=Paper=
{{Paper
|id=None
|storemode=property
|title=Knowledge representation in Health Research:
the modeling of Adverse Events Following Immunization
|pdfUrl=https://ceur-ws.org/Vol-897/ecs_1.pdf
|volume=Vol-897
|dblpUrl=https://dblp.org/rec/conf/icbo/Courtot12
}}
==Knowledge representation in Health Research:
the modeling of Adverse Events Following Immunization==
https://ceur-ws.org/Vol-897/ecs_1.pdf
Knowledge representation in Health Research: the
modeling of Adverse Events Following Immunization
Author Mélanie Courtot
Supervisor Ryan R. Brinkman
Collaborator Alan Ruttenberg
Studies/Stage Ph.D. Student, Year 2
Affiliation BC Cancer Agency, Terry Fox Laboratory, Vancouver,
Canada
E-Mail mcourtot@gmail.com
Aims and Objectives of the Research
The overall goal of my thesis is to improve safety signal detection in vaccine adverse events
reports within the Canadian healthcare system. A system that would provide an easy and
fast way to isolate reports of potential interest for human experts to focus their attention on
would significantly decrease time and resources currently needed for signal detection.
Justification for the Research Topic
Free-text reporting of Adverse Events Following Immunization (AEFIs) leads to
inaccurate and incomplete data. Accurate representation of adverse event is a crucial part
of clinical research: it may initiate further investigation of potential problems in vaccine safety
or efficacy, and facilitate subsequent dissemination of safety-related information to the
scientific community and the public [1,2]. However, current methods used for adverse events
reporting are not sufficient, mitigating their usefulness. There is no standardization of the
terminology used in the current Electronic Data Capture System used by Public Health
Agency of Canada – at best a Medical Dictionary of Regulatory Activities (MedDRA [3]) code
is assigned after parsing the clinician’s input, but this code is not linked to any definition.
Several studies highlight the potential issues in using MedDRA for adverse event reporting,
ranging from inaccurate reporting (as several terms are non-exact synonyms) to lack of
semantic grouping features impairing processing in pharmacovigilance [4-8]. Additionally,
only the final adverse event code as determined by the system is saved, and information
about sub-parts are lost, therefore restricting ability of the physician to go back to the set of
symptoms observed to establish the diagnostic, and limiting the ability to query the resulting
datasets.
Research Questions
Is it possible to automate the diagnosis of adverse events for safety signal detection?
(1) Can the logic of a clinical guideline be encoded as an ontology to allow for semantic querying, i.e.,
be complex enough to encode all logical aspects while maintaining reasoning capabilities?
(2) Can a mapping between the ontology and another resource used to annotate existing AE reports
datasets be established, and allow for ontologically-supported diagnosis to be inferred from the current
reports?
(3) Is it possible to apply known algorithms to classified datasets to detect statistically significant
patterns?
(4) Is the detection of those patterns more efficient in terms of time and cost than performed by human
review?
Research Methodology
Regarding vaccine adverse events, the Brighton Collaboration [9] has done extensive work
towards standardization. This global network of world-renowned experts aims to provide high
quality vaccine safety information. They create methodological standards for accurate risk
assessment, including standardized case definitions of AEFIs. While they do not provide a
causal assessment between a given adverse event and the immunization process, the case
definitions are designed to define the levels of diagnostic certainty of reported AEFIs.
Aim 1. To develop an ontology to logically represent Brighton definitions, resulting in
increased quality and accuracy of AEFIs reporting. In collaboration with the Brighton
Collaboration, I will create an application ontology, the Adverse Events Reporting Ontology
(AERO). AERO defines individual signs and symptoms textually. They will then be logically
defined by being positioned into a hierarchy, and linked between them and an overall
diagnosis.
Aim 2. Establish a mapping between MedDRA and the AERO. MedDRA - the Medical
Dictionary for Regulatory Activities - is currently being used to encode adverse events
reports into reporting systems such as the Canadian Adverse Events Following Immunization
Surveillance System (CAEFISS) or the Vaccine Adverse Event Reporting System (VAERS)
in the US. Each term in AERO will be linked to the corresponding term(s) in MedDRA.
Aim 3. Perform automatic case classification on structured datasets. Using the mapping
built in aim 2, we will be able to process automatically the existing MedDRA annotations on
the data, to infer if a Brighton criteria has been met or not. For example, the tool would be
able to suggest anaphylaxis diagnosis based on a set of MedDRA annotations.
Aim 4. Detect safety signals. Using the classified datasets, known statistical methods can
be applied to determine if the proportion of reported cases according to the Brighton is
statistically significant.
Research Results to Date
The AERO project is available at http://purl.obolibrary.org/obo/aero. AERO is listed on the
OBO library at http://obofoundry.org/cgi-bin/detail.cgi?id=AERO and under BioPortal at
http://bioportal.bioontology.org/visualize/45521. The initial effort has been described in [10].
Based on preliminary results of the AERO work, the Brighton Collaboration expressed
interest in partnering. As of November 2011, I am leading a Brighton working group which
goal is to support further development of the AERO and work towards using it to perform
automatic case classification of adverse event reports, as described above.
References
1. Iskander JK, Miller ER, Chen RT The role of the Vaccine Adverse Event Reporting system (VAERS) in
monitoring vaccine safety. Pediatr Ann 2004 Sep;33(9):599-606.
2. The safety of medicines in public health programs: pharmacovigilance, an essential tool,
http://www.who.int/medicines/areas/quality_safety/safety_efficacy/Pharmacovigilance_B.pdf
3. Medical Dictionary of Regulatory Activities. http://www.meddramsso.com/
4. Merrill G: The MedDRA paradox. AMIA Annual Fall Symposium 2008, 470-474.
5. Richesson R, Fung K, Krischer J: Heterogeneous but “standard” coding systems for adverse events: Issues
in achieving interoperability between apples and oranges. Contemp Clin Trials 2008, 29:635-645.
6. Bousquet C, Lagier G, Lillio–Le-Louet A, Le Beller C, Venot A, Jaulent M: Appraisal of the MedDRA
conceptual structure for Describing and Grouping Adverse Drug Reactions. Drug Safety 2005, 28:19-34.
7. Mozzicato P: Standardised MedDRA queries: their role in signal detection. Drug Safety 2007, 30:617-619.
8. Almenoff J, Tonning J, Gould A, Szarfman A, Hauben M, Ouellet-Hellstrom R, Ball R, Hornbuckle K, Walsh L,
Yee C, et al: Perspectives on the use of data mining in pharmaco-vigilance. Drug Safety 2005, 28:981-1007.
9. The Brighton Collaboration - http://www.brightoncollaboration.org.
10. Mélanie Courtot, Ryan R.Brinkman and Alan Ruttenberg.Towards an Adverse Event Representation ontology
Proceedings of the International Conference on Biomedical Ontologies (ICBO) 2011, http://icbo.buffalo.edu/