The paper, which was uploaded to open-access and preprint science hub arXiv, describes artificial intelligence models that would automate EHR note-taking for clinicians.
Specifically, Google is proposing a new language modeling task that could predict what physicians would write in their patient notes based on existing data from a patient’s medical record, such as demographics, labs, medications and past notes. Researchers used a publicly-available, de-identified dataset to train the generative models, and then compared the model’s automated notes against those by clinicians in the dataset for accuracy and completeness.
“We find that much of the content can be predicted, and that many common templates found in notes can be learned,” Google Brain Software Engineer Peter Liu writes in the paper. “Such models can be useful in supporting assistive note-writing features such as error-detection and auto-complete.”
The paper is only a pilot, and the researchers haven’t explored whether the models have improved clinical productivity. They suggest future work could involve combining EHR data with information from outside a patient’s medical record, such as imaging data or transcripts of patient-physician interactions.
To read the paper, click here.
More articles on EHRs:
Montgomery General Hospital signs 1st ‘Meditech-as-a-service’ deal for company’s cloud-based EHR
How one Mid-Atlantic health system put a Broadway spin on an Epic go-live
Healthcare facilities aren’t implementing ONC’s EHR safety guidelines, study finds
At the Becker's 11th Annual IT + Revenue Cycle Conference: The Future of AI & Digital Health, taking place September 14–17 in Chicago, healthcare executives and digital leaders from across the country will come together to explore how AI, interoperability, cybersecurity, and revenue cycle innovation are transforming care delivery, strengthening financial performance, and driving the next era of digital health. Apply for complimentary registration now.