Linguamatics NLP for City of Hope Registries - Srisairam Achuthan
Srisairam Achuthan described how collaborative teams at City of Hope (CoH) use NLP in research discovery to automate and improve the creation of disease registries, which are organized systems containing patient data from multiple sources, all related to a specific disease diagnosis. For use cases he drilled into details of the Multiple Myeloma Disease Registry and the Transplant Registry for Outcomes Research at CoH.
The current CoH practice required disease registrars to manually extract relevant information from multiple sources, which was time-consuming and not the best use of skilled staff time. CoH created an NLP workflow for automatic extraction of clinical and diagnostic information in a structured format from various unstructured reports associated with patients. This comprised a form builder and data capture and analytics/reporting applications, on top of a structured database, and provided the benefits of enhanced collaborations, self-service access to data, future-proofing, and a “place of truth”.
Srisairam described how NLP helps in automating the abstraction process for clinical attributes, especially in discretizing data elements embedded within unstructured text and in capturing manual abstraction workflows. He stressed the importance of identifying and analyzing data elements ahead of time. He then discussed in detail the 11 steps needed to create the NLP structured database from the multiple unstructured documents, and how CoH iteratively refined the Linguamatics NLP queries to maximize precision and recall in a training set and then test data.