Article type
Year
Abstract
Background: the Cochrane PICO ontology provides a framework for organizing information about trials of interventions that is valuable both in reporting studies and in linking evidence across studies. PICO stands for the four components: population, intervention, comparison and outcome. Each of these components is linked to lower-level entities, which in turn require specification. Here we describe the development of three of these lower-level ontologies in the specific context of developing a behavioir change intervention ontology in the Human Behaviour Change Project (HBCP).
Objectives: to develop ontologies for populations, settings and modes of delivery.
Methods: for each ontology, we conducted the following steps:
1) development of the definition and preliminary classification by reviewing existing classification systems (top-down) and reviewing 100 published behaviour change intervention reports (bottom-up);
2) refinement of the intervention ontology by trialling the preliminary classification on 100 behaviour change reports;
3) stakeholder review by behavioural science and public health experts to further refine the ontology;
4) inter-rater reliability testing of coding using the ontology by two independent coders;
5) specifying relationships within the ontology.
Results: the resulting population ontology has 329 entities with six upper-level entities (sociodemographic characteristics, socioeconomic factors, social roles, health status, behaviour and psychological status) and six sub-levels, while the setting ontology has 70 entities with three upper-level entities (geographical location, attribute of location and iIntervention site) and four sub-levels, and the modes of delivery ontology has 86 entities across three upper levels (informational, somatic and cross-cutting) and five sub-levels.
Conclusions: the ontologies developed add to the organisational framework provided by PICO and can be used in synthesising evidence across trials. While the population and setting ontologies have general applicability, the mode of delivery ontology is more specific to behaviour change trials. These ontologies are currently being used in HBCP as a basis for machine learning to extract information from behaviour change intervention trial reports.
Patient or healthcare consumer involvement: when the researchers and expert stakeholders reach agreement on the first version of the ontologies, patients and members of the public will be invited to comment on their clarity.
Objectives: to develop ontologies for populations, settings and modes of delivery.
Methods: for each ontology, we conducted the following steps:
1) development of the definition and preliminary classification by reviewing existing classification systems (top-down) and reviewing 100 published behaviour change intervention reports (bottom-up);
2) refinement of the intervention ontology by trialling the preliminary classification on 100 behaviour change reports;
3) stakeholder review by behavioural science and public health experts to further refine the ontology;
4) inter-rater reliability testing of coding using the ontology by two independent coders;
5) specifying relationships within the ontology.
Results: the resulting population ontology has 329 entities with six upper-level entities (sociodemographic characteristics, socioeconomic factors, social roles, health status, behaviour and psychological status) and six sub-levels, while the setting ontology has 70 entities with three upper-level entities (geographical location, attribute of location and iIntervention site) and four sub-levels, and the modes of delivery ontology has 86 entities across three upper levels (informational, somatic and cross-cutting) and five sub-levels.
Conclusions: the ontologies developed add to the organisational framework provided by PICO and can be used in synthesising evidence across trials. While the population and setting ontologies have general applicability, the mode of delivery ontology is more specific to behaviour change trials. These ontologies are currently being used in HBCP as a basis for machine learning to extract information from behaviour change intervention trial reports.
Patient or healthcare consumer involvement: when the researchers and expert stakeholders reach agreement on the first version of the ontologies, patients and members of the public will be invited to comment on their clarity.