ChatGPT can connect to Epic health records: access, safeguards and limitations

OpenAI has introduced an integration that can bring authorised patient information from Epic electronic health record environments into ChatGPT for Healthcare. It has also launched a Healthcare Public Data plugin that connects structured information from official medical and government sources.

This is not a feature that lets any ChatGPT user search private health records. A participating healthcare organisation must configure the integration, existing permissions remain relevant and the EHR connection is not offered to individual accounts.

What the Epic integration can do

Clinicians can ask about changes since a patient’s previous visit, recent laboratory results, medication changes, specialist recommendations or unresolved follow-up work. OpenAI says ChatGPT gathers the relevant authorised record context, summarises developments and points back to supporting chart information.

The company describes two deployment patterns. Staff may bring EHR context into ChatGPT, or a supported healthcare deployment may place ChatGPT inside the EHR layout so the user can work without leaving the patient chart.

Nine official public-data connections

The Healthcare Public Data plugin connects to nine official sources. OpenAI specifically identifies ClinicalTrials.gov, CMS Coverage, RxNorm, DailyMed and PubMed. The aim is to work with precise records, identifiers, policy versions, drug labels and trial criteria instead of relying on an unsourced general answer.

For example, a research group could compare eligibility requirements for recruiting trials, while a pharmacy team could check current DailyMed label information. The integration may reduce the time required to gather records, but the resulting interpretation still needs qualified human review.

Who can access it

• ChatGPT for Healthcare customers can ask workspace administrators to enable the EHR integration and public-data plugin.
• ChatGPT Enterprise customers must confirm eligibility and configuration for a regulated workspace.
• Eligible individual clinicians in the United States can install the public-data plugin, but cannot use the EHR integration through an individual account.

OpenAI has not announced general consumer access or India-specific EHR availability in the September 1 release.

Security and evaluation claims

The healthcare workspace includes controls such as role-based access, single sign-on and audit logs. OpenAI says applicable customers can use a Business Associate Agreement for HIPAA-compliant workflows. Those controls do not remove the need for least-privilege access, local policy, monitoring and incident-response procedures.

OpenAI reports that physicians reviewed 4,363 ratings across 27 EHR-related clinical use cases and rated 99.1% of responses safe. In a separate evaluation using large US healthcare datasets, more than 93% of responses for each of five tested sources received “good” or better accuracy ratings. These are company-reported evaluations, not a guarantee that every answer is clinically correct.

The appropriate role is assistive: assembling authorised context and official evidence for professionals to review. It should not turn an AI summary into an automatic diagnosis, treatment decision or final patient record without verification.

Source: OpenAI healthcare integration announcement.

Official healthcare integration artwork. Image: OpenAI.