Stanford Medicine has introduced ChatEHR, a groundbreaking artificial intelligence-based software that allows clinicians to interact with patient medical records using conversational queries. Developed by a team led by Nigam Shah, chief data science officer at Stanford Health Care, this tool aims to expedite tasks such as chart reviews and information retrieval by letting users ask direct questions about patient histories and receive immediate, contextually relevant responses pulled securely from electronic health records.
Currently in pilot at Stanford Hospital, ChatEHR is accessible to a select group of 33 healthcare professionals, including physicians, nurses, and nurse practitioners. The tool is designed to integrate seamlessly into existing electronic medical record workflows, making it easier for clinicians to gather essential details for patient care. Early users report that the system frees them from sifting through extensive records, instead providing rapid answers about allergies, test results, and care histories. The tool is information-only, not intended to render medical advice, and decisions remain in the hands of healthcare professionals.
ChatEHR is credited with enhancing efficiency in high-pressure situations like emergency admissions and patient transfers, where reviewing voluminous records can slow down care. The solution can also generate summaries of large document sets and perform ´automations´—evaluative tasks such as transfer eligibility decisions or highlighting candidates for specialized care. Development began in 2023, inspired by advances in large language models. Future plans include scaling ChatEHR hospital-wide, refining accuracy using open-source frameworks like MedHELM, and adding transparency features such as citing record sources for generated answers. The project aligns with Stanford Medicine´s responsible artificial intelligence guidelines, focusing on safety, educational resources, and robust technical support. The work is also supported by Stanford´s department of medicine and the Center for Biomedical Informatics Research.
