UCLA researchers have developed an AI system that turns fragmented electronic health records (EHR) normally in tables into readable narratives, allowing artificial intelligence to make sense of complex patient histories and use these narratives to perform clinical decision support with high accuracy. The Multimodal Embedding Model for EHR (MEME) transforms tabular health data into "pseudonotes" that mirror clinical documentation, allowing AI models designed for text to analyze patient information more effectively.
Why it matters
Electronic health records contain vast amounts of patient information that could help doctors make faster, more accurate decisions in emergency situations. However, most cutting-edge AI models work with text, while hospital data is stored in complex tables with numbers, codes, and categories. This mismatch has prevented healthcare systems from fully leveraging advanced AI capabilities. Emergency departments, where quick decisions can be critical, particularly need tools that can rapidly process comprehensive patient histories to predict outcomes and guide treatment decisions.
What the study did
Researchers created a novel approach that converts tabular electronic health record data into text-based "pseudonotes" using medical documentation shortcuts commonly used by healthcare providers. On other words, instead of treating the EHR as a collection of codes, pseudonotes creates a story composed of multiple narratives. The system breaks patient data into concept-specific blocks (medications, triage vitals, diagnostics, etc.), transforming each into text using simple templates, and then encodes each one separately using language models. It essentially emulates a form of medical reasoning.
They then fed this text to advanced language models, treating different types of health information, like lab results, diagnoses, and medications, as separate but related data streams. The team tested their system against traditional machine learning methods, specialized healthcare AI models, and prompting-based approaches using real emergency department prediction tasks.
What they found
Across over 1.3 million emergency room visits from the Medical Information Mart for Intensive Care (MIMIC) database and UCLA datasets, MEME consistently outperformed existing approaches across multiple emergency department decision support tasks. The multimodal text approach, which processes different components of health records separately, achieved better results than trying to combine all information into a single representation. The system demonstrated superior performance compared to traditional machine learning techniques, EHR-specific foundation models like CLMBR and Clinical Longformer, and standard prompting methods. The approach also showed good portability across different hospital systems and coding standards.
What's next
The research team plans to test MEME's effectiveness in other clinical settings beyond emergency departments to validate its broader applicability. They also aim to address limitations observed in cross-site model generalizability, working to ensure the system performs consistently across different healthcare institutions. Future work will focus on extending the approach to accommodate new medical concepts and evolving healthcare data standards, potentially making advanced AI more accessible to healthcare systems.
From the experts
"This bridges a critical gap between the most powerful AI models available today and the complex reality of healthcare data," said Simon Lee, PhD student at UCLA Computational Medicine. "By converting hospital records into a format that advanced language models can understand, we're unlocking capabilities that were previously inaccessible to healthcare providers. The fact that this approach is more portable and adaptable than existing healthcare AI systems could make it particularly valuable for institutions working with different data standards."
About the study Article:
About the Research Team Simon A. Lee (Department of Computational Medicine, UCLA), Sujay Jain (Department of Electrical and Computer Engineering, UCLA), Alex Chen (Department of Computational Medicine, UCLA), Kyoka Ono (Department of Statistics and Data Science, UCLA; Department of Natural Sciences, International Christian University, Mitaka, Tokyo, Japan), Akos Rudas (Department of Computational Medicine, UCLA), Jennifer Fang (LA Health Services, Enterprise Clinical Informatics; Harbor-UCLA Medical Center; UCLA Department of Emergency Medicine), and Jeffrey N. Chiang (Department of Computational Medicine, UCLA; Department of Neurosurgery, UCLA).
