=Paper=
{{Paper
|id=Vol-1149/bd2014_rogers
|storemode=property
|title=None
|pdfUrl=https://ceur-ws.org/Vol-1149/bd2014_rogers.pdf
|volume=Vol-1149
}}
==None==
https://ceur-ws.org/Vol-1149/bd2014_rogers.pdf
ABSTRACTS : scientific
Spatio-temporal visualisation of disease
incidence and respective intervention
strategies
Stefan Von Cavallara, Matthew Davisa, Kelly L. Wyresa, Matthias Reumannb, Martin J. Sepulvedac,
Priscilla Rogersa
a
IBM Research – Australia, 204 Lygon Street, Carlton, VIC 3053, Australia
b
IBM Research – Zurich, Saeumerstrasse 4, 8803 Rueschlikon, Switzerland
c
IBM Research – Watson, 1101 Kitchawan Road, Yorktown Heights, NY, 10598 USA
SUMMARY
The ability to effectively collect and leverage information offers rich new insights, which can greatly inform
decision making within healthcare practice and health systems. In this study, we show how information can be
Dr Priscilla Rogers collected, analysed and presented in new ways to inform key decision makers in understanding the prevalence
Manager of disease and the response to interventions. We demonstrate this for malaria, using data sources from
IBM Research Australia Kenya, where malaria is one of the three biggest causes of mortality for children under the age of 5 in the
sub-Saharan continent of Africa1.
INTRODUCTION
prrogers@au1.ibm.com Health research and “big data” have the potential to provide powerful new insights. Frameworks, which have
the ability to collect and store data in a secure manner, organise and prepare the data, conduct advanced
analytics, and display the data in a visually compelling way offer many possibilities to improve healthcare
planning and delivery. A pressing problem in developing nations is childhood mortality. Despite the availability
Priscilla Rogers is a Research Staff Member at IBM
of effective intervention strategies, pneumonia, diarrhoeal disease and malaria remain the top three causes
Research - Australia, and the manager of the Laboratory’s of mortality for children <5 years in Africa1. In this region there remains a lack of accurate disease incidence
Healthcare Research team. In this role, she is responsible and healthcare data with a spatiotemporal resolution appropriate for effective intervention planning and
for driving the research agenda in healthcare, and is implementation. In this study, we developed a web-based visualisation portal using select and simulated data
passionate about data-driven healthcare and pioneering sources that would allow relevant bodies, such as public health officials, to visualise the disease burden of a
technologies for transformative plays in the sector. Priscilla particular disease and the potential impact of available and relevant interventions.
studied Mechanical Engineering at Monash University, and
holds a PhD in acoustic microfluidics for lab-on-a-chip DESCRIPTION
device applications, also from Monash.
The Cognitive Healthcare and Health Systems Hub is a framework that provisions the ability to collect, store
and analyse data from many data providers in a secure manner. The visualisation portal allows relevant
bodies to obtain and visualise tailored insights into data within the Hub. This includes a portal to visualise the
disease burden of a particular disease and the potential impact of available and relevant interventions. The
portal employs a geospatial 3D environment in which to visualise data, enabling the user to view information
from a sub-county/town/house level through to a country wide overview. By further augmenting the disease
model data with information representing other entities, such as points of interest (i.e. healthcare clinics,
pharmacies etc) we can obtain a deeper insight into the disease and appropriate resources available to the
different regions.
At its most basic, the disease model data is visualised by utilising a combination of D32 and Cesium3. The D3
framework is used to process the visualisation data and transform it into a form ready for presentation. The
Cesium framework is used to present the visualisation data within a geospatial 3D environment, for example
displaying sub-county boundary lines, sub-county names, points of interest etc., right through to rendering the
level of red shading within each sub-county region to represent the estimated incidence of disease. Enhanced
user functionality is provided by the framework, allowing the user to both navigate the 3D environment and
clearly see how the disease burden changes with space and time.
The Spatiotemporal Epidemiological Modeler (STEM), an open source tool to simulate models of disease, was
used to generate data for an epidemic malaria outbreak and associated public health interventions in Kenya4.
STEM contains a malaria disease model to simulate disease transmission by region using environmental and
earth science data sources, including elevation, temperature, precipitation, and vegetation coverage5,6.
38 #bd14 | big data conference
To model malaria in Kenya and integrate the intervention scenarios, we made several changes to the included models and data in STEM. First, we replaced the
default Kenya data with higher resolution political and earth science data sources, enabling simulations at sub-county resolution. Next we modified the program
to modulate disease model parameters by region. This allows us to target interventions for specific regions. For the “distribute bed nets” intervention, the malaria
model’s biting rate was modified; for “spray insecticide”, we constrain the mosquito population. To feed simulation data directly to the visualisation portal, we also
implemented a JSON data logger in STEM. Model parameters for the baseline and intervention scenarios were derived from literature.
The portal enables the user to select which scenario output to visualise. Such visualisation allows intuitive access to information essential for those interested in
understanding transmission dynamics or planning disease control strategies. Future versions could also link directly to the simulation engine, allowing scientists to
modify rate parameters real time to the simulation and actively test alternate hypotheses. Such comparisons will greatly inform public health policy-makers.
Figure 1. Visualisation portal displays simulated disease incidence in Kenya (no interventions are applied).
CONCLUSION
In this study, we simulate disease incidence data for various scenarios using a mathematical model for analysing and visualising the infectious disease spread in
human populations over time and over geographies. We have shown how leveraging information can be used to inform users on the prevalence of disease and the
ways in which the diseases are best treated and prevented.
REFERENCES
1. World Health Organization, Heslth Statistics and Informatics Department. Summary: Deaths by cause, in WHO Regions, estimates for 2008. 2011.
2. Cesium, Analytical Graphics Inc. http://cesiumjs.org/
3. D3, Data Driven Documents, Mike Bostock, http://d3js.org/
4. STEM, http://www.eclipse.org/stem
5. S. Edlund, M. Davis, J. Pieper, A. Kershenbaum, N. Waraporn, J. Kaufman, A global study of malaria climate susceptibility. Epidemics 3 (Boston, 2011).
6. G. Macdonald, The Epidemiology and Control of Malaria, London, Oxford University Press, 1957.
3 - 4 april 2014 | melbourne 39