Biomedical terminologies and ontologies are increasingly used to annotate patient data that are generated in the course of the clinical care process, for example in the context of an Electronic Health Record. The purpose is to make these data comparable and interoperable across caregivers, institutions and systems with the goal of providing better care on the one hand, and of advancing biomedical science on the other. Where data pertain to individuals and are thus specific, ontologies capture what is generic. For an ontology or terminology to maintain its usefulness, it must be updated at regular intervals. Updates occur because of changes in both the domain itself and in the understanding of scientists and clinicians. A serious drawback of current approaches to ontology versioning is that the changes made in new releases are either not documented, or information is limited as to which entries in the ontology appeared, disappeared, or became fused or split. Only in very rare cases is information provided about the reasons for the changes made. This makes backward compatibility of legacy annotations difficult, if not impossible at all. The long-term vision behind the work proposed is one in which biomedical ontologies do not just reflect the state of the art in biomedical science in terms of what entities exist in reality and of how they are related, but that they also keep track of whether the changes introduced in successive versions of ontologies reflect (1) changes in the underlying reality, (2) in the views of ontology authors - or in associated scientific theories, or (3) are corrections of editorial mistakes. The hypothesis is that such an approach to ontology versioning can serve as a means to measure quality improvements in successive versions of an ontology in an objective manner, which will also help to ensure an incremental improvement in terminologies and thus in clinical data. The project will test this hypothesis by applying this novel approach to ontology versioning to the Systematized Nomenclature of Medicine (SNOMED) according to the following plan: . a detailed analysis of SNOMED-CT's existing history mechanism, covering all versions that have been released since its inception, with the objective of detecting those ways in which this mechanism falls short with respect to ontology versioning principles (aim1), . development of a prototype in which these principles are implemented together with a metric for measuring quality improvements in successive versions and that can serve as a plug-in for ontology authoring systems such as Prot?g?, ODE or SWOOP (aim2) . use of this prototype to reorganize a representative sample of SNOMED-CT's history information in line with the principles of ontology versioning (aim3), and . computation of the quality improvement of SNOMED-CT over time in order to demonstrate the usefulness of the approach and foster its acceptance in other ontologies (aim 4).

Public Health Relevance

High quality biomedical terminologies and ontologies are important instruments for the integration and analysis of biomedical data about anything from molecules to patient populations as they enable the results of such analyses to serve improvements in patient care and advances in the state of the art in diagnosis and treatment. Maintaining the quality of such instruments is accordingly a crucial task, but one that is still difficult to carry out because objective methods for measuring quality are still lacking. We propose a metric for quality assurance of biomedical terminologies and ontologies and a strategy for its implementation that is designed to bring about an enhancement in our ability to reuse data for new research purposes, an improved organization of research, and opportunities for new kinds of information-based research cooperation and clinical care.

Agency
National Institute of Health (NIH)
Institute
National Library of Medicine (NLM)
Type
Exploratory/Developmental Grants (R21)
Project #
1R21LM009824-01A1
Application #
7660638
Study Section
Special Emphasis Panel (ZLM1-ZH-C (J2))
Program Officer
Ye, Jane
Project Start
2009-03-31
Project End
2011-03-30
Budget Start
2009-03-31
Budget End
2011-03-30
Support Year
1
Fiscal Year
2009
Total Cost
$285,300
Indirect Cost
Name
State University of New York at Buffalo
Department
Type
Organized Research Units
DUNS #
038633251
City
Buffalo
State
NY
Country
United States
Zip Code
14260
Ceusters, Werner; Bona, Jonathan P (2016) Analyzing SNOMED CT's Historical Data: Pitfalls and Possibilities. AMIA Annu Symp Proc 2016:361-370
Ceusters, Werner (2011) SNOMED CT's RF2: Is the future bright? Stud Health Technol Inform 169:829-33
Ceusters, W; Capolupo, M; de Moor, G et al. (2011) An evolutionary approach to realism-based adverse event representations. Methods Inf Med 50:62-73
Ceusters, Werner (2011) SNOMED CT revisions and coded data repositories: when to upgrade? AMIA Annu Symp Proc 2011:197-206
Ceusters, Werner; Smith, Barry (2010) A unified framework for biomedical terminologies and ontologies. Stud Health Technol Inform 160:1050-4