VACCINATED — Visual analytics for characterizing a pandemic spread VAST 2010 Mini Challenge 2 award: Support for future detection

摘要

Given a set of hospital admittance and death records, the challenge was to characterize the spread of a pandemic in terms of the attack and mortality rates, spatiotemporal patterns of onset and the recovery time. We began the analysis by preprocessing the hospital admittance records using the University of Pittsburgh''s CoCo classifier [1]. CoCo is a text classification software that takes hospital admittance fields and classifies them into chief complaint categories (Botulinic, Constitutional, Gastrointestinal, Hemorrhagic, Neurological, Rash, Respiratory, and Other). The choice of the CoCo classifier was based on its online availability as well as its well documented classification performance metrics, see [1]. Once the data was classified, we utilized and extended work done by the Purdue University Visual Analytics Center on healthcare analysis [2]. Our system consists of a combination of linked views, showing time series views of syndromes and death rates through line graph views (Figure 1 — Top), stacked graph views showing deaths (Figure 1 — Bottom), geographical map views showing the impact by country (not illustrated in this paper), and summary windows providing statistical breakdowns of the data (not illustrated in this paper). All views are linked through an interactive time slider that allows users to explore the data over time. Extensions to our previous work [2] include the stacked graph view, summary windows, new control chart methods, and an interactive ‘tape measure’ tool.