Workflow issues for Health-mapping mashups

Full text

Workflow issues for Health mapping “mashups” of
OGC Web Services
Didier G. Leibovici *, Suchith Anand *, Jerry Swan *,
James Goulding *, Gobe Hobona *, Lucy Bastin #, Sergiusz Pawlowicz *,
Mike Jackson * and Richard James §
* Centre for Geospatial Science, University of Nottingham, UK # Aston University, Birmingham, UK
§Centre for Health Care Associated Infections, University of Nottingham, UK
ABSTRACT
This paper explores potential uses of OGC web services in the context of health mapping and
epidemiological studies. The situations or scenarios cover (i) different user perspectives, e.g. public
and community, research, health professionals and (ii) different interaction levels, e.g. simple data
“mashups” (overlay), use of web processing services (WPS) and participating GIS. Some particular
aspects of the domain such as privacy for the individual and computational flexibility/performance
raise issues in chaining web services. We propose some solutions at both the architecture and
geocomputational level, as well as modifications to the standards involved.
INTRODUCTION
Historical perspectives about health and epidemic studies using spatial information can be very
informative about the benefits these disciplines can get from Geographical Information System –
Science- (GIS). Using only pen and paper, combining cholera death and water pump locations, Dr
John Snow demonstrated, 150 years ago, that spatial studies could be of great value in understanding
and predicting infectious disease outbreaks. Yesterday’s desktop GIS provided more computational
power, making the most of the visualisation display, spatial properties and statistical spatial analysis
methodologies.
The internet and web services can now transfer and share all these capabilities with everyone.
This evolution concerned three major methodological areas: (i) data acquisition, storage and
exchange, (ii) data analysis, conflation and geocomputation and (iii) user-driven/contextual
functionalities, i.e. for researchers, practitioners, decision makers and the general public. The
experience of the user and the demanding flexibility he/she expects from these three methodological
components are the drivers of geospatial science (Chang et al. 2009, Gao et al. 2008).
Web technologies and infrastructures, particularly the interoperability of web services, is just
beginning to address the challenges of new and future health GIS services for people’s well being,
epidemic monitoring, health and the environment, derived from spatial “mashups” and pertinent
spatial statistics. Recent developments in web 2.0, positioning technology, location aware pervasive
computing, as well as ubiquitous positioning, allowing the creation of contextual foot-prints or
“scent-trails” (path and activities) provide additional means towards achieving pertinent spatial
modelling. These new means can express their full potential only when embedded within
interoperable web services infrastructures.
In the International Journal of Health Geographics, Boulos et al., 2008, give, within a long series
of articles entitled “Web GIS in practice”, an interesting list of potential use of some modern
technologies, particularly addressing non IT specialists, including health practitioners. These authors
also use the metaphor “mashups” to explain the principle of overlaying of maps or weaving of web
services in order to serve, or conflate different information sources. These geospatial mashups can be
conceptually expressed and stored as workflows: the chaining of data access (WSF, WCS) and
processing transformations/algorithms/models (WPS), represented as a graph for example using

BPMN (see further). Instead of focusing on particular tools, this paper investigates the role and use of
the interoperability of web services via the standards from the Open Geospatial Consortium (OGC)
for the domain of public health and epidemiology. Some typical scenarios are described using OGC
web services (OGC, 2010), then particular important issues for the domain are explored with
solutions that may lead to modify or extend the specification of the standards involved.
HEALTH MAPPING SERVICES
In order to place emphasis on the interoperability for web services for geospatial use cases, the
standards from OGC and ISO organisations prevail. For any disease, a simple data conflation can be
conducted by using Web Map Services (WMS, see OGC, 2010), where an analysis could be, for
example, a visual appreciation of the relative density of MRSA cases (Methicillin Resistant
Staphilococcus Aureus, see Lowy, 2010) for a certain area (selection of a bounding box): overlay for
a given area of a density map of cases and a population density map. A more interesting query (as
suggested by figure 1) provide, for example, a distribution of cases by age and gender for a selected
hospital or area delineated. This can be performed using WFS servers (Web Feature Services, see
OGC, 2010). Within the domain of spatial epidemiology of MRSA (Grundmann et al. 2010), such
functionality can be seen on the website www.spatialepidemiology.net/SRL-Maps/maps/, which uses
Google maps API for the web client (using a WFS or not to access the GML files).
Figure 1: Hypothetical architecture using Web Feature Service to provide MRSA cases
information per General Practice and/or per hospital located in the requested map area.
Notice that we mentioned the query is done for a selected area, but this depends in fact on the
ability of the client to gather the attribute values of each selected geometry, and evaluate/display the
aggregated distribution. This can be either implemented as such within the client or the client calls a
WPS (Web Processing Service, see OGC 2010).
So an “ideal” health mapping client must be able to look for existing data services (WMS, WFS,
WCS), then query the one selected, either to build up a data mashup and then run a selected WPS to
provide a result from an analysis (a statistic, a statistical map), or, build up a more complex workflow
involving more than one WPS also harvested. One can already interpret the selection of the mashup
and the use of a WPS, as a workflow.

Apart from accessing and analysing data, web services is also used to perform surveys and collect
data, which is also an important part of health and epidemic studies (if not the most time-consuming
and expensive). Sensor Web Enablement (SWE, see OGC 2010) and location based services (LBS)
provided, for example, by a mobile phone enable relevant ways of conducting a population survey
and acquiring data. As seen in figure 2, an LBS using a mobile phone can deliver a public service in
refining a standard diagnostic questionnaire with local information regarding the risk of a specific
disease, and personal information coming may be from wearable devices (e.g., temperature) or from
medical history either stored on the phone or remotely accessed (may be encrypted).
Figure 2: Web2.0 and Participating GIS principles provided by mobile phone.
A whole survey can also use mobile devices and GPS tracking of consenting patients. The latter
acts as a LBS giving feedbacks to patients to get better assessment of their risks due to the
environment (including other known cases), and as a basis for simulation studies, besides the current
study of risk factors.
SOME ISSUES AND SOLUTIONS
Confidentiality and privacy are the first constraints that are often discussed in epidemiological
studies. Obviously if the data from the survey are accessible via the internet, this becomes even more
crucial. Data access must be controlled depending on the information requested and the user profile.
GeoXACML the extension of XACML from OASIS (XACML, 2005) and current developments of
GeoRM (OGC, 2010) enable definition and control mechanisms for a web service to be accessed.
In the previous paragraph we already mentioned the workflow functionality that is desirable not
only in the disciplinary domains discussed in this paper, but also in environmental modelling in a
broad sense. An important feature for monitoring situations is to be able, to store a temporary result
issued from chaining web services, to define and store its workflow along with its lineage. In public
health and epidemiological studies, the semantic of the queries stored in the metadata of the workflow
is of particular importance when for example, the progress of the knowledge about a disease leads to
refining of selection criteria. Some standards, such as BPMN (Business Process Model Notation,
BPMN 2009) from the OMG and XPDL (XML Process Definition Language, XPDL 2008) from the

WfMC or the more well known BPEL (Business Process Execution Language, BPEL 2007) are
discussed within an OGC Domain Working Group to approach the issues for geospatial workflow.
At the intersection of the two previous issues is the concern about obtaining the most pertinent (or
appropriate) results based on the available data at the finest (or appropriate) scale. To solve the
antagonism between privacy and accessing fine scale data, we propose extending the web services
standard in order to allow full access to data at computational level (from the WPS) but for the results
to be output at a coarser scale, (Figure 3.). The access right of the WMS, WFS or WCS contains a link
to an upscaling (generalisation) processing service. The scale is maintained within a chain of WPS
until the output from the last WPS. The WPS standard has also to be modified to allow the final
chaining to perform the upscaling or generalisation.
Figure 3: Principle of blind access to fine scale data for WPS preserving privacy.
At a geocomputational level, some heavy computing processing can diminish the performance; a
lot of statistical methodologies can be expressed within an updating paradigm where a current
statistical map or map associated with a current statistic is updated according to an incoming dataset.
The generic signature of such “updating” WPS then contains: datasets, current result, new datasets.
Some data assimilation methodologies proceed in the same way. Disease clustering is a very time
consuming geoprocessing statistical method (Kulldorff 1995, Lawson et al. 2006). Following this
“updating” approach we are modifying some spatial statistical algorithms exploring collocated events,
(Leibovici et al. 2008, 2009 and 2010), to allow better performance for the WPS encapsulating them.
These WPS are working with the R package (R Development Core Team, 2007) as back-end
(Williams et al. 2010), which we believe to provide added flexibility about implementing classical,
modified and new statistical (spatial) methods in epidemiology and public health.
DISCUSSION
This is a position paper describing some necessary developments of the OGC standard (WPS, and
workflow in development) in order to promote its use in the domain of public health and
epidemiology for different users. Even though geospatial “mashups” raise common issues for many
fields, some particular aspects for public health and epidemiology need attention. Confidentiality and
privacy are amongst the main sources of these issues. We described a solution regarding the
antagonism between obtaining pertinent fine scale results from a workflow and preserving the privacy

of citizens: the “blind access” principle. We also presented the principle of “updating WPS” to
increase performance at computational level.
Privacy regarding tracking capabilities was not fully discussed here, but in the first instance it
could appear to pose similar confidentiality principles as in traditional population surveys,
nonetheless the spatial components (“scent-trail” or contextual foot-prints) can contain even more
sensitive information. The principle of “blind access” is also be applicable for this kind of data.
A recent review on software for disease surveillance (Robertson and Nelson, 2010) pointed out
that flexibility and performance are very important features. The reviewed desktop applications but
mentioned the potential of web applications, beneficiating from increased computing power and
architecture; we believe the principles developed in the present paper to be effective in these aspects
for spatio-temporal data as well. Different experiments and scenarios implementing the above
principles for enhancement of the web service standards will be demonstrated at the AGILE
conference.
BIBLIOGRAPHY
BPMN 2009 Business Process Model Notation version 2 beta 1. OMG (Object Management Group),
http://bpmn.org/
BPEL 2007 Standard WS-BPEL 2.0. OASIS, http://docs.oasis-open.org/wsbpel/2.0/wsbpel-v2.0.pdf
Boulos M., Scotch M., Cheung K., and Burden D., 2008 Web GIS in practice VI: a demo playlist of
geo-mashups for public health neogeographers. International Journal of Health Geographics,
7(1), 38.
Chang, A., Parrales M., Jimenez J., Sobieszczyk M, Hammer S, Copenhaver D, et al., 2009.
Combining Google Earth and GIS mapping technologies in a dengue surveillance system for
developing countries. International Journal of Health Geographics, 8(1), 49.
Gao S., Mioc D., Anton F., Yi X., Coleman D., 2008 Online GIS services for mapping and sharing
disease information. International Journal of Health Geographics, 7(1), 8.
Grundmann H, Aanensen DM, van den Wijngaard CC, Spratt BG, Harmsen D, et al. 2010
Geographic Distribution of Staphylococcus aureus Causing Invasive Infections in Europe: A
Molecular-Epidemiological Analysis. PLoS Med 7(1): e1000215.
Kulldorff M., and Nagarwalla N., 1995 Spatial disease clusters: Detection and inference. Statistics in
Medicine, 14(8), 799-81
Lawson A., Gangnon R., and Wartenberg D., 2006 Developments in disease cluster detection.
Special Issue: Statistics in Medicine 25, (5)
Leibovici D.G., 2009 Spatio-temporal Multiway Decomposition using Principal Tensor Analysis on
k-modes: the R package PTAk. Journal of Statistical Software (accepted August 2009)
Leibovici D.G., Bastin L., and Jackson M., 2008 Discovering Spatially Multiway Collocations.
GISRUK Conference 2008, Manchester, UK.
Leibovici D.G., Bastin L., and Jackson M., 2009 Higher Order Cooccurrences in Point Pattern
Analysis and Decision Tree Clustering. Computers & Geosciences, (submitted)
Leibovici D.G., Bastin L., Anand S., Swan J., Hobona G., and Jackson M., 2010 Spatially Clustered
Associations in Health GIS. GISRUK Conference 2010, London, UK.

Lowy F.D., 2010 Mapping the Distribution of Invasive Staphylococcus aureus across Europe. PLoS
Med 7(1): e1000205
OGC 2010 OpenGIS® Standards and Related OGC documents. Open Geospatial Consortium:
website, http://www.opengeospatial.org/standards
R Development Core Team 2007 R: A Language and Environment for Statistical Comput-
ing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, URL
http://www.R-project.org.
Robertson, C and Nelson, T (2010) Review of software for space-time disease surveillance.
International Journal of Health Geographics, vol. 9, 2010, p. 16
XACML 2005 eXtensible Access Control Markup Language. Organization for the Advancement of
Structured Information Standards http://www.oasis-
open.org/committees/tc_home.php?wg_abbrev=xacml
XPDL 2008 XPDL 2.1: XML Process Definition Language version 2.1. WorkfLow Management
Coalition. http://www.wfmc.org/xpdl.html
Williams, M*, Cornford, D, Bastin, L, Jones, R and Parker, S (2010) Automatic processing, quality
assurance and serving of real-time weather data. Computers and Geosciences. (accepted)