How Care-Connect and Spryfox use NLP in Making Patient-level Decisions

In the absence of a fully integrated electronic health record system, documents are an important source of health economic and clinical information. For example, hospital discharge summaries contain useful health resource utilization information. Healthcare NLP tools are used to extract information from these unstructured documents, but reaching regulatory-grade accuracy often requires augmenting them with human-in-the-loop workflows.

We propose a framework for quality assurance for such systems, using clinician developed logic to filter and rank entities to be manually reviewed – thereby reducing workload while maintaining a high level of safety. The framework has been implemented on top of John Snow Labs’ Healthcare NLP & LLM models and applied in real-world clinical data abstraction projects for synthetic control and risk stratification

About the speakers

Hoifung Poon

Christian Debes

Co-Founder, Head of Data Analytics & AI at Spryfox GmbH

Dr. Debes is co-founder and head of data analytics and AI at Spryfox. He has a PhD in machine learning (from 2010) and 13 years of experience in industrial data science. His current focus is on application of AI in health, including the use of NLP. Besides his work at Spryfox, Dr. Debes is also lecturer for data science at TU Darmstadt. Before his time at Spryfox he had leading positions at Merck and AGT where he build and successfully managed larger data science teams.

Hoifung Poon

Fiona Kiernan

Chief Economist at Care-Connect

Dr Fiona Kiernan is Chief Economist with Care-Connect, a managed care provider for patients with chronic conditions. She is a unique hybrid of an economist and clinician. Along with her medical degree from University College Dublin, she also holds a Masters in Health Economics from the London School of Economics and Political Science, and a PhD in Economics from University College Dublin. She designs risk stratification tools and value-based pricing models using data from multiple clinical sources including electronic health records.



Sessions: April 2nd – 3rd 2024
Trainings: April 15th – 19th 2024


Presented by