OpenAI has announced that electronic health records (EHRs) can now be connected to ChatGPT for Healthcare. Epic and a Healthcare Public Data plugin have been integrated with the chatbot, allowing authorized patient data to be accessed within ChatGPT for Healthcare.  

In the announcement, OpenAI explained that with this new connection, clinicians can query a patient record to find answers to specific questions, such as those related to recent medication changes, lab results, or unresolved issues. The company noted that ChatGPT for Healthcare can also point to evidence within the chart that supports its answer.  

Within supported deployments, ChatGPT for Healthcare can also be integrated directly into an EHR environment, allowing clinicians to access AI-assisted workflows within the patient record itself.  

Nine official public health-care data sets will also be available within the AI platform, including ClinicalTrials.gov, CMS Coverage, RxNorm, DailyMed, and PubMed, bringing information from trusted sources together in one place. OpenAI explained that clinicians can use these resources to check and verify clinical trial eligibility criteria, medication warning labels, and coverage policy information.  

“As a pilot partner, we’re exploring how the new EHR integration with ChatGPT for Healthcare can help clinical teams understand what has changed and what matters most across a complex patient record. By bringing relevant information together more quickly and comprehensively, the technology has the potential to reduce time spent synthesizing data and give clinicians more time with patients. We’re also engaging frontline teams to validate these capabilities in practice and help shape where they can add the most value,” stated Suresh Gunasekaran, President and CEO of UCSF Health.  

However, Epic noted in a post on X that the connection between Epic and ChatGPT was built using freely available FHIR application programming interfaces.  

Physician Evaluation 

OpenAI stressed that it partners with hundreds of physicians to evaluate and help improve health-related outputs from ChatGPT. To date, physicians from 60 countries and 26 medical specialties have reviewed more than 700,000 model responses, providing feedback aimed at improving model performance and health-care–related tools.  

For the EHR connection, physicians evaluated responses across 27 clinical use cases, ranging from visit preparation summaries, imaging reviews, patient education materials, prior authorization documentation, and clinical risk score calculations to diagnostic guidance review. Across 4,363 ratings covering all 27 clinical use cases, an average 99.1% of responses were rated as safe. Medication history and safety review had the lowest safety rating at 97.9% across 284 tasks tested. Many use cases achieved a 100% safety rating.  

Physicians also reviewed hundreds of chatbot responses to complex clinical questions in a two-round evaluation. They assessed the accuracy of responses drawing on five public health-care data sets: DailyMed, PubMed, ClinicalTrials.gov, RxNorm, and CMS Coverage. More than 93% of responses were rated as having “good” or better accuracy. CMS Coverage had the lowest accuracy rating, at 93.2%, compared with the highest rating of 98.6% for DailyMed.  

The Business of Health Care 

ChatGPT has also added plugins that allow the platform to connect with systems such as Microsoft SharePoint, Google Drive, Salesforce, and Slack to assist health-care professionals with administrative and operational tasks. Through these connections, users can draw on research and operational information to create reports, analyses, and presentations.  

Health-care systems can also access ChatGPT Work and Codex through a governed workspace designed to support broader organizational AI needs. 

“At AdventHealth, the value of AI starts with our people. By putting tools like ChatGPT Work, Codex, and connected business data in the hands of our team members, AI can reduce routine work and help practical innovations move more quickly into action. The goal is simple: give caregivers and teams more time for the human connection that enables our connected, whole-person care,” said Robert Purinton, Chief AI Officer at AdventHealth.  

ASCO AI in Oncology is published by Conexiant under a license arrangement with the American Society of Clinical Oncology, Inc. (ASCO^®). The ideas and opinions expressed in ASCO AI in Oncology do not necessarily reflect those of Conexiant or ASCO. For more information, see Policies.