Authors: Chao Pang, Xinzhuo Jiang, Nishanth Parameshwar Pavinkurve, Krishna S. Kalluri, Elise L. Minto, Jason Patterson, Linying Zhang, George Hripcsak, Noémie Elhadad, Karthik Natarajan

Published on: February 06, 2024

Impact Score: 8.27

Arxiv code: Arxiv:2402.04400

Summary

• What is new: Use of GPT model combined with CEHR-BERT for generating synthetic EHR sequences in OMOP format.
• Why this is important: Existing synthetic EHR data generation methods neglect temporal dependencies in patient histories, limiting realism and applicability.
• What the research proposes: Training a Generative Pre-trained Transformer (GPT) model with a novel patient representation from CEHR-BERT to generate temporally coherent patient sequences.
• Results: Generated patient sequences can be accurately converted to OMOP format, enhancing the realism and utility of synthetic EHR for research.

Technical Details

Technological frameworks used: GPT with CEHR-BERT patient representation

Models used: Generative Pre-trained Transformers, CEHR-BERT

Data used: EHR data converted to patient sequences

Potential Impact

Healthcare data analytics firms, EHR software providers, medical research institutions

Want to implement this idea in a business?

We have generated a startup concept here: HealthGPT.