Build a “Cochrane ontology” that defines key components of Cochrane content and their relationships with one another
This is an essential step in conceptualising and formulating a Cochrane metadata and semantic store. However, it has the potential to provide other significant benefits.
- Provide a clear reference point for communicating the Cochrane linked data work to both internal and external stakeholders.
- Allow us to develop structured mechanisms for others who might wish to interact with or license our data, or link to it from their own programmes and products.
- Provide a common framework, shared across the Collaboration for use in designing tools and processes to improve efficiency in review production and authoring.
- Provide leadership - A number of initiatives currently underway are attempting to define models (ontologies) for closely related areas, but no one to date has come forward with a semantic model for Systematic Reviews, for their key components such as risk of bias assessments or for the relationships between reviews, the trials they contain, and the multiple reports in the literature that may be associated with each trial. A Cochrane model that did so would stand a good chance of becoming a Web 3.0 standard approach.
Project Activities to date
Three first phase ontologies are being developed
- Review model: just enough to represent the hooks for annotaing a review
- PICO model for capturing the domain of PICO
- Annotation model for associating PICO with a review
See the PICO Ontology section for a diagram of these 3 models and how they relate to one another.
The PICO model
We have identified 3 different sorts of PICOs within reviews and we plan to build our models so that all 3 can be annotated. They are:
- The Review (or study selection criteria) PICO. This is documented in the Methods:Criteria for considering studies for this review section of the review
- The Comparison PICOs - These are taken from the analyses in the Review - initially from the forest plots, although this may be extended at some point in the future
- The Included Study PICOs - with each included study having its own set of PICOs
The Annotation Model
This model will allow us to track details of the annotation process - who did the annotation, when, what was it based on. The latter point is important for Study PICOs because the view of a study within a single review may be incomplete (e.g. may only include some arms of a multi-arm trial or selected subsets of trial populations).