=Paper=
{{Paper
|id=None
|storemode=property
|title=Using CDISC ODM and the RDF Data Cube for the Semantic Enrichment of Longitudinal Clinical Trial Data
|pdfUrl=https://ceur-ws.org/Vol-952/paper_7.pdf
|volume=Vol-952
|dblpUrl=https://dblp.org/rec/conf/swat4ls/LerouxL12
}}
==Using CDISC ODM and the RDF Data Cube for the Semantic Enrichment of Longitudinal Clinical Trial Data==
https://ceur-ws.org/Vol-952/paper_7.pdf
Using CDISC ODM and the RDF Data Cube for the
Semantic Enrichment of Longitudinal Clinical Trial Data
Hugo Leroux1 and Laurent Lefort2
1
The Australian E-Health Research Centre, CSIRO, Brisbane, Queensland, Australia
2
CSIRO ICT Centre, Canberra, ACT, Australia
{firstname.lastname}@csiro.au
Abstract. The development of ontology, linked data standards and tools for
semantic enrichment, opens new opportunities to analyse and reuse the clinical
data collected as part of clinical trials and longitudinal studies. This paper pre-
sents our approach on the semantic enrichment of the data collected as part of
the Australian, Imaging, Biomarker and Lifestyle study of Ageing (AIBL).
AIBL is a large scale longitudinal clinical study into neurodegenerative diseases
that has been designed to support investigations of the predictive utility of vari-
ous biomarkers, cognitive parameters and lifestyle factors as indicators of Alz-
heimer’s disease.
The objective of this paper is to highlight the complementarities of Clinical
Data Management Systems standards, such as CDISC ODM, with novel ap-
proaches to manage large volume of heterogeneous linked data resources, such
as the W3C RDF Data Cube. We start by describing the standards, ontologies,
linked data resources and tools that we use to aggregate the study data. Next,
we describe the structure of the Linked Clinical Data Cube and the tools that we
use to map the recorded medication intake information to the relevant standards
in the Australian context: SNOMED CT and the Australian Medical Terminol-
ogy. We also discuss how our approach could be extended to take advantage of
past and present Linked Open Data initiatives in the Health Care and Life Sci-
ences community.
Keywords: Ontology, Semantic enrichment, Clinical Trial, Longitudinal Study,
Medication data, Data Cube
1 Introduction
The implementation of clinical trial management systems using relational databases
has expedited the dissemination of clinical trial data among collaborators and pro-
vided the potential to swiftly query the data repository to extract information. Existing
Clinical Data Management Systems (CDMS) software use a generic electronic case
report form (CRF) data structure to adapt to multiple clinical trials, and are imple-
mented in relational database management systems through a set of monolithic tables.
SWAT4LS-2012, November 28-30, Paris, France.
© The authors
Some CDMS products such as OpenClinica1 use a data dictionary, derived from the
clinical study protocol, to handle the variable fields of these generic tables. These
products generally support the Clinical Data Interchange Standards Consortium Op-
erational Data Model (CDISC ODM) export file format [1], an XML-based standard
that specifies how to include this lightweight study metamodel as metadata alongside
the study data.
The RDF Data Cube vocabulary [2], developed by the W3C Government Linked Data
working group, is a vocabulary for the publication of statistical data in RDF. The
specification defines a generic observation data structure that matches the CRF data
structure used in clinical data management systems.
In this work, we present how we can apply the RDF Data Cube specification to se-
mantically enrich longitudinal clinical study data to allow users to query the clinical
trial data more effectively and efficiently. We describe how the proposed Linked
Clinical Data Cube can support the various case report forms derived from the study
protocol in a uniform manner.
The approach proposed in [3] is evaluated on clinical data obtained as part of the Aus-
tralian, Imaging, Biomarker and Lifestyle study of Ageing (AIBL) data. AIBL [4] is a
large scale longitudinal clinical study into neurodegenerative diseases that has been
designed to support investigations of the predictive utility of various biomarkers,
cognitive parameters and lifestyle factors as indicators of Alzheimer’s disease (AD)
with a cohort of over one thousand participants residing in two Australian cities, Perth
and Melbourne. Each recruited participant completed blood and neurological testing
and some underwent brain imaging testing. To fulfil the longitudinal nature of the
study, each participant undergoes a re-examination every eighteen months. The par-
ticipants also volunteered a broad range of lifestyle and medical information, includ-
ing medication information that is currently obtained from participants in paper form
and then manually entered into the OpenClinica CDMS by study staff [5].
For the enrichment of medication data, we will use the Australian Medicines Termi-
nology (AMT) and SNOMED CT-AU [6]. However, among the difficulties associ-
ated with collating self-reported medication usage is the potential for misidentifica-
tion of the correct medication being recorded along with inconsistencies and lack of
precision relating to crucial information, such as dosage and frequency of use. We
show here how we can extend our Linked Clinical Data Cube to reuse links between
medication resources [7] and in particular reuse the ones which are already available
as linked data [8-9] or which share common identification keys [10].
This paper has four main sections: in section 2, we introduce and compare the CDISC
ODM and of the RDF Data Cube specifications. In section 3, we describe the generic
structure of the Linked Clinical Data Cube. In section 4, we describe in more details
how we can convert the recorded medication intake information recorded for the
AIBL study and map it to the relevant standards in the Australian context: SNOMED
CT and the Australian Medical Terminology. Finally, in section 5, we discuss how
our approach could be extended to take advantage of past and present Linked Open
Data initiatives in the Health Care and Life Sciences community.
1
https://openclinica.com/
2 Comparison of CDISC ODM and RDF Data Cube
2.1 CDISC ODM (with OpenClinica extensions)
The AIBL study data was migrated from a proprietary CDMS to the OpenClinica
CDMS a year ago [5]. One of the compelling factors in choosing OpenClinica was
that it adopts the CDISC ODM structure to define the logical organisation of the study
metamodel and as the basis for the standard export format for the clinical trial data.
The CDISC ODM standard [1] defines a format that facilitates the sharing of clinical
data and metadata from multiple sources. Furthermore, it assumes that the clinical
study follows a predefined structure that defines the study, subject, events, forms,
item groups and items. An abstracted version of the CDISC ODM structure that illus-
trates the regions of interest for our project is shown in Fig. 1.
Fig. 1. The OpenClinica CDISC ODM data model (container and descriptor classes)
The CDISC ODM data model is specifically designed for a data capture context:
An item is a single measurement or analysis result collected during a study e.g. a
blood pressure reading,
An item group is a set of related measurements or analysis results,
A form (or case report form) is a collection of items and item groups for capturing
and displaying clinical trial data,
A study event corresponds to a patient visit or other encounter where the data cor-
responding to one or multiple forms is collected.
Furthermore, the definitions of these data structures are also included in the CDISC
ODM export format. In OpenClinica, this data dictionary can be edited by super users
with the help of an Excel template that plays the role of a configuration file. The users
of the tools are encouraged to share their CRFs and have access to a library of peer
reviewed ones derived from authoritative standards sources such as the CDISC Clini-
cal Data Acquisition Standards Harmonization (CDASH) initiative.
2.2 The RDF Data Cube vocabulary
The RDF Data Cube vocabulary [2], published by the W3C Government Linked Data
working group, is a vocabulary for the publication of statistical data in RDF [11]. This
specification is available as a working draft, but it has been evaluated by a number of
government agencies (Eurostat, European and UK Environment agencies) who have
published large scale datasets. It has also triggered new work on the Online Analytical
Processing (OLAP) of Linked Data sources ([12-13]).
The basic principles behind the design of the RDF Data Cube vocabulary are illus-
trated in Figure 2.
Fig. 2. Cube (Dataset), Slice and Observation
A cube is a dataset that is divided into slices according to several dimensions. Each
slice contains a number of observations. The arrows in Fig. 2 represent the links be-
tween the cube and the slices and between the slices and the observations. These extra
links at multiple levels of data aggregation allow the data consumers to navigate and
query linked data. The RDF Data Cube vocabulary defines three types of data items:
dimensions for the identification keys, measures and attributes for the recorded data
and metadata. The slices group subsets of observations within a dataset where all the
dimensions except one (or a small number) are fixed.
The RDF Data Cube vocabulary specifies container classes and descriptor classes and
the set of properties to link them (Fig. 3). The main descriptor class, has a link to
properties classes (qb:componentProperty) to specify the data items which are
used. The other classes (qb:DataStructureDefinition, qb:SliceKey,
qb:ComponentSpecification) and properties (qb:structure,
qb:sliceStructure, qb:componentAttachment) are used to indicate what
properties are used at what level of aggregation (qb:Observation, qb:Slice
and qb:DataSet).
Fig. 3. The RDF Data Cube data model (container and descriptor classes)
2.3 Comparison of data slicing approaches
Both approaches bundle together the data (container classes) and metadata (descriptor
classes) and support the registration of code lists. The CDISC ODM format supports
only one way of slicing the data with five primary dimensions (Subject-Study event-
Form-Item group-Item). The RDF Data Cube structure is more flexible for later pub-
lication of the collected data where the choice and ordering of the dimensions depend
largely on the queries that we wish to make from the system.
In the context of the AIBL study, as it is longitudinal in nature, we are particularly
interested in analysing the different variables from an Event perspective; for example
to determine the rate of change of a particular observation across different time
points. On the other hand, we are also interested in analysing variables at an Item
level. One such example would be to track the concentration of a particular protein in
the blood together with the cortical thickness of a portion of the brain.
3 Structure of the Linked Clinical Data Cube
3.1 Overview
We use a Linked Clinical Data Cube template that answers our needs for analysing
the AIBL study from an Event and an Item perspective. The primary dimensions for
the CDISC ODM data model (Fig. 1) are the subject (or patient) identification and the
study event (or time point) identification. The other dimensions (Form – Item group
and Item) are domain-dependent and are specified by the data dictionary content. To
provide access to the recorded data at both levels, we have designed a Linked Clinical
Data Cube based on two nested RDF Data Cubes as shown in Fig. 4.
Fig. 4. Event and Item data cubes
This Linked Clinical Data Cube comprises two nested data cubes that depict the inter-
connectedness between an Event Cube, an Event Slice, an Item Cube and an Item
Slice. Our top-level data cube manages slices of Study Event Observations
that contain observations that are collection of Form (CRF) data. Our bottom-level
data cube manages slices of Item Observations that contain observations that
are either Form data or item data. To minimize the duplication of data, the Event
Observations contains links to the low-level Item Observations containing the meas-
ures and attributes rather than their values. This provides us with a high level of flexi-
bility to analyse observations at both an Event and Item level concurrently.
3.2 Coupling with domain ontologies
The RDF Data Cube vocabulary (QB) has been coupled to domain ontologies for the
publication of long term climate data time series as linked data [14] with the help of
the W3C Semantic Sensor Network ontology 2 [15]. For the Linked Clinical Data
Cube, we will need different ontologies for each item data cube corresponding to a
different domain of application.
Users of CDISC-compliant tools are encouraged to use standard Case Report Forms
(CRFs) to directly comply with other CDISC standards such as the CDISC Clinical
Data Acquisition Standards Harmonization [16] and the CDISC Study Data Tabula-
tion Model [17] specifications. SDTM and CDASH information can also be added to
2
http://purl.oclc.org/NET/ssnx/ssn
CDISC ODM content as annotations (using the Alias element) at all levels of defini-
tions (for all the descriptor classes shown in Figure 1).
Fig. 5 illustrates how we can potentially integrate a lightweight “skeleton ontology”
based on these CDISC standards with ontology modules from the RDF Data Cube
vocabulary (classes with the qb prefix) and SSN ontology (classes with the ssn pre-
fix) to construct our Linked Clinical Data Cube.
Fig. 5. Coupling the AIBL Linked Clinical Data Cube with CDISC-mappable ontologies
Fig. 5 shows the ontology modules, their classes and relationships (plain lines are
used for sub-class-of relationships and dashed lines for object properties linking
classes). The different colours used in Fig. 5 indicate which classes from which on-
tology modules should be coupled together (e.g. via multiple inheritance relation-
ships):
qb:Slice, ssn:Observation and TrialTable (Interventions,
Findings and Visits),
qb:Observation, ssn:Observation, InterventionsObservations
and FindingObservations,
qb:ComponentProperty, ssn:Property and Variable,
ssn:FeatureOfInterest, Subject and LocationOfMeasurement,
ssn:Platform and Site,
ssn:Deployment and Visit.
4 Application to the AIBL medication data
4.1 Data collected for the AIBL study
The AIBL study has been collecting medications information in part to monitor the
effects of some pharmaceuticals that could affect cognitive function. Information
relating to the medications intake of each participant was recorded on a questionnaire,
in paper form, and manually entered, by study staff, in OpenClinica. A sample of the
recorded information including the medication’s name, prescribed dosage, frequency
and duration of use is shown in Table 1.
Table 1. The medication information as recorded in OpenClinica
Subject Study Item Medication name Dosage Frequency Length of
id Event id group id time taken
4 3 3 Cartia 3 years
26 3 1 Arthro-aid (glucosa- 750mg 1 bd
mine hydrochloride)
The goal is to map this medication information to taxonomy of medication codes in
order to provide a hierarchical classification of the drugs. One significant challenge
linked to this task is in the identification of the correct medication given the inaccu-
racy due to inconsistency with the naming and imprecision regarding the dosage,
frequency and duration of use. In the case of the medication’s name, a mix of Trade
Name, Active Ingredients and informal name have been used to describe the pre-
scribed medication. Furthermore, the participants have omitted to record several fields
including the prescribed dosage, frequency and duration of use when filling in their
questionnaires. In the next section, we describe our approach to mapping the medica-
tion information to two Australian standards for medication terminology: AMT and
SNOMED CT-AU.
4.2 Mapping to SNOMED and AMT
Our choice of AMT and SNOMED CT is based on their complementarities. AMT
provides unique codes and accurate standardised names to unambiguously identify all
commonly used medicines in Australia with eight key top-level concepts [18] includ-
ing Trade Product. SNOMED CT organises content into several hierarchies, including
the Substance, Clinical finding, Body structure and Observable entity hierarchy and
its foundation in Description Logic makes it a good candidate to decomposing the
complex medications concept hierarchy and describing our domain ontology.
The processing pipeline [6] for mapping the medication information is shown in Fig.
6 and summarised below.
Fig. 6. Processing pipeline for mapping the medications data (extract from [6])
The medication records are extracted from OpenClinica at the start of the pipeline. A
data cleansing process is conducted to manually amend the inconsistencies, described
in the previous section, from these records. This is followed by two mapping phases.
In Phase 1, the system attempts a match of the medication name to an AMT concept
below the Trade Product hierarchy. The search operation returns zero or more candi-
date mappings. If more than one concept is returned, the strategy adopted to match the
AIBL medication to an AMT concept is to calculate the Least Common Ancestor
(LCA) [6]. During Phase 2, for every medication name not adequately identified in
Phase 1, the system attempts a match to a SNOMED CT-AU Substance Identifier.
The use of the Substance hierarchy is designed to broaden the search in an attempt to
address the more obscure medication name not identified in Phase 1 [6].
4.3 Handling AIBL medication records in the Linked Clinical Data Cube
The Medication Data Cube is an instance of the Item data cube described in Fig. 4. Its
primary dimensions are the subject id and study event id. The originally available
dimension for the Medication reference is the Medication name. The AMT and
SNOMED-CT identifiers can be used as alternative dimensions when available as
described in Fig. 7. The name, dosage, frequency and duration of use are available as
measures or attributes.
Fig. 7. Data cube dimensions for the AIBL Medication Data
Fig. 8 extends the discussion from section 3.2 by illustrating how the references from
the SNOMED CT and AMT ontologies augment the skeleton ontology depicted in
Fig. 5. Linking to AMT and SNOMED CT concepts provide the possibility to obtain
additional information based on links between the concepts or trade products branches
and other branches in the AMT and SNOMED CT ontologies. We will also exploit
the mappings at the substance level between these two resources as defined in [18].
Fig. 8. Medication observation reference to SNOMED or AMT ontologies
We also intend to use the DrugBank 3 , 4 database and the Anatomical Therapeutic
Chemical (ATC) and Defined Daily Dose (DDD) taxonomy5 defined by the WHO
Collaborating Centre for Drug Statistics Methodology to supplement the medication
data as depicted below in Fig. 9.
Fig. 9. Additional dimensions resulting from the linking to ATC DDD
3
http://www.drugbank.ca/
4
A RDF version of DrugBank is available from http://linkedlifedata.com/.
5
http://www.whocc.no/atc_ddd_index/
5 Discussion
5.1 Benefits of the approach
There are several challenges associated with mapping clinical trial concepts to estab-
lished ontologies and linked data resources to enrich clinical data such as the AIBL
study data [3]. We propose a three-tiered approach which helps to answer some of
these challenges.
The first tier applies the Data Cube principles to overcome the monolithic nature of
the CDISC ODM file structure. This is the approach illustrated by Fig. 4 which ex-
poses the clinical data across multiple dimensions.
The second tier involves the semantic enrichment of the AIBL data using references
from the curated medication classification obtained by mapping the medication data
to AMT and SNOMED CT. This is outlined in section 4.3 and illustrated by Fig. 7.
This process has the potential to further expose the clinical data across the additional
dimensions.
The third tier relates to the linkage of the clinical data to other resources, namely the
ATC DDD, DrugBank and all the other linked data resources that possess references
to them. This approach is depicted in Fig. 9 and provides the opportunity to introduce
yet supplementary dimensions through which to expose the data. For the users of the
AIBL data published as linked data, the benefits of our approach are tied to the extra
information provided by the linked resources as adding links to DrugBank and ATC
DDD create new opportunities to query the data.
DrugBank also defines drug and food interactions. The former provides an important
step in the exploration of drug-drug interactions that also provide some insight into
potential risks and contraindications associated with the intake of the medication. The
latter will be useful when we explore the association between the participant’s drug
intake and type and quantity of food consumed. DrugBank also provides information
on the gene-drug interactions medication target which could expedite the discovery of
biomarkers.
The five levels of taxonomy of medications code provided by ATC DDD (Fig. 9) also
provide means to aggregate the study data for statistical purposes. This is complemen-
tary to what would be possible with the help of the taxonomies supplied by AMT and
SNOMED-CT.
5.2 Future work
The Linked Clinical Data Cube will not reach its true potential unless it is coupled
with multiple domain ontologies to enrich its referencing capabilities. The work
within the AIBL Linked Clinical Data Cube will be to organise and logically map the
logical information contained within the various CRFs to domain ontologies (Fig. 1 in
[5]). We plan, in the short term, to conduct a survey of existing domain ontologies,
from the literature, to identify suitable candidates that adequately define the semantics
of the test data comprising the study.
As a first step, we will need to identify the primary dimensions and the set of identifi-
able classes that define the Linked Clinical Data Cube. For the first tier, we need a
modular ontology that covers the definitions introduced by CDISC standards, in par-
ticular CDASH [16]. As shown in Fig. 5, the skeleton of this ontology can reuse a
good subset of the Semantic Sensor Network ontology but it should also define key
CDASH classes such as Intervention, Findings, Visit and Subject. We
also need additional modules for each type of CRF defined for the AIBL study data
[5]. One of the roadblocks to this task is the need to release RDF versions of the
CDISC CDASH [16] and STDM [17] standards in sync with versions of these stan-
dards used in the tools. To ease the conversion from CDISC ODM to RDF and en-
courage developers of new CRFs to map their definitions to a common reference, the
reusable CRF templates supplied by the CDISC consortium should also include anno-
tations pointing to CDASH definitions published as RDF. There have been several
attempts by the CDISC Consortium to develop an RDF version of these two standards
but these have, as yet, not been completed.
The second step entails producing more complete mapping tables between our con-
cepts and those defined in linkable resources on the web, in particular AMT and
SNOMED CT. There are opportunities to improve the semi-automated mapping algo-
rithm implemented for AMT and SNOMED CT with the help of other medication
resources e.g. DrugBank, NDF-RT 6 and RxNorm 7 . Schulz [19] identifies various
shortcomings within SNOMED CT in relation to completeness and consistency.
5.3 Related work
Many researchers have developed approaches to facilitate the semantic enrichment of
biomedical research data. Some of these approaches [20] have focussed on integrating
the clinical data with ontologies while other approaches [21] have investigated the use
of linked data resources. However, little effort has been directed at combining these
two complementary approaches.
Some of the ontologies developed in the context of translational research [22] and
clinical trials [23-25] are partially applicable to our needs. But they do not adequately
cover the observation aspects that are required for our data cube. Several of these
ontologies also have a large number of dependencies to other ontologies.
The Linked Open Drug Data8 (LODD) and the Linked Life Data (LLD) projects pro-
vide additional resources that can be used to extend the AIBL Linked Clinical Data
Cube. Both projects aim to build a large scale knowledge cloud that can be used for
drug discovery. LODD [8] federates the efforts by participants of the W3C Health and
Life Sciences (HCLS) Interest group9 to convert available resources into linked data.
LLD [9] provides a semantic data integration platform for the biomedical domain
comprising many of the data sources belonging to LODD plus some new ones. The
6
http://evs.nci.nih.gov/ftp1/NDF-RT/
7
https://www.nlm.nih.gov/research/umls/rxnorm/
8
http://www.w3.org/wiki/HCLSIG/LODD/Data
9
http://www.w3.org/wiki/HCLSIG
resulting datasets contains more than 8 million triples representing the knowledge
within over 2 millions links relating to medications, diseases, clinical trials, gene in-
formation and pharmaceutical companies among others.
Among the various use cases reported via the W3C HCLS Interest group are efforts to
explore links to identify and verify genes linked to Alzheimer’s disease (AD).
Through the links between the drug, medications, disease and clinical trial reposito-
ries, we hope to leverage on efforts by others to further explore the effects of pre-
scribed medications, for AD sufferers, on the various genes comprising the pathways
of interest. Other applications of LODD include the identification of potential side-
effects linked to the intake of drugs that have conflicting stimuli on the disease path-
ways.
The SALUS project [26] is a former attempt to adapt CDISC standards to build a
Semantic Framework to improve interoperability between clinical research and clini-
cal care domains. We adopt a similar approach to them but their focus is on service
mappings rather than linked data sets. The Semantic Cockpit [27] project aims to
develop a data slicing framework comparable to what we propose on the basis of the
RDF Data Cube. The goal of this project is to intelligently assist business analysts by
discriminating unimportant information and using reasoning to only present useful
information to the analyst.
6 Conclusions
Several new opportunities exist to analyse and reuse the clinical data gathered as part
of clinical trials through the development of ontology, linked data standards and tools
to semantically enrich this data. We have presented an approach for the semantic
enrichment of clinical trial data obtained as part of the AIBL study, a large-scale lon-
gitudinal study into neurodegenerative diseases. We have outlined the design of the
Linked Clinical Data Cube. The Linked Clinical Data Cube takes advantage of the
strength of the RDF Data Cube in defining the slices, dimensions and observations
within the data and applying them to the CDISC ODM data model to provide in-
creased flexibility in the formulation of queries and allow the users to query the clini-
cal data more effectively and efficiently. We have also outlined the use of the AMT
and SNOMED CT-AU taxonomies to enrich the medication data. Finally, we have
presented our method to extend our Linked Clinical Data Cube to reuse links between
medication resources, in particular the ones that are already available as linked open
data. The main contribution of our approach is that we propose the use of ontologies
and linked data resources together to semantically enrich the clinical data, thanks to
the cohabitation of the container and description classes in our solution. Our strength
is in the potential for semantic enrichment of any CDMS tools that adopts the CDISC
standard.
Acknowledgements. The authors would like to express their gratitude to Simon
McBride, Simon Gibson and Dr Michael Lawley for their assistance in scoping the
Medications case study and to Drs Alejandro Metke and Kerry Taylor for their feed-
back on the paper.
References
1. CDISC: Clinical Data Interchange Standards Consortium - Operational Data Model (2011)
http://www.cdisc.org/odm
2. Cyganiak, R., Reynolds, D., Tennison, J.: The RDF Data Cube Vocabulary, W3C Working
Draft 05 April 2012. World Wide Web Consortium (2012)
3. Leroux, H., McBride, S., Lefort, L., Kemp, M., Gibson, S.: A method for the semantic en-
richment of clinical trial data. Stud Health Technol Inform, 178, pp. 111-116 (2012)
4. Ellis, K. A., Bush, A. I., Darby, D., De Fazio, D., Foster, J., Hudson, P., Lautenschlager,
N. T., Lenzo, N., Martins, R. N., Maruff, P., Masters, C., Milner, A., Pike, K., Rowe, C.,
Savage, G., Szoeke, C., Taddei, K., Villemagne, V., Woodward, M. and Ames, D.: The
Australian Imaging, Biomarkers and Lifestyle (AIBL) study of aging: methodology and
baseline characteristics of 1112 individuals recruited for a longitudinal study of Alz-
heimer's disease. Int. Psychogeriatry, 21(4), pp. 672-687 (2009)
5. Leroux, H., McBride, S., Gibson, S.: On selecting a clinical trial management system for
large scale, multi-centre, multi-modal clinical research study. Studies in health technology
and informatics, 168, pp. 89-95 (2011)
6. McBride, S., Lawley, M., Leroux, H., Gibson, S.: Using Australian Medicines Terminol-
ogy (AMT) and SNOMED CT-AU to better support clinical research, Studies in Health
Technology and Informatics, 178, pp. 144-149 (2012)
7. Sharp, M., Bodenreider, O., Wacholder, N.: A Framework for Characterizing Drug Infor-
mation Sources. AMIA 2008 Symposium Proceedings (2008)
8. Samwald, M., Jentzsch, A., Bouton, C., Stie Kallesøe, C., Willighagen, E., Hajagos, J.,
Marshall, M. S., Prud'hommeaux, E., Hassenzadeh, O., Pichler, E.:
Linked open drug data for pharmaceutical research and development.
Journal of Cheminformatics, 3(1), pp. 19- (2011)
9. Williams, A. J.; Harland, L.; Groth, P.; Pettifer, S.; Chichester, C.; Willighagen, E. L.;
Evelo, C. T.; Blomberg, N.; Ecker, G.; Goble, C. & Mons, B.: Open PHACTS: semantic
interoperability for drug discovery. Drug Discovery Today (2012)
10. Saitwal, H. and Qing, D. and Jones, S. and Bernstam, E. and Chute, C.G. and Johnson,
T.R.: Cross-terminology mapping challenges: A demonstration using medication. Journal
of Biomed. Inform. 45, pp. 613-625 (2012)
11. Cyganiak, R., Hausenblas, M., McCuirc, E.: Official Statistics and the Practice of Data Fi-
delity. In: Wood, D. ed.: Linking Government Data. Springer, pp. 135-151 (2011)
12. Kampgen, B., O'Riain, S., Harth, A.: Interacting with Statistical Linked Data via OLAP
Operations. In: International Workshop on Linked APIs for the Semantic Web (LAPIS
2012), (2012) http://lapis2012.linkedservices.org/
13. Etcheverry, L., Vaisman, A.A.: QB4OLAP: A Vocabulary for OLAP Cubes on the Seman-
tic Web. In: Third International Workshop on Consuming Linked Data (COLD 2012),
CEUR Workshop proceeding, vol. 905, CEUR-WS.org (2012),
14. Lefort, L., Bobruk, J., Haller, A., Taylor, K. and Woolf, A.: A Linked Sensor Data Cube
for a 100 year homogenized daily temperature dataset. In: 5th International Workshop on
Semantic Sensor Networks (SSN-2012), CEUR-Proceedings, vol. 904, CEUR-WS.org
(2012).
15. Compton, M., Barnaghi, P., Bermudez, L., Garcia-Castro, R., Corcho, O., Cox, S., Gray-
beal, J., Hauswirth, M., Henson, C., Herzog, A., Huang, V., Janowicz, K., Kelsey, W. D.,
Phuoc, D. L., Lefort, L., Leggieri, M., Neuhaus, H., Nikolov, A., Page, K., Passant, A.,
Sheth, A., Taylor, K.: The SSN ontology of the W3C semantic sensor network incubator
group. Web Semantics: Science, Services and Agents on the World Wide Web 15(3),
(2012)
16. CDISC CDASH Team: Clinical Data Acquisition Standards Harmonization (CDASH),
Version 1.1 Clinical Data Interchange Standards Consortium (2011)
http://www.cdisc.org/cdash
17. CDISC SDS Team: CDISC Study Data Tabulation Model, Version 1.2 Clinical Data Inter-
change Standards Consortium (2008) http://www.cdisc.org/stdm
18. Michel, J., Lawley, M. J., Chu, A., Barned, J.: Mapping the Queensland Health iPharmacy
Medication File to the Australian Medicines Terminology Using Snapper. Studies in
Health Technology and Informatics 168, pp. 104-116 (2011)
19. Schulz, S., Suntisrivaraporn, B., Baader, F., Boeker, M.:
SNOMED reaching its adolescence: Ontologists' and logicians' health check.
Int. J. Med. Inform., 78, pp. S86-S94 (2008)
20. Dumontier, M., Villanueva-Rosales, N.: Towards pharmacogenomics knowledge discov-
ery with the semantic web. Brief Bioinform 10(2), pp. 153-163 (2009)
21. Marshall, M. S., Boyce, R., Deus, H. F., Zhao, J., Willighagen, E. L., Samwald, M.,
Pichler, E., Hajagos, J., Prudhommeaux, E., Stephens, S.: Emerging practices for mapping
and linking life sciences data using RDF: A case series. Web Semantics: Science, Services
and Agents on the World Wide Web 14(0), pp. 2-13 (2012)
22. Luciano, J., Andersson, B., Batchelor, C., Bodenreider, O., Clark, T., Denney, C., Do-
marew, C., Gambet, T., Harland, L., Jentzsch, A., Kashyap, V., Kos, P., Kozlovsky, J.,
Lebo, T., Marshall, S., McCusker, J., McGuinness, D., Ogbuji, C., Pichler, E., Powers, R.,
Prud'hommeaux, E., Samwald, M., Schriml, L., Tonellato, P., Whetzel, P., Zhao, J.,
Stephens, S., Dumontier, M.: The Translational Medicine Ontology and Knowledge Base:
driving personalized medicine by bridging the gap between bench and bedside. J. Biomed.
Semantics, 2 (Suppl 2), (2011)
23. Sim, I., Carini, S., Tu, S., Wynden, R., Pollock, B., Mollah, S., Gabriel, D., Hagler, H.,
Scheuermann, R., Lehmann, H., Wittkowski, K., Nahm, M., Bakken, S.:
The human studies database project: federating human studies design data using the ontol-
ogy of clinical research. In: AMIA Summits Transl. Sci. Proc. 2010, pp. 51-55 (2010)
24. Ogbuji, C.: A Framework Ontology for Computer-Based Patient Record Systems. In: 2nd
Int. Conf. on Biomedical Ontology (ICBO-2011), CEUR-Proceedings, vol. 833, CEUR-
WS.org (2011)
25. Kong, Y., Dahlke, C., Xiang, Q., Qian, Y., Karp, D., Scheuermann, R.: Toward an ontol-
ogy-based framework for clinical research databases. J. Biomed. Inform. 44(1), pp. 48-58
(2011)
26. Laleci, G., Yuksel, M., Dogac, A.: Providing Semantic Interoperability between Clinical
Care and Clinical Research Domains. IEEE trans. On Information Technology in Bio-
medicine, vol. PP(99), to appear (2012)
27. Neumayr, B., Schrefl, M., Linner, K.: Semantic Cockpit: An Ontology-Driven, Interactive
Business Intelligence Tool for Comparative Data Analysis. In Olga De Troyer, Claudia
Bauzer Medeiros, Roland Billen, Pierre Hallot, Alkis Simitsis, Hans Van Mingroot, ed.:
Advances in Conceptual Modeling. Recent Developments and New Directions. Springer
Berlin / Heidelberg, pp. 55-64 (2011)