Stanford’s ChatEHR lets doctors talk to the chart — and it talks back

© khunkornStudio - stock.adobe.com

Stanford Health Care is testing a new artificial intelligence (AI) tool that could reshape how clinicians interact with electronic health records (EHRs), starting with a single prompt:

“Hi, 👋 I’m ChatEHR! Here to help you securely chat with the patient’s medical record.”

ChatEHR, developed by Stanford Medicine, uses large language model (LLM) technology to help physicians, nurses and other clinicians pull specific data, summarize charts and query a patient’s medical history in natural language.

The software, which began pilot testing this year with 33 Stanford clinicians, is integrated directly into the EHR and pulls information securely from patient files.

Built into the workflow

“AI can augment the practice of physicians and other health care providers, but it’s not helpful unless it’s embedded in their workflow and the information the algorithm is using is in a medical context,” said Nigam Shah, M.B.B.S., Ph.D., chief data science officer at Stanford Health Care. “ChatEHR is secure; it’s pulling directly from relevant medical data; and it’s built into the electronic medical record system, making it easy and accurate for clinical use.”

The project began in 2023, when Shah and a team including Anurang Revri, vice president and chief enterprise architect for Stanford Health Care, recognized the potential of LLMs for clinical utility. Rather than building a separate app, the developers opted to integrate ChatEHR directly into the existing Epic-based EHR to reduce friction.

“This is a unique instance of integrating LLM capabilities directly into clinicians’ practice and workflow,” said Michael Pfeffer, M.D., chief information and digital officer for Stanford Health Care. “We’re thrilled to bring this to the workforce at Stanford Health Care.”

Faster triage, fewer clicks

The early use cases are designed to reduce administrative burden and time spent hunting for information. Users can ask ChatEHR questions like, “Does this patient have any allergies?” or “What does their latest cholesterol test show?” or “Have they had a colonoscopy? Were the results normal?”

“Making the electronic medical record more user-friendly means physicians can spend less time scouring every nook and cranny of it for the information they need,” said Sneha Jain, M.D., a clinical assistant professor of medicine and early tester of the software. “ChatEHR can help them get that information up front so they can spend time on what matters — talking to patients and figuring out what’s going on.”

The tool is also being tested for use in high-pressure environments like the emergency department, where doctors need rapid access to medical histories.

“It’s not just the chest pain they’re having at that moment that matters — it’s their whole story,” said Jonathan Chen, M.D., Ph.D., a hospital system physician and assistant professor of medicine and biomedical data sciences. “Having ChatEHR boil that down into a relevant summary would make that process smoother.”

From summaries to smart automation

Beyond searching and summarizing, Stanford’s team is building automations that evaluate patient data and recommend actions. One automation, for example, assesses eligibility for transfer to Sequoia Hospital’s patient care unit.

“That automated evaluation saves us the administrative burden of sifting through patient information and helps us quickly determine if a patient can be transferred, opening access to care here at Stanford Hospital,” Shah said. Future automations could help identify hospice eligibility or flag high-risk post-surgical patients.

To ensure quality and safety, the team is evaluating ChatEHR using MedHELM — an open-source framework for assessing LLM performance in real-world medical settings. They’re also working to incorporate source citations, showing clinicians exactly where the tool pulled specific data within the chart.

Scaling up

The pilot phase will inform broader rollout plans. Shah said ChatEHR is being deployed in line with Stanford’s responsible AI guidelines, including accuracy testing, performance monitoring and user education.

“We’re rolling this out in accordance with our responsible AI guidelines, not only ensuring accuracy and performance, but making sure we have the educational resources and technical support available to make ChatEHR usable and useful to our workforce,” Shah said.

The software is not designed to offer clinical judgment or make decisions; it’s strictly an information-gathering tool. But if the pilot continues to prove effective, the implications for clinician efficiency and chart navigation could be significant, especially for physicians spending long hours navigating dense records or managing large patient volumes each day.