What is it about?
Electronic health records contain a huge amount of useful information about a patient. Unfortunately, most modern machine learning methods struggle to make the most of this rich data source. We tweak ideas from natural language processing and produce models that are well suited to summarising a patients medical record and predicting what will happen on their care journey. We test this approach in the context of chronic kidney disease, to predict whether patients will experience end-stage kidney disease.
Featured Image
Photo by Europeana on Unsplash
Why is it important?
This work demonstrates the feasibility of using Large Language Models (LLMs) to support disease prediction. Additionally, it highlights two key unanswered questions about LLMs which should be considered when adapting such models to health record data.
Perspectives
Digital health is very much a team sport and it was a pleasure to collaborate and learn from colleagues in different disciplines.
Arrash Yassaee
Imperial College London
Read the Original
This page is a summary of: HEalthRecordBERT (HERBERT): leveraging transformers on electronic health records for chronic kidney disease risk stratification, ACM Transactions on Computing for Healthcare, July 2024, ACM (Association for Computing Machinery),
DOI: 10.1145/3665899.
You can read the full text:
Contributors
The following have contributed to this page
