7th October 2021 - 2pm - Martin Chapman

Title: Phenoflow - A Microservice Architecture for Portable Workflow-based Phenotype Definitions

Slides are available for download below:

Slides from Martin Chapman: Phenoflow (1.68 MB PDF)

Abstract: The use of EHR-based phenotype definitions to extract patient cohorts of interest for research is now commonplace. To reduce overhead in the identification of a cohort from a new dataset, we want to be able to reuse definitions for certain conditions as much as possible. However, several barriers to reuse exist, such as the lack of a common standard for structuring definitions and little provision for deriving standard computable forms of a definition to be executed against a dataset in practice. 

In this talk, I will discuss our work on Phenoflow, a standard definition model that supports a wide range of definition types (from code-based to NLP-based phenotypes), and an associated authoring architecture, which parses definitions from a wide range of sources, restructures them according to the Phenoflow model, and allows customisable, executable forms of each definition to be generated. I shall discuss our initial work in standardising definitions and evaluating the model, and our plans for future work, including how our standard model can support important tasks such as definition validation.

Speaker: Dr Martin Chapman, Research Associate, King's College London

The meeting was held online via Zoom.