Developing an Ontology for Randomised Controlled Trials Henry Potts & Sanjay Modgil Design-a-Trial team, University College London
Randomised controlled trials ‘Gold standard’ in clinical research Units randomly assigned to groups, in which intervention(s) are compared to a control and an outcome measured Not restricted to medicine (agricultural products, advertising, archaeological conservation)
Design-a-Trial http://www.design-a-trial.net Decision support software to aid inexperienced clinicians write a protocol for a randomised controlled trial (RCT) Critiquing, pre-emption, natural language generation, ... Back-end: Prolog; user interface: Visual Basic http://www.design-a-trial.net
An ontology A structure to encode data about RCTs Used as basis for Design-a-Trial v.2 [Sanjay’s talk] Propose as an Interlingua for other trial software
History EON/GLIF-2 for guidelines, not trials BreastCancerProto.pont ProtegeWin example concentrating on procedural aspects Our RCT ontology Ongoing work by Tu, Noy etc.
Trial_Model_Entity GLIF-2 etc. are bipartite in structure Guideline_Model_Entity (procedural) Reference_information Our ontology is tripartite add Trial_Model_Entity Allows representation of trial-level concepts Trial architecture, Outcomes, Interventions, Randomisation, ...
Multiple superclasses Protégé allows multiple superclasses links different levels of representation crossing Guideline_Model_Entity vs. Trial_Model_Entity divide ties Outcome_measure to an Action_Step Thanks to... Samson Tu for this paradigm shift
Guideline_Model_Entity changes GLIF uses an ordinal time structure Introduce temporal referents Composite steps
Reference_information changes Activity Compliance_procedure Consent_procedure Structure of Intervention_procedure Outcome_variable etc.
Conclusions Copy of (mostly) documented ontology available Aim to make it downloadable from the web site and write up for publication DaT going into commercial development