Courtney Irwin (MetadataWorks) will present two recent case studies utilising natural language processing methods to improve the utility and interoperability of clinical data for the National Health Service in the UK.
MetadataWorks, a UK based startup, aims to support organisations to manage and make better use of their data assets through their data catalogue product. MetadataWorks has focused on the complex domain of clinical data, working with health organisations to catalogue data assets, develop and promote data standards, and improve interoperability.
Recently, MetadataWorks demonstrated the use of a text similarity algorithm as a potential replacement for the manual mapping process required to transition between pathology coding standards. Working with NHS Digital, the information and technology partner to the UK health and social care system, MetadataWorks reviewed the results of the algorithm with pathology specialists, demonstrating that the algorithm successfully identified 96.5% of the approved code matches. The process and results were approved by clinical and technical experts in stakeholder consultations ahead of a planned coding standards transition to be rolled out to general practitioners, hospitals, and laboratories across the UK.
Following this, MetadataWorks extended the research and application of this work with a grant from InnovateUK, the UK’s innovation agency, to support the automation of metadata documentation. Utilising similar methods as above, and extending the research by experimenting with trained clinical word embeddings, MetadataWorks are building a prototype to suggest matches between clinical dataset elements and common clinical data models and standards.
The talk will explore these two use cases, and the benefits and limitations of these algorithms to improve clinical data management at a national scale, to promote standardization, increasing data interoperability, and supporting health research that improves peoples lives.