Developing an ontology centred on the PICO (patient, intervention, comparison, outcome) model to support the linkage of Cochrane evidence to promote its usability and discoverability

Article type
Authors
Beecher D1, Last A1
1Cochrane CET/ITS
Abstract
Background: Cochrane has vastly rich content and data stores that are not utilized to their full potential, locked away in data silos and static presentation formats. The Cochrane Linked Data Project is focused on the adoption of linked data technologies to provide the structured flexibility needed to support a more effective way for the finding, sharing and use of all Cochrane content. An ontology is one such linked data technology. An ontology defines a common vocabulary within a domain, a formal description of concepts; their properties and the relations among them. This provides a shared common understanding of a domain enabling the opening up of the data. Given the nature of the content and data, no existing vocabulary was a suitable reflection of the work of Cochrane, therefore the Cochrane Ontology was developed as part of the project.

Objectives: the construction of a custom ontology, specifically designed for the Cochrane domain, to enable the open production, publication, dissemination and usability of Cochrane data and content.

Methods: we chose the PICO model as the foundation for the Cochrane Ontology, owing to the intrinsic nature of the content and data. These four concepts; Patient, Population or Problem; Intervention; Comparison and Outcome therefore required description.

For reasons of standardization and interoperability, where applicable the creation of a new classification system should be based upon existing terminologies. The decision was to focus on the controlled vocabularies that form part of the Unified Medical Language System (UMLS). We assessed the individual concepts within each of the vocabularies against the Cochrane content and data in terms of suitability. In some instances the concepts fitted the linguistic needs of Cochrane without much alteration required, in other cases existing vocabularies needed to be merged and extended to provide the subject coverage required. We did not find a suitable classification for Outcomes, therefore we created a collaboration between Cochrane and Care Outcomes Measures in Effectiveness Trials (COMET) initiative to develop this vocabulary. This will be of value to the entire healthcare linked data community.

The PICO model provided the essential conceptual structure for the modelling of the multiple, independently developed vocabularies into a single coherent classification system.

Results: at present the Cochrane Ontology is used for PICO annotation of Cochrane Systematic Reviews, which in the future will be searchable via a PICO search tool (in development ). The use of the ontology for other projects within Cochrane is also under discussion, for example, to help structure data for Review Manager in future development phases of this tool.

Conclusions: an evaluation of the Cochrane Ontology is being carried out this year, the conclusions of which will be reported at the Santiago Colloquium.