steven Wood, Sheffield Teaching Hospitals NHS Foundation Trust
Maria-Cruz Villa-Uriol, Department of Computer Science
What is this paper about?
Our paper proposes a technique that facilitates the analysis and visual exploration of temporal event data. We build a multilevel overview of large volumes of temporal event sequences, and we let users transform how much detail they want to see, either across clusters of sequences (vertical level-of-detail) or longitudinally (horizontal level-of-detail). Our technique uses machine learning and advanced visualisation to allow an interactive exploration of the data. The proposed technique has been implemented into a visualization system called Sequence Cluster Explorer (Sequen-C) that allows multilevel and detail-on-demand exploration through three coordinated views, and the inspection of data attributes at cluster, unique sequence, and individual sequence level.
Why is the research important and/or novel?
Digitalisation has brought great opportunities for data analysis. Temporal event data are continuously being recorded. Events are associated with a timestamp, which indicates when they occurred and for how long they lasted. The current challenge is developing novel techniques able to benefit from the data richness present in these datasets, and how to obtain valuable insights that go beyond basic summary statistics where anomalous temporal patterns are easily missed. Our work proposes a technique, able to manipulate these data in an interactive fashion and able to facilitate the identification of common and deviating patterns that could be of use for data stakeholders.
Anything else that you would like to highlight about the paper?
We have used Sequen-C to study various real-world datasets in the healthcare domain.
The paper highlights two case studies: CUREd and MIMIC-III. These case studies demonstrate how the technique can aid users to obtain a summary of common and deviating pathways, and explore data attributes for selected patterns. CUREd is a unique linked research database that collates routine NHS data from a number of Urgent and Emergency Care service providers in Yorkshire and the Humber region from 2011-2017 including ambulance services, calls to emergency services and visits to emergency departments. Using Sequen-C we were able to identify geographical regions with a significantly larger number of patient calls to emergency services, or identifying clusters of callers patients with repeated calls to emergency services.
MIMIC-III is a publicly available single-center database comprising information relating to patients admitted to critical care units at a large tertiary care hospital in Boston, Massachussets. In our paper we showed how Sequen-C could be used to study hospitalised patients suffering from atrial fibrillation.
This work was co-developed with Sheffield Teaching Hospitals NHS Foundation Trust and the Centre for Urgent and Emergency Care Research at The University of Sheffield. The Health Foundation partially funded some of the work via the project "PathAnalyse: Towards the redesign of outpatient services using visual process analytics".
The University’s four flagship institutes bring together our key strengths to tackle global issues, turning interdisciplinary and translational research into real-world solutions.