OpenAI announced the integration of ChatGPT Health with Epic, one of the leading electronic health record (EHR) systems in the United States, which hosts data on more than 325 million patients. The move allows clinicians to review patient information through ChatGPT and ask questions about it, rather than interacting with the health record solely through its traditional interface.
The integration includes appointment notes, test results, medication information, and specialist documentation. The system can use this data to summarize the record, review changes in medical history, create a clinical timeline, and help the physician prepare for upcoming appointments. In some systems, ChatGPT can be integrated directly into the electronic record workflow, allowing an initial assessment to be conducted before the appointment without leaving the patient’s file.
Read-Only Integration
OpenAI says the integration gives ChatGPT permission to read patient records only and does not allow it to write any data to health records. This restriction establishes clear limits on the tool’s role within the healthcare environment, as its stated function is limited to review, analysis, and assistance with preparation, without directly modifying the medical record.
At the same time, the company launched the Healthcare Public Data app for healthcare workers. The app can retrieve information from ClinicalTrials.gov, CMS Coverage, RxNorm, DailyMed, and PubMed. According to OpenAI, these sources can be used to assess clinical trial eligibility criteria, analyze medication information, review reimbursement policies, and compile data related to healthcare providers.
What Changes in Practice?
The announcement does not merely add a new source of patient data; it places conversational models within a workflow that physicians rely on daily. The practical value lies in reducing the time needed to search through a lengthy record and connect disparate elements such as laboratory results, medications, and clinical notes. However, the source does not specify the institutions where the integration has begun, clarify the scope of its availability, or explain the mechanisms for human review for each use.
OpenAI explained that organizations that sign a business associate agreement (BAA) can use ChatGPT Work, Codex, apps, and connectivity within a workflow compliant with health data requirements. It also emphasized that ChatGPT should not be used for diagnosis or treatment.
Safety Remains a Critical Issue
The expansion follows OpenAI’s announcement that it had made ChatGPT Health available to all consumers in the United States during the previous month, saying that users ask approximately 300 million health-related questions each week. The company also said it had collected feedback from more than 4,300 physicians across 27 clinical use-case scenarios, and that 99.1% of the answers were considered safe.
These figures come from OpenAI and do not eliminate open questions about the accuracy of the answers or the safety of relying on them for medical decisions. The source also points to the continuing debate over the use of artificial intelligence in healthcare and to lawsuits facing the company related to health advice provided by ChatGPT. Therefore, despite the greater breadth of data it can read, the integration remains an assistive tool rather than a replacement for medical assessment and clinical responsibility.