event-icon
Description

Common Data Models (CDMs) help address the portability barrier of electronic phenotyping algorithms. In this presentation, we share empirical knowledge of how interoperability takes more than what is currently defined in current state-of-the-art data models and how they can be enhanced, leveraged by our experience with manually annotating the free text of the eligibility criteria of 1,000 clinical trials. We discuss how those criteria, as a depiction of clinical researchers’ “real-world” phenotyping needs, highlight three types of difficulties in specifying a phenotype: domain uncertainty, granularity uncertainty, data table uncertainty; and we postulate that these challenges might be addressable by implementing heuristic conventions as a component of the data model.

Learning Objective: Learn how the definitions of a common data model can form a barrier to the development of an interoperable electronic phenotype, based on the experience of using the OHDSI OMOP CDM to annotate the eligibility criteria of 1,000 recent Phase 4 clinical trials from ClinicalTrials.gov.

Authors:

Fabricio Kury (Presenter)
Columbia University in the City of New York

Li-heng Fu, Columbia University in the City of New York
Chi Yuan, Nanjing University of Science and Technology
Ida Sim, University of California San Francisco
Simona Carini, University of California San Francisco
Chunhua Weng, Columbia University in the City of New York

Presentation Materials:

Tags